NEW ECONOMIES

Timestamps

(0:00) Meet Dylan Abruscato
(2:20) Why OpenAI acquired TBPN
(6:48) What changed after the acquisition
(7:43) Will frontier models come on the show?
(8:17) What makes TBPN so vibrant
(9:32) Prepping for a daily live show
(10:46) Why live streaming is hot
(12:04) Are niche audiences more valuable?
(12:58) How the ad integrations work
(15:44) Calling partners post-acquisition
(17:10) Building alongside legacy media
(18:17) The copycats
(19:38) Will another TBPN-style acquisition happen?
(20:08) OpenAI’s marketing services
(21:02) Thinking about distribution channels
(22:10) Dream guests
(22:59) TBPN’s next 12 months
(23:22) Audience Q&A

Lessons from this episode with Dylan

1. The OpenAI deal was for the team, not the show
Dylan says most people misread why OpenAI bought a live tech talk show.

“I think what a lot of folks got wrong about the deal is that it had very little to do with the show itself and much more to do with the interest in us and the brains behind the operation.”

Many of TBPN’s 30 sponsorship deals included marketing and comms advisory work, a tactic the team first used to win sponsors. OpenAI’s offer was simple: keep doing the show, wind down the advisory clients, and make OpenAI the only one.

TBPN was profitable, had raised no money and had no plans to sell. Fidji Simo reached out the morning after the Super Bowl.

2. Build the show you’d want to watch
The two stunts that came right before the deal both started as ideas the team found funny.

“The truth is we’re just building the show that we always wanted to watch.”

The first was a hyper-local 15-second ad that aired in San Jose at noon but still counted as a Super Bowl spot. It spelled out TBPN using over a thousand logos of past guests, and those founders shared it everywhere. The second was claudewithads.com, a reskinned Claude that gave one real answer and then rotated through sponsor ad reads.

The team’s brainstorming rule is that if an idea makes them laugh, they run with it.

3. Live is the last unskippable ad format
Dylan has worked in live TV at SNL, CNBC and HQ Trivia, and has argued for live for a decade.

“It’s a truly unskippable ad format, and that’s kind of allowed us to command such high prices.”

Almost nothing is watched live anymore. Sports is the exception, and even award shows are now consumed as recaps on X. TBPN’s audience is smaller than HQ’s was, but advertisers value it more because no one can fast-forward.

4. Go intentionally niche
TBPN never explains the jargon. Where CNBC would spell out what AWS stands for, John and Jordi just say AWS or GPU, and the audience respects it.

“It really only takes one or two founders watching the show to switch to this platform for the advertisement to pay for itself.”

A small audience of founders, CFOs and VCs is worth more to a tech sponsor than a mass audience. As an example, John once read a Ramp ad straight to Mark Zuckerberg as he took his seat.

5. Copy Formula 1’s sponsorship model
TBPN sold only year-long deals, split into a couple dozen categories with one sponsor per category.

“Everyone got 100% share of voice in their category on everything that we did over the course of the year.”

Sponsors got the same ad read in every episode, plus the ticker, merch, newsletter links and end cards on tweets. Dylan’s view is that repetition is what counts when a founder is finally ready to switch products. TBPN also never tried to upsell new placements partway through the year.

The commercial mindset was there from day one. On the first episode, with one viewer, John did fake ad reads for sponsors the show didn’t have. When TBPN wound down its sponsors after the acquisition, viewers asked for the ad reads back.

6. Chemistry plus a fixed format
The show depends on the friendship between its two hosts.

“It’s like a true one-of-one friendship, and it’s just so relatable.”

John and Jordi meet at the gym at 6 a.m. every day and work out the day’s lead story in the sauna. They go live at 11, and after the show ends they still call each other on the drive home.

The format never changes:

Hour one: John and Jordi on the current thing, then live reactions to the timeline.
Hour two: one or two tier-one interviews.
Hour three: a lightning round of startup fundraising news.

Most guests are booked last minute. Mag Seven–level names take six months or more of back-and-forth.

7. Be a companion to legacy media, not a competitor
TBPN never tried to beat traditional outlets at breaking news.

“We always viewed TBPN as a companion show.”

When founders or publicists offered scoops, TBPN sent them to the Journal or Bloomberg. The founder would come on a few hours later to break the story down. Every show opens with a hard copy of the Journal.

Dylan doesn’t want copycats making another three-hour daily show. He would rather see new formats in tech media, like a late-night tech show or an interactive Shark Tank.

Where to connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Dylan on X: https://x.com/DylanAbruscato

Watch and follow TBPN: https://www.tbpn.com

Previous episodes include

Our partners:

Harmonic: Your go-to startup database. Visit Harmonic to learn more.

Prelint: Keeps every decision reviewed, recorded, and traceable. Head over to Prelint to learn more.

VCX: The public ticker for private tech. Head over to GetVCX.com to learn more.

Eric Ries

Ollie Forsyth — Fri, 16 Oct 2026 12:53:53 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Eric Ries, creator of The Lean Startup and author of Incorruptible, joins us to explain why “financial gravity” pulls even the best companies away from their mission. He explains why the standard best practices founders are handed “are designed to make you lose control of your company.”

Eric published The Lean Startup 15 years ago and went on to found the Long-Term Stock Exchange. He co-founded the AI research lab Answer.AI and helped start Virgil, an AI-enabled law firm for early-stage companies. In his latest book, Incorruptible, he studies the companies that have resisted this force, from Costco to Novo Nordisk to Anthropic.

In our in-depth episode, we discuss:

How the default “any lawful act or activity” clause in a Delaware charter quietly commits founders to maximizing shareholder value, and why lawyers rarely tell them
The story of Sol Price, FedMart and the origins of Costco, and the “governance fortress” that has protected Costco’s $1.50 hot dog for 40 years
The three mission protective provisions every founder can set up in a matter of weeks: becoming a PBC, adopting a director’s oath, and appointing a mission guardian
The difference between a mission statement and a mission, and what Silicon Valley Bank’s charter revealed after its collapse
Why “it’s always too early until it’s too late,” how to win over the roughly 25 people who must agree to a change, and why Anthropic’s Long-Term Benefit Trust has become a commercial advantage

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Eric Ries

403KB ∙ PDF file

Timestamps

(0:00) Meet Eric Ries
(2:14) Eric’s New Book
(3:30) How Startups Are Built Today
(6:20) Why Companies Self-Destruct
(16:37) Who’s to Blame When Companies Fail
(21:25) The Blueprint for a High-Performing Company
(26:52) How Companies Protect Themselves
(33:13) Building Company Ethos and Structure
(46:05) Why Companies Get Bullied
(50:32) When Is It Too Late to Save a Company?
(55:15) The Company Eric Would Have Saved
(1:00:05) What Eric Is Excited About Next

Lessons from this episode with Eric

1. Why great companies are destroyed by financial gravity
Eric explains why blaming a greedy investor or a reckless founder misses the real cause of most corporate collapses.

“There is a form of what I call financial gravity that operates on these companies. The ones that collapse, the ones that erupt in scandal, whether I’m talking about Boeing or FTX or Wells Fargo or Silicon Valley Bank, they’re all the same story.”

He compares it to a submarine with a weak hull. Asking whether the water is to blame is the wrong question. The right question is who designed the submarine.
Gravity acts on every company, yet some have lasted 50, 75 or even 500 years. Studying the ones that never collapse is how you learn to build differently.

2. How the firing of Sol Price led to the founding of Costco
Sol Price capped his markups at 14% and treated customers as clients he owed a fiduciary duty to. His investors fought him for 20 years, then changed the locks on his office overnight.

“It is so much easier to destroy than to create. They managed to destroy in seven years what had taken Sol more than 20 to build.”
Within seven years of firing him, the investors had completely liquidated FedMart. Two weeks after his dismissal, Sol leased the office upstairs and started the Price Club, which later merged into Costco.
Costco has survived for 40 years because of what Eric calls a governance fortress. That structure lets its directors act as a bulwark against Wall Street pressure.

3. The six words in your charter that hand your company to shareholders
Most founders have never read the first sentence of their own corporate charter. Eric says that sentence decides who the company really serves.

“Since 1986 in the United States, whenever you see ‘any lawful act or activity,’ the courts have decided that actually means maximize shareholder value.”
The phrase sounds open-ended. Legally, it makes your company a financial instrument whose job is to enrich shareholders.
The same duty forced Twitter’s board to sue Elon to complete his deal. It also let Philip Morris take over an inhaled therapeutics company and destroy it.

4. Three mission protective provisions that take weeks, not years
Eric gives founders a short list of protections they can put in place early. Some take only days.

“Welcome to best practices. They are designed to make you lose control of your company by force or by corruption.”
First, become a public benefit corporation, a two-page filing in Delaware. Second, require every director to take an oath to protect the mission, much like the Hippocratic oath. Third, set up a mission guardian, such as a foundation or a purpose trust.
Companies owned by an industrial foundation, such as Novo Nordisk, are five times more likely to reach year 50. Eric’s explanation: “Checks and balances work.”

5. A mission statement is not a mission
Eric’s own deposits were at Silicon Valley Bank when it collapsed. Afterwards, he looked up what the bank had legally committed to.

“They had no mission at all. The whole thing was a scam.”
The bank’s mission statement promised to move the innovation economy forward. Its charter said “any lawful act or activity.” Even if its interest rate bets had paid off, they would have done nothing for customers.
His test: Google spends billions to make sure its quarterly reports are on time, while “Don’t be evil” was one sentence in the employee handbook. What a company invests in shows what it values.

6. It’s always too early until it’s too late
Lawyers and investors often tell founders to deal with governance after product-market fit. Eric says that advice really means never.

“In most companies, there are at most 25 people that need to consent to anything like this.”
Veeva Systems became a public benefit corporation after it was already a $40 billion public company. It spent a year on investor relations and won 90% of the shareholder vote.
Eric’s line for pitching a change to investors: “It is my judgment as the founder of this company that it is essential to our strategy that we win the trust of [some group].” Framing it this way forces people to say exactly what they object to.

7. Why ethics is an accelerant
Eric uses Anthropic to show that integrity can be a source of commercial strength.

“People talk about ethics like it’s an impediment to business success, but here it is clearly an accelerant.”
Anthropic is governed by the Long-Term Benefit Trust, which appoints directors to its for-profit board. Eric argues this structure is a big part of why enterprises trust it more than its competitors.
Costco, Patagonia, Vanguard, Mondragon and Anthropic have different missions and cultures. The one thing they share is that each rejected the standard best practices for how a company should be structured.

Where to connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Eric on X: https://x.com/ericries

Eric's new book - Incorruptible: https://www.incorruptible.co

Previous episodes include

Our partners:

Harmonic: Your go-to startup database. Visit Harmonic to learn more.

Prelint: Keeps every decision reviewed, recorded, and traceable. Head over to Prelint to learn more.

VCX: The public ticker for private tech. Head over to GetVCX.com to learn more.

This Week In Startups

Ollie Forsyth — Mon, 12 Oct 2026 14:38:39 GMT

Hey readers,

Welcome back to your weekly tech digest, a quick look at the biggest stories shaping the tech world this past week.

What did you miss? OpenAI rolled out GPT-6 to every ChatGPT user, and the headline feature is “Intelligent UI”: answers that can now include charts, buttons, forms and small interactive tools built on the fly.

On the startup fundraising front, TypeSafe AI raised $870M at a $7.5B valuation in a round led by a16z, just weeks after launching Jev, a model that returns scores and probabilities instead of prose and says it passed 1M users in days. Elsewhere, AI evaluation platform Arena nearly doubled its valuation to $3.1B in ten months with a $200M Series B, and chip simulation startup Vinci raised $250M at a $1.5B valuation.

If that wasn’t enough, it was also SF Tech Week, where we hosted two new media summits: one with media execs alongside Yahoo, and one with creators alongside EO. More on both soon.

This edition covers:

👀 The latest AI news
💰 Top startup news & funding rounds (sourced using Harmonic)
💸 Latest VC funds
🤝 Acquisitions
🎤 Must-watch interviews from our podcast show

and much more.

The latest AI news

OpenAI launched GPT-6: Rolled out to all ChatGPT users with “Intelligent UI,” which builds charts, buttons and interactive tools into answers on the fly. Paid plans get GPT-6 Sol, free users get Luna, across 1.2B+ weekly users.
Mistral shows off Large 4, its biggest AI model yet. It’s huge, but only uses a small part of itself at a time, so it stays fast and cheap to run. Anyone can try it now, and Mistral says it will let people download it for free by the end of October. It was built entirely in Europe.
Arena, the community-driven AI model evaluation platform, raised a $200M Series B at a $3.1B valuation, co-led by Lightspeed and Khosla Ventures, with Andreessen Horowitz and Felicis.

Top startup news & funding rounds

We take a closer look at some of the hottest startup rounds of the past week.

Vinci, the AI-powered real-time hardware simulation platform, raised a $250M Series B at a $1.5B valuation, co-led by Advent, Temasek and Xora Innovation, with AMD, Eclipse and Khosla Ventures.
Valon, the AI mortgage servicing operating system, raised a $150M Series D at a $2.3B valuation with participating investors including Andreessen Horowitz and Ribbit Capital.
Vesta, which provides a swarm of agents to mortgage lenders, raised $30M led by Conversion Capital.
Mecka AI, a startup that collects and analyzes human motion data to train humanoid robots and other kinds of robots, raised a $60M Series B round led by Sequoia.
General Medicine, the virtual healthcare access platform, raised a $120M Series B led by Andreessen Horowitz, with Matrix, Eli Lilly and Mercy.
Capitolis, the capital markets optimization platform, raised a $120M Series E at a $1.9B valuation, led by Citi, with BNP Paribas, State Street, Barclays and UBS.
Parallel, the autonomous freight rail company, raised a $100M Series C led by AVP, with Hillspire and Riot Ventures.
Universal Quantum, the UK scalable quantum computing company, raised a $100M Series A co-led by DCVC and Firgun Ventures, with Roblox founder David Baszucki.
Spiko, the Paris-based tokenized cash platform that earns yield on corporate treasuries, raised a $90M Series B at an $800M valuation, led by NEA, with Index Ventures and Speedinvest.
Hone, started by former OpenAI and Cognition staffers, raised a $60M seed at a $285M valuation, led by Index Ventures and Benchmark.

(sourced using our partner, Harmonic)

Latest VC funds

Union Square Ventures raised $900M. $500M to back early-stage startups and $400M for its opportunity fund.
Forbion raised $2.6B across two funds, its largest raise ever. It’s a major boost for European biotech at a time when industry leaders are warning that the region is falling behind the US and China.
Endeavor Catalyst closed an oversubscribed $320M Fund V, its largest yet, taking AUM past $850M.
PSV closed Fund II at $63M to back Nordic founders before product-market fit, now across Denmark, Sweden and Norway.
Monte Carlo Capital raised $14.5M in a first close for Fund II to back seed-stage deep tech, AI and spacetech startups.

Acquisitions

TechWolf announced that it has signed a definitive agreement to be acquired by SAP.
Informa is acquiring Clarion Events for £2.24B (about $3B), going all-in on live B2B events.
Uber is buying ezCater for $2.3B in cash to bring catering to Uber Eats.
Cloudflare acquired Deno to improve its Workers programming model.

Clip of the week

How honest is too honest between co-founders? Join the conversation.

@lqiao has six co-founders. They're all engineers from hyperscalers.\n\nTheir glue isn't shared views. It's brutal intellectual honesty.\n\nDifferent perspectives, one common language: logic.\n\nWhat's the one thing your team shares so …","username":"ollieforsyth","name":"Ollie Forsyth","profile_image_url":"https://pbs.substack.com/profile_images/2077093891971624960/4nBmTets_normal.jpg","date":"2026-10-06T16:07:51.000Z","photos":[{"img_url":"https://substackcdn.com/image/fetch/$s_!DBmn!,w_1028,c_limit,f_auto,q_auto:best,fl_progressive:steep/l_play_button_usfui2,w_88,e_colorize:0/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F__ss-rehost__tw-video-preview-13_2107502931813797888.jpg","link_url":"https://t.co/sNMmzyOwrv"}],"quoted_tweet":{},"reply_count":6,"retweet_count":0,"like_count":22,"impression_count":1383,"expanded_url":null,"video_url":"https://video.twimg.com/amplify_video/2107502931813797888/vid/avc1/1280x720/5-mVvFjjopoFI4Wn.mp4?tag=16","video_preview_media_key":"13_2107502931813797888","belowTheFold":true}" data-component-name="Twitter2ToDOM">

Must-watch interviews from our podcast show

Linda Tong, CEO of Webflow, joins us to explain how she’s leading Webflow’s transformation from a $4 billion website builder into an agentic marketing platform, why she believes the internet is filling up with AI-generated “garbage” from people who never asked what’s worth building, and how she thinks about winning in a market where AI capability itself isn’t a moat.

During the episode, we also explore an interesting trend and tension: what it takes to build “taste” into AI-generated products when every model can now write clean code but none of them can reliably read a brand. Linda also shared how enterprise pricing negotiations are exposing the gap between what companies pay for and the value they actually get, and how she expects websites themselves to evolve from static pages into something closer to Google Maps: constantly adapting and reshaping around each visitor.

Previous episodes include

Fiverr CEO: The New Rules to Stay Employed in the AI Economy | Micha Kaufman.
Fireworks CEO: AI Startups Are Scaling Into Bankruptcy (And How to Fix It) | Lin Qiao.
Coinbase CTO: The Strategy to Achieve 99% AI Written Code | Rob Witoff.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).
Axios CEO: How to Get Ahead of Everyone Who’s Using AI | Jim VandeHei.

Our partners:

Harmonic: Your go-to startup database. Visit Harmonic to learn more.

Prelint: Keeps every decision reviewed, recorded, and traceable. Head over to Prelint to learn more.

VCX: The public ticker for private tech. Head over to GetVCX.com to learn more.

Greenhouse

Ollie Forsyth — Thu, 08 Oct 2026 15:55:43 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Daniel Chait, co-founder and CEO of Greenhouse, joins us to explain why AI has trapped hiring in a “doom loop” where, for the first time he can remember, “neither side is really happy.”

He co-founded Greenhouse in 2012 and has built it into one of the world’s leading hiring platforms, used by more than 7,500 companies with a team of over 700. His Dream Job feature lets job seekers flag one application a month as their top choice, and he says those candidates now convert from application to hire at around five times the rate of everyone else.

Daniel also shares why 1 to 1.5% of applicants fail an ID check, how North Korean operatives use deepfakes to get through interview rounds, and why he believes starting with an AI interview can remove bias from the most biased stage of hiring.

In our in-depth episode, we discuss:

How auto-apply tools and AI résumé filters created the hiring “doom loop,” and why job seekers and employers no longer trust each other
The three-layer “Swiss cheese” model Greenhouse uses to catch fake candidates, and why North Korean infiltrators are so hard to spot once they’re hired
Why an AI interview at the first stage can take racial bias out of résumé screening, and how it cuts around six days from the process
Why Greenhouse’s value has shifted from workflow to data and context, and whether it will come out of this moment as Blockbuster or Netflix
How Daniel decides which bets to place, why “we can do anything, we just can’t do everything,” and what it would mean to hire AI agents through Greenhouse

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

Daniel Chait (Greenhouse) Transcript

373KB ∙ PDF file

Timestamps

[00:00] Meet Daniel Chait
[02:16] The state of hiring right now
[04:03] Why employers get 1,000 applications in a day
[05:31] The AI doom loop explained
[10:31] Why AI résumé filters fail
[11:35] The “Jared problem”: how AI amplifies bias
[13:32] North Korean spies and fake candidates
[16:03] 1 in 100 applicants fail ID checks
[16:32] The Swiss cheese model of hiring security
[20:34] How deepfake hiring rings get caught
[22:30] Why Daniel is still an AI optimist
[23:34] Dream Job: the feature with a 5x hire rate
[26:05] How COVID, rates and ChatGPT changed hiring
[29:02] Are entry-level jobs disappearing?
[33:00] Is university still worth it?
[35:01] Why durable skills beat technical skills
[36:00] Is job hopping a red flag?
[37:34] Changing business models
[40:00] Staying relevant when anyone can clone software
[43:03] Blockbuster or Netflix?
[45:31] Greenhouse’s new mission
[47:30] Why AI interviews can remove racial bias
[51:30] What happens to humans in hiring?
[53:32] How AI interviews save 6 days per hire
[55:03] From software engineers to agent managers
[56:30] Could companies one day hire AI agents?
[1:00:02] How Greenhouse places bets
[1:03:03] A week as Daniel’s chief of staff
[1:04:35] Great leaders have no followers
[1:05:35] Leading 700 people as CEO
[1:07:02] What keeps Daniel motivated

Lessons from this episode with Daniel

1. Why AI has trapped hiring in a doom loop
Daniel explains how one unremarkable company got a thousand applications on day one, and why the tools meant to help job seekers have made the system worse for everyone.

“That’s kind of what we’ve referred to as the AI doom loop: this idea that the more each side is using the kind of current generation of AI tools to help themselves, the worse the overall system has gotten for everyone.”
For an in-person manufacturing job in the Midwest, the company was flooded with applicants from overseas and people with graduate degrees looking for tech roles.
Job seekers have put their search on autopilot, so AI applies to thousands of jobs on their behalf. That fills every pipeline with noise, which pushes people to send out even more applications.

2. The Swiss cheese defense against fake candidates
Between 1% and 1.5% of applicants on Greenhouse fail an ID check. Daniel breaks down the layered security model used to catch them before they get hired.

“In our data, about one to one and a half percent of job applicants hard fail an ID check.”
Every security layer has holes, so you stack them until the holes stop lining up: a voice AI interview first, then digital signals like IP address and device fingerprint, then full ID verification.
Scammers hate a 25-minute voice interview at the start of the process, and most of them drop out right away.

3. The dating app trick that makes candidates five times more likely to get hired
Greenhouse’s Dream Job feature lets job seekers flag one application a month as their top choice. Daniel explains why that scarcity creates a signal that didn’t exist before.

“We’ve had thousands of people get their dream job and they convert from application to hire at about five times the rate of other job applications.”
Because candidates only get one a month, they spend it on a job they actually want and think they can win.
The idea was borrowed from early decision in college admissions and super likes on dating apps.

4. Why Greenhouse’s moat is no longer its software
Customers no longer click through Greenhouse’s interface. They connect it to Claude or ChatGPT and do the work there, which changes where the value sits.

“It used to be that the value of Greenhouse was the workflow.”
The value is now in context and data: knowing which moves are legal, which candidates need a response, and which of millions of members are real people actively looking.
“Do you want to emerge from this moment as Blockbuster or Netflix? It’s just that simple.”

5. How AI interviews can design out the bias of résumé screening
Daniel points to a 2003 study showing that résumé review is one of the most biased stages in hiring, and argues that AI can remove that stage altogether.

Identical résumés with only the name changed showed about a 50 percentage point gap in interview invitations, depending on whether the name sounded stereotypically white or Black.
When interviews are scarce, companies gatekeep who gets one. An AI interviewer lets them interview everyone.
“I don’t care what your name is. I don’t care if you were on the women’s volleyball team or the men’s lacrosse team. None of that. It’s just about how do you answer the questions?”

6. What it would mean to hire an AI agent through Greenhouse
In an internal brainstorm, Greenhouse asked what happens when a customer says half of this year’s 100 hires will be agents. Daniel works through the question from first principles.

“What would it mean to put up a job posting for an agent? What would it mean for the agent to apply to that job or be interviewed or get an offer?”
Every step of hiring has a job to do. A job ad tells whoever or whatever can fill a need that the need exists, and an interview tests whether a candidate can actually do the work.
“Some of it’s a little science fiction for now,” but he says anyone in his seat had better be thinking about it.

7. Why great leaders have no followers
Daniel shares the leadership lesson from Fred Kofman’s Conscious Business that shapes how he runs a company of 700 people.

Picture runners on a track: the one in front looks like the leader, but the others aren’t chasing him. They’re all running toward the same finish line.
“As a leader, my job is not to get people to follow me. My job is to find the finish line and tell everyone as clearly as I can where it is.”
He doesn’t book sales, answer support tickets or ship code. His job is to set the vision and make it clear to customers, investors and employees.

Where to connect with us

Follow Ollie on LinkedIn: https://www.linkedin.com/in/ollieforsyth

Follow Daniel on LinkedIn: https://www.linkedin.com/in/dhchait

Visit Greenhouse: https://greenhouse.com

Our Partner for this episode is Harmonic - your go-to startup database https://harmonic.ai

Previous episodes include

Fiverr

Ollie Forsyth — Thu, 01 Oct 2026 14:21:17 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Micha Kaufman, co-founder and CEO of Fiverr, joins us to explain why anyone who has worked in their profession for more than five years is now “a dinosaur.”

He has spent 17 years building Fiverr into a global marketplace connecting millions of freelancers with businesses, and took it public on the New York Stock Exchange in 2019. His internal memo warning staff about AI went viral a year ago, and he says the employees who embraced it are now doing two to three times more than before.

Micha also shares why AI has flipped the 80/20 split between human and software work, why an AI that knows everybody “just makes everybody more average,” and why human judgment and accountability will be the biggest moat any professional can have.

In our in-depth episode, we discuss:

The asteroid analogy behind Micha’s viral AI memo, and what changed inside Fiverr in the 12 months after he sent it
Why the 20% of work AI can’t do is “infinitely important,” and the risk of outsourcing judgment to machines
Why most AI-generated code is built on poor-quality open source, and what that means for developers whose job was 60–70% copy-pasting
Why Micha wants to stop calling freelancers “freelancers,” and why he believes getting better is each person’s own responsibility
How Micha runs a public company through a period of uncertainty by encouraging impatience, and whether half of Fiverr’s clients could soon be AI agents

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES: Micha Kaufman (Fiverr)

227KB ∙ PDF file

Timestamps

(00:00) Meet Micha Kaufman
(02:17) Is the 9-to-5 dead?
(05:20) The viral AI memo
(05:50) The dinosaur and the asteroid
(12:25) Using AI vs really using AI
(14:33) The one thing AI lacks: accountability
(18:55) Has AI made us lazy?
(19:33) How AI flipped the 80/20 of work
(21:02) Why AI is a nightmare for designers
(23:23) What happens to freelancers doing basic work?
(27:32) Why "freelancer" is the wrong word
(28:38) Whose job is it to get better?
(32:04) Why most AI-generated code is bad
(38:32) Why adaptability wins
(41:10) Running a public company in uncertain times
(43:06) Why Micha encourages impatience
(44:02) The hard part of deploying AI
(45:19) Fiverr's next 12 months

Lessons from this episode with Micha

1. Why every experienced professional is now a dinosaur
Micha’s viral memo started with one analogy. He explains why anyone with more than five years in their profession is living through an extinction event, and why the outcome is transformation, not the end.

“If you’ve been doing something as a professional for more than five years, you are a dinosaur. And you are a dinosaur at the moment an asteroid makes impact.”
The crater is only the start. The debris, the darkness, the ice age and the food scarcity that follow are what decide which species survive.
Survival is not about evolution, which takes millions of years. It is about morphing into something new, and some species flourish while others go extinct.

[0:05:57 – 0:07:20]

2. The one thing AI will never have
Micha argues that the real line between humans and machines is not intelligence. It is accountability, and he traces the reason back to our mortality.

When ChatGPT gives harmful advice, the chat window still says, “I’m AI and I make many mistakes.” That line sums up AI’s lack of accountability.
Humans are accountable because we are mortal. Reputation is our most precious asset, and keeping it requires accountability.
“Human in the loop” is a buzzword. What it really comes down to is judgment and accountability, and the risk is that people start outsourcing both to software.

[0:14:33 – 0:15:57]

3. How AI flipped the 80/20 of work
Before AI, humans did most of the work and software did the rest. Micha explains why that ratio has reversed, and why the part left to humans matters more than ever.

Up until AI, humans did about 80% of the work and software did about 20%. Now it has flipped.
The 80% that AI handles “doesn’t get you anywhere. It just gets you faster.”
The 20% left to humans “is not 20% important. It’s infinitely important.” The danger is starting to outsource that 20% as well.

[0:19:36 – 0:21:02]

4. The X factor: why the same AI gives different results
If everyone has the same AI tools, where does the advantage come from? Micha argues it comes from the person using them.

Whatever AI produces is the average, because everybody has access to it. Something looking slick doesn’t mean it will work.
“If I prompt Claude Design and you prompt Claude Design, we’re not going to get the same results.”
The difference is design taste and knowing what works. “It’s not Claude Design. It’s me. This is the X factor.”

[0:26:10 – 0:27:32]

5. It’s not your CEO’s job to make you better
Micha shares the message he gave his own team, one he admits he might get criticized for.

“Look, it’s not my job to make you better. Not better human beings, not better professionals, not better husbands and wives, not better parents. It’s your job.”
He holds himself to the same rule: it’s not the board’s job or the market’s job to make him better. It’s his own.
Fiverr’s role is to help those who help themselves. “I would do anything to help those who help themselves.”

[0:29:05 – 0:30:22]

6. Why most AI-generated code is bad
Code is being written at blazing speed, but Micha says most of it is poor quality. The reason is built into how AI models learn.

AI is a statistical model. It answers a request from wherever it has the most data, and for code, that is open source.
“The vast majority of open source is terrible.” Repurposing it means code that is insecure, inefficient, full of bugs and weak on edge cases.
Five years ago, about 60 to 70% of a developer’s job was already copy-pasting from open source. AI now does that part.

[0:32:35 – 0:34:05]

7. Why Micha encourages impatience with the CEO
Running a public company takes long-term patience, but Micha pushes his team to be impatient, including with him.

The framework starts with being impatient with yourself. Only then can you demand the same of others.
If Micha says, “Okay, I’ll think about it,” that’s not a good answer. If he says a week, his team should ask: why not tomorrow? Can we talk about this tonight?
The balance: push hard on what can move fast, but “you can’t define 100 years in a day. It takes time.”

[0:42:50 – 0:44:02]

Where to connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Micha on X: https://x.com/michakaufman

Visit Fiverr: http://fiverr.com

Our partner for today's episode is Harmonic, your go-to startup database: https://harmonic.ai

Previous episodes include

Fireworks AI

Ollie Forsyth — Tue, 29 Sep 2026 14:18:05 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Lin Qiao, co-founder and CEO of Fireworks AI, joins us to explain why “a lot of companies will build AI into bankruptcy” this year. She previously led PyTorch at Meta, and her company now powers every major coding company, including Cursor, which it began working with at $2M ARR and has watched grow a thousandfold.

Lin also shares why the frontier labs’ bet on general intelligence leaves most companies without a moat, how customizing models on private data can make AI five to 10 times cheaper, and why Fireworks expects to process 100x its current 40 trillion tokens a day within a year.

In our in-depth episode, we discuss:

Why product-market fit no longer means a durable business in the AI era, and how companies can scale straight into bankruptcy
The framework for owning your own intelligence: put 10–20% of AI spend into training and cut total cost of ownership by four to eight times
Why the product itself is no longer the moat when coding agents make anything easy to copy, and why your private data is
How Lin runs a company with six additional co-founders, doubling headcount every quarter, on a culture of “extreme ownership”
What Lin learned from Jensen Huang about leadership, and why she believes the future belongs to millions of specialized models rather than a few general ones

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES: Lin Qiao (Fireworks)

282KB ∙ PDF file

Timestamps

(0:00) Meet Lin Qiao
(2:17) Why Jensen Loves Fireworks
(4:15) Jensen's Unique Philosophy
(7:33) What "Human Judgement" Really Means
(11:06) When Founders Should Stop Interviewing
(13:10) Why Join Meta Just to Leave and Build
(17:06) Building With 6 Co-Founders
(22:00) Managing Co-Founder Conflict
(24:47) Why Firms Build on Fireworks
(37:21) The Framework for Owning Intelligence
(38:40) How Long It Takes to Build an Open Model
(43:42) Respecting Privacy in Training
(45:57) The Next Moat
(48:53) Are Data Centers & GPUs an Opportunity?
(50:55) Open vs. Closed Model Strategy
(52:34) Ollie Becomes Lin's Chief of Staff
(54:18) How Lin Uses AI
(56:12) 40 Trillion Tokens a Day

Lessons from this episode with Lin

1. Why companies are scaling AI into bankruptcy
In the SaaS era, product-market fit and a durable business meant the same thing, because the cost of running software was low. Lin explains why AI has split those two concepts apart.

Once you hit product-market fit in SaaS, you scaled as fast as possible, because your biggest COGS was people.
In AI, your customers can love your product while the cost of serving them exceeds your income. “When you scale, you literally scale into bankruptcy.”
For startups, this means running out of money before the next round. For enterprises, it means the CFO looks at the cost forecast and refuses to approve the rollout.

[0:33:30 – 0:35:30]

2. The framework for making AI 10x cheaper
Every company Lin talks to asks the same thing: it sounds like an investment, so how do we justify it? Her answer is to think in terms of total cost of ownership.

Put 10–20% of your overall AI spend into training your own model, and the remaining 80–90% into inference.
Customizing the model cuts inference costs by five to 10 times, and total cost of ownership still falls four to eight times, even after the training investment.
Training isn’t one-off: “We literally launch new models every week,” so the model needs to evolve with the product.

[0:37:00 – 0:38:30]

3. Why your data is the only moat left
Coding agents have made products easy to build and easy to copy. Lin argues the real moat is the data a company collects from customers using its product.

Every company has unique product taste, judgment and design, but when anyone can build anything, “their uniqueness may not be that unique.”
A company’s customer preferences and feedback are its alpha, and that data will never be shared with anyone else.
Yet that data isn’t being used to create any intelligence. That’s the gap Fireworks calls specialized intelligence.

[0:32:00 – 0:33:30]

4. “Are you sure you’re going to start a company with seven of you?”
When Lin pitched Benchmark’s Eric Vishria for her first round, he was surprised by the size of her founding team. Most companies have two or three co-founders; Fireworks has seven.

The subtext of his question was that big founding teams usually fall apart through drama and founder dynamics.
Lin’s answer is that the co-founders are “brutally intellectually honest” with each other and share an engineering background, which makes logic their common language.
Four years in, she says they have become each other’s strength and a bigger force than any one of them alone.

[0:21:00 – 0:22:00]

5. The Navy SEAL principle behind Fireworks’ culture
Fireworks doubled from 150 to 300 people in a single quarter, and Lin still interviews everyone. The trait she looks for is “extreme ownership.”

The army is extremely hierarchical, and everyone stays in their lane. On a battlefield, those boundaries stop mattering: you take responsibility, make the call and watch each other’s backs.
At Fireworks, “no problem is anyone else’s problem.” It doesn’t matter who wrote the code.
It isn’t enough to call out a gap; you have to see it through and fix it yourself.

[0:09:00 – 0:10:30]

6. What makes Jensen Huang different from every other CEO
Lin believes leadership is not a privilege but judgment, and says Jensen operates unlike any conventional CEO.

She was shocked by how much detail he understands, from capacity allocation to the technical details of every topic.
Traditional companies rely on hierarchy to move information up and down the chain; Jensen simply knows everything across Nvidia.
She predicts AI will make this style of leadership far more common, because information will flow freely without deep hierarchy.

[0:05:00 – 0:06:30]

7. How Cursor grew 1,000x, and why knowledge work is next
Fireworks started working with Cursor when it was at $2M ARR, and has watched it grow a thousandfold in three years. Lin saw 2025 as the year of coding and this year as the year of knowledge work.

Every major coding company now runs on Fireworks.
Lin describes knowledge work as a bushy tree with coding as the trunk. Every profession, from dentistry to legal, has its own depth.
Each tip of that tree is an AI-native startup building agents for one specific domain, which is why AI applications are now diversifying so quickly.

[0:25:30 – 0:27:30]

8. From 40 trillion tokens a day to 100x
Fireworks processes 40 trillion tokens a day. Asked where that will be a year from now, Lin weighs two opposing forces.

Demand will definitely go up.
Token efficiency will rise too: smaller models will deliver the same intelligence and need fewer thinking tokens to reach a good result.
Even with those compounding effects, she believes reaching 100 times more tokens within a year is possible.

[0:56:00 – 0:57:00]

Where to find and connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Lin on X: https://x.com/lqiao

Visit Fireworks: https://fireworks.ai

Our partner for today’s episode is Harmonic, your go-to startup database: https://harmonic.ai

Previous episodes include

This Week In Startups

Ollie Forsyth — Mon, 28 Sep 2026 13:45:50 GMT

Hey readers,

Welcome back to your weekly tech digest, a quick look at the biggest stories shaping the tech world this past week.

What did you miss? Meta put Muse, its personal AI agent (video below), on glasses and Mac and unveiled Muse Charm, a keychain gadget with a 2-inch screen that Zuckerberg pushed the team to ship this year instead of next. Muse already runs across Meta’s app, WhatsApp and the web, and the new hardware is Meta’s bet on keeping it within reach all day. The open question is how many people actually want an AI agent on their face or dangling from their keys.

In other news, Anthropic committed $11.6B to Akamai over seven years, sending Akamai’s stock up more than 15% after hours. Cognition’s Devin has doubled its revenue to $1B ARR since May, and DeepSeek also crossed $1B ARR after raising its API prices.

On the deals front, Databricks acquired Row Zero, Cloudflare bought AI training-data marketplace Human Native, and Qualcomm picked up PickNik Robotics. And in fundraising, Bessemer closed $5.75B, while Eileen Burbidge raised $27M for her solo-GP fund, NFG Ventures.

This edition covers:

👀 The latest AI news
💰 Top startup news & funding rounds (sourced using Harmonic)
💸 Latest VC funds
🤝 Acquisitions
🎤 Must-watch interviews from our podcast show

and much more.

👀 The latest AI news

Meta puts Muse on glasses, Mac and a new pocket device, Muse Charm.
Daniel Ek’s body-scanning startup Neko Health opened doors for its first U.S. location in New York.
Anthropic commits $11.6B to Akamai over seven years.
Cognition’s Devin passes $1B ARR, double what it reported in May, weeks after raising at a $48B valuation.
DeepSeek hits $1B ARR after raising API prices up to 4.5x and is closing a $7.5B raise through a Shanghai listing.

💰 Top startup news & funding rounds

We take a closer look at some of the hottest startup rounds of the past week.

TEKEVER, the Lisbon-based maker of AI-driven autonomous drones and surveillance systems, raised $580M Series D at a $6.4B valuation, co-led by Baillie Gifford and UC Investments.
Island, the enterprise browser platform that builds security controls into both human and AI-agent activity, raised $400M Series F at a $6.4B valuation, led by Evolution Equity Partners, with Sequoia Capital, Coatue, and Insight Partners.
Snorkel AI, the AI data development platform for building and evaluating enterprise AI, raised $350M at a $3.5B valuation, led by Insight Partners, with Lightspeed Venture Partners, GV, and Addition.
Heidi Health, the Melbourne-based AI medical scribe that writes up clinical notes from consultations, raised $335M at a $1.26B valuation, co-led by General Catalyst and Blackbird, with Point72 Ventures and Headline.
Enveda, the Boulder-based biotech using AI to find new drugs in natural compounds, raised $311M Series E at a $2B valuation, led by Catalio Capital Management, with Lightspeed, ICONIQ, T. Rowe Price, Baillie Gifford, and Lux Capital.
OpenEvidence, the AI medical search and clinical decision support platform for physicians, raised $250M at a $15B valuation, with Andreessen Horowitz participating.
Precision Neuroscience, the New York-based brain-computer interface company letting people with paralysis control devices, raised $250M Series D at a $1B valuation, led by Pershing Square, with ARK Invest, B Capital, and Mubadala Capital.
Rightway, the New York-based pharmacy benefits manager that pairs clinicians with technology, raised $155M Series E, led by Francisco Partners, with Thrive Capital and Khosla Ventures.
Numeral, the San Francisco-based AI-native platform for sales tax, VAT, and GST compliance, raised $100M Series C, led by Insight Partners, with Benchmark, Mayfield, and Y Combinator.
Ema, the “universal AI employee” that automates HR, IT, and finance work, raised $77M Series B, led by Creaegis, with Accel and Prosus Ventures.
Confido, the New York-based AI platform that brings finance and trade operations together for consumer packaged goods brands, raised $55M Series B, led by Insight Partners, with Y Combinator.
Ande, the AI-native platform connecting enterprises with corporate entertainment, raised $52M Series A, with Lightspeed Venture Partners, Redpoint, and Bain Capital Ventures.
Solcoa Industries, the Alameda-based producer of rare earth metals for electric motors and generators, raised $45M, co-led by Bain Capital Ventures and J.P. Morgan, with Felicis.
Corridor, the New York-based benefits brokerage for small and mid-sized businesses, raised $25M Seed, led by Bain Capital Ventures, with OpenAI, Scale AI, BoxGroup, and Ramp.
Ando, the team chat platform that brings AI agents into conversations as teammates, raised $20M Seed, co-led by Index Ventures, Accel, and Emergence Capital.
Basalt Health, the Nashville-based post-acute care tool that works inside electronic medical records, raised $20M Series A, led by NEA.

(sourced using Harmonic)

💸 Latest VC funds

Bessemer Venture Partners closed $5.75B to back founders from seed to growth.
Connect Ventures raised $55M (targeting $80M) for Fund V to back European founders.
Eileen Burbidge has raised $27M for her solo-GP fund, NFG Ventures.
Ibex Investors raised $87M for new Israel-focused secondary fund.

🤝 Acquisitions

Row Zero, the cloud spreadsheet for big data, was acquired by Databricks.
Human Native, the London-based marketplace that licenses content as AI training data, was acquired by Cloudflare.
PickNik Robotics, the US-based robotics software company behind MoveIt, was acquired by Qualcomm.

Our AI Microdrama Series

We've launched Billion Dollar Tea, an AI-powered show spilling the juiciest tea from the world's biggest companies through real stories and imagined what-ifs. Episode 4 continues our Elizabeth Holmes arc and imagines her comeback: what would she build next? Follow the show here which is in collaboration with Plot Party.

We had one of the Theranos whistleblowers on the podcast last year. Watch our conversation with Tyler Schultz here.

Upcoming events

September 29th: Meet the ElevenLabs team | Why You Should Be Using ElevenLabs. Register here.
October 7th: New Media Summit in collaboration with Yahoo in S.F. Our latest guest speaker is Jessica Jensen, CMO at LinkedIn. Apply here to attend.

Must-watch interviews from our podcast show

Alan Trefler, Founder and CEO of Pegasystems, joins us to unpack how a 40-year-old, $6 billion public company is thinking about AI, why he refuses to charge customers for tokens, and what it actually takes to survive five generational shifts in technology without losing your edge.

You might not know this about Alan, but Pega’s first two clients, Citibank and Bank of America, signed on in 1984 and are still customers today. He shared why he thinks the real risk in enterprise AI isn’t the technology, it’s the incentive: “These guys are a little bit like drug dealers… passing out packets of tokens and getting people hooked.”

Previous episodes include

Superhuman CEO: The Strategy to Turn Standalone Apps Into a $13 Billion Suite | Shishir Mehrotra.
Europe Is Quietly Building the New Silicon Valley | Saul Klein.
$2 Billion Founder: The Secret to Building a High-Growth Startup in 24 Months | Tanay Kothari.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Why most people are underusing AI and why the winners won’t just be OpenAI and Anthropic | Reid Hoffman.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).

Coinbase

Ollie Forsyth — Sat, 26 Sep 2026 14:16:33 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Rob Witoff, CTO of Coinbase, joins us in his first interview since returning to the crypto company. He explains why 98–99% of Coinbase’s code is now written by agents, how his engineers went from writing code by hand to supervising 3–10 agents each in a single year, and why a Friday-morning “spicy feedback” agent may be giving Coinbase bigger productivity gains than AI in product development.

In our in-depth conversation, we discuss:

Why Coinbase engineers are producing 100x the code they did a year ago, and why agent managers of agent managers are coming next
Why it’s now cheaper to build a prototype than to ask for permission, and why a picture is worth a thousand words but a prototype is worth a million
The weekly AI feedback agent every Coinbase employee receives, and the “spicy” section that tells Rob to get out of the weeds
Why crypto is so volatile, and why "it's never as good or as bad as it seems
Will Coinbase become the financial super app?

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast Rob Witoff (Coinbase)

337KB ∙ PDF file

Timestamps

(0:00) Meet Rob Witoff
(2:30) Crypto's Understated Opportunity
(4:09) Coinbase's Next Opportunity
(5:30) Inside Coinbase's CTO Role
(8:10) Managing a Third of the Company Under Engineering
(10:28) Taste & Judgment in the Modern Era of Engineering
(13:33) Inside Coinbase's Engineering Team: Zero to One
(15:35) How to Set Up Engineers for Success
(18:28) How Rob Sets Up His Agents
(20:17) How Disruptive Are Model Updates
(22:51) Would Coinbase Ever Launch Its Own AI Model?
(26:12) Agentic Trading
(26:52) Why Are Markets So Volatile?
(29:02) What Is the Super App Opportunity?
(31:20) Why Now Is the Moment for Agentic Financial Services
(34:38) The Fundamentals of Security
(35:38) The Agentic Consumer Experience
(41:34) Prediction Markets
(43:52) Stablecoins
(45:45) Will Stablecoins Be Country-Owned?
(47:35) The New Financial Infrastructure
(49:52) How Coinbase Structures Internal Feedback
(54:16) Ollie Joins as Rob's Chief of Staff
(57:41) Has 2026 Been the Inflection Point for AI?

Lessons from this episode with Rob

1. Why 98–99% of Coinbase’s code is now written by agents, and why engineers are becoming supervisors of loops and systems

A year ago, Coinbase engineers were writing code by hand. Today almost all of it is written by agents, and Rob says he didn’t predict it would change this fast.

Most engineers now supervise three to five agents, and sometimes 10, throughout the day. Rob calls that a very different muscle to exercise.
The next step is supervising whole loops and systems, which shifts the job to deciding what problems should be solved in the first place.
Coinbase is building a “Coinbase brain” that holds its product and architecture strategy and directs its agents and loops.

2. The weekly AI feedback agent that tells Rob to get out of the weeds

Every Friday at 9am, everyone at Coinbase gets a full assessment of how they’re working, built from their Slack messages and transcribed meetings. It covers what they’re doing well and where their gaps are.

Rob’s favorite part is the “spicy feedback” section, which asks: what is the really hard feedback you need to hear?
People might not want to say “Rob, get out of my way,” but his agent isn’t afraid to say it.
Lessons that used to take him six months to learn, he now iterates on every week.

3. Why it’s now cheaper to build a prototype than to ask for permission

Rob doesn’t want people coming to him or their lead with an idea. He wants them to go and build it, and that applies to 100% of the company, including non-engineers.

During one tax season, people were debating how to fix a confusing tax form. A non-engineer built a fully working prototype, and it immediately became the thing they built.
A picture is worth a thousand words, but a prototype is worth a million.
Once people can see, touch and feel something, it gets far more attention and momentum.

4. How Coinbase builds taste and judgment in engineers

Rob says taste in modern engineering means thinking like an entrepreneur about what you’re building and why. That starts with being a daily user of your own product.

Employees can now direct deposit their pay into the Coinbase app, a small feature that makes it easy for everyone to use the product and find its rough edges.
Judgment comes from battle scars: you have to make mistakes to learn the big lessons.
Large companies can’t let staff make mistakes that hit customers, so Coinbase builds safe environments where engineers can break things without customers being affected.

5. Why Coinbase flattened its layers so the engineer doing the work is in the room

Large companies have traditionally had many layers, from manager to director to senior director and up. Rob says that is good for coordinating but slow for executing.

Over the last year, Coinbase deliberately removed layers between decision-makers and the people doing the work.
His rule of thumb is that for any big architectural decision, the engineer who will execute it is in the room when it’s discussed.
This removes the games of telephone that have historically made it hard to execute at scale.

6. How to trust an AI agent with your money: read-only first, hard limits second, human sign-off third

Rob describes a path that applies to almost any use of agents. Agents start with read-only access and earn more control over time.

Even read-only access can show you how your portfolio is allocated and whether it still matches your risk tolerance.
Once agents can trade on your behalf, they need hard limits on which funds they can move.
Escalating to a human has to be a core part of the workflow, so the agent never makes an assumption you wouldn’t make.

Where to find and connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Rob on X: https://x.com/rwitoff

Visit Coinbase: https://coinbase.com/

Our partner for today's episode is Harmonic, your go-to startup database: https://harmonic.ai

Previous episodes include

If you enjoyed this episode, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Axios

Ollie Forsyth — Mon, 21 Sep 2026 12:02:27 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Jim VandeHei, Co-Founder and CEO of Axios and Co-Founder of Politico, joins us to explain why he believes the open web will largely die within 2 to 5 years as LLMs become the main way people get their information, why unique, human-created content will soar in value once everyone has access to Einstein-level AI, and how his Simplify framework helps people delete the busy work that eats up roughly 25% of every week.

In our in-depth conversation, we discuss:

Why trust scores for media will never work
Why Jim can’t understand how anyone is a solo founder, and how he has worked with his co-founders: Mike Allen and Roy Schwartz for nearly 20 years
Who gets paid when LLMs train on creators’ work: the courts, licensing or subscriptions?
The plan to save local news with AI, one community at a time
Why Axios fires talented people who aren’t good people culturally, and what “killers with humility” means
The Simplify framework: how to stop wasting 25% of your week

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Jim VandeHei

372KB ∙ PDF file

Timestamps

(0:00) Meet Jim VandeHei
(2:11) The State of Media Today
(4:09) Will We Have Trust Scores?
(6:48) Are We in a Better Place Today?
(11:08) How Will Creators Be Compensated by LLM Models?
(13:32) Timely vs. Timeless Content
(17:51) The Opportunity for Creators
(19:10) AI Summaries
(23:32) How Will Creators Make Money?
(25:14) The Opportunity for Local News
(31:25) Inside Axios
(35:25) How to Navigate Tough Days
(39:09) Surrounding Yourself with Great People
(47:46) What Makes Dyslexic Founders Tick
(49:14) 50% More, with 50% Less
(52:22) How to Prioritize Your Life and Work
(56:23) Ollie Joins as Jim's Chief of Staff

Lessons from this episode with Jim

1. How Mike Allen’s humility shaped Axios’s culture, and why what a leader does is instantly contagious
Jim credits his co-founder Mike Allen with the company’s culture. Mike brings cupcakes to events and stays behind to clean up so the cleaning crew doesn’t have to.

If the most famous, powerful person at the company has no ego, nobody else has an excuse to have one.
Whether you lead one person or many, your behavior spreads fast: treat people well and they will treat people well.
Bad behavior spreads the same way, and he says it will move like cancer if you allow it.

2. Why AI companies pay creators nothing today, and why Jim believes they will have to compensate them
Jim says AI companies currently believe fair use lets them take whatever they want off the internet, and that legally they might be right. He argues that AI still needs new information, new ideas and human expression, so it will have to pay for them.

He expects the courts to decide over the next one to four years how much power AI companies have to use other people’s work.
Whether the companies are morally right is a separate question from whether they are legally right.
He sees three possible routes to payment: paying people to produce content, licensing deals, or subscriptions built into your personal LLM.

3. Why unique, human-made content will soar in value when everyone has Einstein-level AI
Jim says he works under the assumption that within three years, everyone will have access to Einstein-level intelligence on every topic. In that world, he argues, content that is truly unique becomes more valuable, not less.

He compares it to everyone having equal access to a trivia contest when Google is in front of them.
People are wired to want survival and also something other people don’t have, so demand for differentiated content won’t go away.
Commodity content that isn’t vital or useful to anyone will have no value.

4. Why your media diet is your responsibility, and why there is no hero in this story other than you
Jim argues that no one can credibly score information for you, so you have to take personal responsibility for what you consume. He treats your media diet like your food diet.

Nobody would agree on who should score media for trust, so it comes down to the individual.
Find at least one source that tries to get closest to the truth in a non-ideological way.
Done properly, he says you can develop almost a bionic brain, with access to the smartest people on any topic.

5. Why you learn more in a year at a startup than in eight years of college
Jim advises anyone young to join or start a startup, and describes it as a living experiment built on the assumption that you’re going to fail.

He says you learn more in a month at a startup than in five years of a regular workplace.
Most startups fail, but you learn about yourself, about fear, and about what you want and don’t want in people.
The pain builds emotional intelligence beyond intellectual intelligence, and gives you a more panoramic view of people and the workplace.

6. Why Jim thinks it has never been a better time to be alive
Jim says he doesn’t share the pessimism of other people, and believes a lot of them are living in a dark parallel universe. He argues that by most standards, this is the best time in history to be alive.

As a species, he believes we have handled bigger problems than this one and will figure it out.
In his view, most people today are richer, more mobile, more educated, healthier and living longer than at any point in history.
He says that almost everything he lists is not debatable.

7. The 10-minute exercise for deciding what to focus on: write down the three things you’d want to be true about your life
Jim answers how you get down to the two or three things that really matter. He starts with a simple exercise you can do even without reading the book.

Imagine you’re at the end of your life and write down the three things you’d want to say were true.
Use them as a North Star for decisions like taking a job or learning a new skill, then repeat the exercise for your career.
If you’re new to a job, work out the three things you need to do to crush it, then check with your boss that you’re on the same page.

Where to find and connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Jim on X: https://x.com/JimVandeHei

Visit Axios: http://axios.com

Jim's new book: https://www.smartbrevity.com/simplify

Our partner for today's episode - Harmonic: Your go-to startup provider: https://harmonic.ai

Previous episodes include

If you enjoyed this episode, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

This Week In Startups

Ollie Forsyth — Sun, 20 Sep 2026 14:58:09 GMT

Hey readers,

Welcome back to your weekly tech digest, a quick look at the biggest stories shaping the tech world this week.

This past week in tech: the debate over AI’s pace moved to center stage. Dario Amodei published “We Must Pace the Frontier,” calling on labs to slow frontier-model development through independent evaluators, inter-lab coordination, and international cooperation, with Sam Altman and Elon Musk endorsing the post within hours of publishing. Sam Altman also said OpenAI won’t IPO in 2026, calling it “an ill-advised moment to go public.”

This edition covers:

👀 The latest AI news
💰 Top startup news & funding rounds (sourced using Harmonic)
💸 Latest VC funds
🤝 Acquisitions
🎤 Must-watch interviews from our podcast show

and much more.

First, some news:

Media is changing fast, as we’ve covered in previous editions on the new media landscape. With that in mind, we’re hosting a New Media Summit with Yahoo in San Francisco on October 7th, during tech week.
The program will feature Dylan Abruscato, President of TBPN, in conversation with Ollie Forsyth, founder of NEW ECONOMIES, followed by a panel discussion with tech journalist Joanna Stern, Yahoo CCO Sona Iliffe-Moon, and Vox Media VP of Podcast Business Lillian Xu, moderated by Jim Louderback.

In town and want to join us on October 7th? Request an invite - this will be a special one.

Request to attend

The latest AI news

Dario Amodei published “We Must Pace the Frontier,” calling on AI labs to slow frontier-model development with independent evaluators, inter-lab coordination, and international cooperation. Sam Altman and Elon Musk endorsed the post within hours.
Sam Altman said OpenAI won’t IPO in 2026, calling it “an ill-advised moment to go public.” He described even a 10% risk of AI-driven human extinction as unacceptable and hinted at an imminent industry pact on safety.
The Boring Company raised $3B at a $23B valuation, led by the UAE, to fund 150+ km of tunnels across the Emirates plus expansion in Vegas and Nashville.

Top startup news & funding rounds

We take a closer look at some of the hottest startup rounds of the week.

Crusoe, the AI cloud infrastructure platform building GPU data centers for high-performance AI workloads, raised $3.9B Series F at a $30.9B valuation, co-led by Valor Equity Partners, Atreides Management, and Mubadala Capital.
Temporal Technologies, the durable execution platform for fault-tolerant distributed workflows, raised $550M Series E at a $12.6B valuation, co-led by Lightspeed Venture Partners, Goldman Sachs Alternatives, Wellington Management, and Tiger Global.
Barq, the Riyadh-based fintech app for payments and money transfers, raised $329.5M Series A at a $1.85B valuation, with Noon Investments, Sohar International, and M20 Capital Group.
Upwind Security, the cloud security platform fusing runtime intelligence with build-time insights, raised $300M Series C at a $3.8B valuation, co-led by TCV and Bessemer Venture Partners.
Exein, the Rome-based runtime security platform for physical AI systems and embedded devices, raised $270M at a $1.7B valuation, led by Headline.
Ridgeline, the cloud-native investment management platform built by Workday founder Dave Duffield, raised $250M Series E at a $1.43B valuation.
EnduroSat, the Sofia-based small satellite and space intelligence company, raised $205M, co-led by Atreides Management and Riot Ventures, with Founders Fund, GV, Lux Capital, and Giant Ventures.
Factory, the AI coding agent platform for autonomous software engineering, raised $200M at a $5.2B valuation, with Sequoia Capital, Blackstone, NEA, Insight Partners, Khosla Ventures, and Clearlake Capital.
Angle Health, the AI-native health insurance platform acting as carrier, TPA, and MGU in one, raised $200M Series C at a $2.7B valuation, led by Vitruvian Partners, with Y Combinator, Portage, Blumberg Capital, and Pruven Capital.
Profound, the AI search analytics platform helping brands optimize visibility across ChatGPT, Perplexity, and other AI engines, raised $180M Series D at a $1.8B valuation, co-led by Kleiner Perkins and Sequoia Capital, with Lightspeed and Khosla Ventures.
Arcee AI, the open-weight small language model company building enterprise-grade domain-adapted AI, raised $150M Series B at a $1.15B valuation, co-led by Vista Equity Partners, Cambium Capital, and Emergence Capital.
CADDi, the AI data platform digitizing and analyzing manufacturing supply chain data, raised $114M Series D at a $1.2B valuation, with Toyota, Woven Capital, Atomico, Salesforce Ventures, and Recruit Holdings.
Thatch, the healthcare benefits platform letting employers set budgets while employees pick their own plans, raised $108M Series C at a $1B valuation, co-led by Index Ventures, General Catalyst, Andreessen Horowitz, and The General Partnership.

(sourced using Harmonic)

New VC Funds

Bain Capital Ventures raised $1.6B to invest in founders building for an abundant post-AGI world.
Portage Ventures raised $600M for fund four.
Crane Venture Partners raised $484M to back extraordinary founders globally, from inception through seed.
Activate closed its maiden $105M fund to back AI-native startups out of India.

Acquisitions

OpenAI acquired GLASS Imaging for $300M+, the computational photography startup founded by two ex-Apple Portrait Mode engineers.

Must-watch interviews from our podcast show

Aydin Senkut, Founder and Managing Partner at venture firm Felicis, joins the podcast at a pivotal moment in tech to unpack what’s happening across venture, whether LPs are concerned, and how to win as a firm in the current environment.

You might not know this about Aydin, but he was actually one of the first employees at Google, working directly alongside Sergey Brin and Larry Page, with a key lesson: “Why the clearest signal from Larry Page wasn’t what he said yes to but the 80% of the time he said no.”

Previous episodes include

Superhuman CEO: The Strategy to Turn Standalone Apps Into a $13 Billion Suite | Shishir Mehrotra.
Europe Is Quietly Building the New Silicon Valley | Saul Klein.
$2 Billion Founder: The Secret to Building a High-Growth Startup in 24 Months | Tanay Kothari.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Why most people are underusing AI and why the winners won’t just be OpenAI and Anthropic | Reid Hoffman.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).

If you enjoyed this edition, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Andreessen Horowitz

Ollie Forsyth — Sat, 19 Sep 2026 15:41:45 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Josh Elman has shaped products at some of the biggest consumer platforms in tech. He helped lead product at LinkedIn, Facebook, and Twitter, invested in startups at Greylock, and took senior operating roles at Robinhood and Apple. Now a Partner at Andreessen Horowitz, he joins us for his first external podcast interview since joining the firm.

In our in-depth conversation, we discuss:

Why anyone can clone a product’s features within days, but nobody can clone where a founder is actually headed next
Why we haven’t seen many new consumer social networking sites
The new era for voice
Is trust the new moat and what does it actually mean?
The pricing challenge for today’s startups
Is S.F still the best place to launch and grow a startup?

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Josh Elman (a16z)

359KB ∙ PDF file

Timestamps

(0:00) Meet Josh Elman
(2:40) Why Josh Is Back in Investing Mode
(4:19) The New Consumer Experiences
(6:00) Where in the Journey Are We?
(8:00) What Comes Next for Consumers?
(10:58) Is Trust the New Moat?
(11:40) The New Way to Search
(13:20) Experiences That Haven't Emerged Yet
(15:01) Trends Josh Is Excited About
(19:08) Whatnot
(21:05) AI Microdramas
(26:43) Why Is Social So Hard?
(32:20) Multi-Agent Communication
(36:19) New Platforms With Distribution
(38:27) The New Era for Voice
(43:01) The Pricing Challenge
(46:47) What Excites Investors Today
(49:14) How Network Effects Have Changed
(51:36) Idea to Clone: Days
(52:39) Why the U.S. Wins at Consumer
(54:34) Is SF Still the Best Place to Start a Company?
(55:43) Josh's Favorite Company
(57:35) Lessons from Ev Williams

Lessons from this episode with Josh

The cycle of adoption

One person built a category-defining agentic tool almost entirely on his own, working out of his apartment. Josh uses it to mark the moment agentic AI went from theory to something people could actually feel the difference of, and why that shift is what’s now pulling the entire industry into this new wave.

Multi-agent communication (agents planning your Friday night)

Josh’s most vivid vision in the episode: agents quietly talking to each other on your behalf, turning a private thought such as, wanting to see a movie into a real plan with friends before you even have to ask.

Why building consumer social is very hard

We know building consumer startups is super difficult. Josh does a relatable comparison of why Discord, Musical.ly, and Instagram all won by solving completely different problems, not by competing head-on. He describes the potential opportunity as, we need something that feels genuinely new, not another variant of what’s already out there.

What excites investors today

Josh lays out his four-part framework for what makes a great consumer product: clear value on first use, an easy substitution for existing behavior, organic word of mouth, and long-term retention.

The new era for voice

The take on why voice will never fully replace typing, and where it actually shines: in the car, prepping for a podcast, or rehearsing a hard conversation before having it for real.

The pricing challenge

We break down the tension building under consumer AI: free, subsidized tools are colliding with the real cost of inference. Josh traces it back to his first week at LinkedIn, physically moving servers, to show how infrastructure costs have shifted from fixed to variable.

Where to find and connect with us

Follow Ollie on X: https://x.com/ollieforsyth

Follow Josh on X: https://x.com/joshelman

Visit Andreessen Horowitz: https://a16z.com

Our partner for today's episode is Harmonic - your go-to startup database: https://harmonic.ai

Previous episodes include

If you enjoyed this episode, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Webflow

Ollie Forsyth — Mon, 14 Sep 2026 16:35:31 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Linda Tong, CEO of Webflow, joins us to explain how she's leading Webflow's transformation from a $4 billion website builder into an agentic marketing platform, why she believes the internet is filling up with AI-generated "garbage" from people who never asked what's worth building, and how she thinks about winning in a market where AI capability itself isn't a moat.

About Webflow

Webflow gives every team the tools to build, manage, and grow a website that drives real revenue. Founded in 2013 by Vlad Magdalin, Sergie Magdalin, and Bryant Chou (via Y Combinator), it’s grown from a niche tool for freelance designers into infrastructure used by 300,000+ businesses. The company raised a $120M Series C in March 2022 at a $4B valuation and has taken in over $330M total.

Watch now: Linda Tong, CEO at Webflow

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Linda Tong

364KB ∙ PDF file

Timestamps

(0:00) Meet Linda Tong
(2:04) The Current State of Building
(2:40) What's Changed for Webflow in the Last 12 Months?
(6:06) Base44 Reaches 10M Users and Lovable Raises $400M
(8:02) How to Support Builders in This Era
(9:18) How Do You Select Which AI Model to Partner With?
(12:00) Is Taste the Next Biggest Moat?
(13:41) Are Agentic Co-Workers Next?
(16:03) Webflow's Current Challenges
(18:46) How to Navigate Human Change
(21:08) The Change of Software Pricing
(25:12) Inside the Webflow CEO Role
(26:32) How Do AI Models Affect Product Roadmaps?
(28:35) Where Does AI Go Next Over the Next 6-12 Months?
(32:52) Will Websites Still Be Relevant?
(34:25) Building Trust With Your Fanatical Users
(36:25) How Linda Runs Webflow
(39:14) A Winning Mindset
(41:10) Are We Just 1% of the Way There?
(43:23) Airtable Acquisition
(45:09) Ollie Joining as Linda's Chief of Staff
(47:19) Personal Time Out
(48:20) Board Members
(50:50) Where Does Webflow Go From Here?

Lessons from this episode with Linda:

1. Linda can build an app with a prompt, but still can’t book a doctor’s appointment less than 6 months out.
Right after saying AI has barely scratched the surface of solving real problems, Linda brings up trying to schedule an appointment and being told the system doesn’t open bookings for another 6 months, so she’d have to call back in 3 months just to schedule it. Her point: we’re maybe 5% of the way to AI actually mattering in daily life.

2. “Just have everything run agentically. Go watch Netflix.” She’s not buying it.
Pushing back on the one-employee-plus-millions-of-agents narrative, Linda says AI still isn’t reliable enough to run unsupervised. You still need people reviewing and coaching the system regularly, and anyone claiming they’ve replaced their whole workforce with agents “is just not real.”

3. “I’m paying for a million seats but only 10 people actually get value out of it.”
Her example of what’s broken in software pricing: when she sits in on procurement negotiations, this is the exact complaint she hears. She argues it’s a sign the pricing was never actually tied to value in the first place, and that the constant tough renegotiations are the tell.

4. Airtable sold for 2.5x revenue, after being privately valued at $11.5 billion.
On how fast valuations are resetting: Airtable, a $500M-revenue company, got acquired at roughly 2.5x enterprise value, a fraction of the $11.5 billion it was last valued at privately. Her takeaway: “times are changing,” and the old rules for pricing a software company no longer hold.

5. Internet garbage: Only 200 million of the internet’s 1.2 billion websites are actually alive.
When Linda joined Webflow four years ago, she looked at the market and found roughly 1.2 billion websites live on the internet, but only around 200 million were active businesses actually being used. She says that ratio hasn’t meaningfully moved since, even as AI makes it trivial to spin up more of the other billion. The ability has shifted to build fast, people have built a lot of websites but that doesn’t automatically translate to tangible value.

6. AI didn’t kill the SDR job. It turned “send more emails” into “review the AI’s calls.”
Her clearest change-management example: an entry-level SDR used to be capped by how many calls and emails they could physically make in a day. Now an AI agent makes the calls and writes the emails, and the job becomes reviewing what the agent learned and deciding what to test next.

7. Her agents don’t do one task and stop. They run until the goal is hit.
She contrasts most “agents” today, which complete a task and hand it back for a human to judge, with what she calls closed-loop agentic workers: give it an outcome like “drive this much pipeline,” and it writes the brief, launches the campaign, measures results, and keeps iterating on its own. Her live example is Webflow’s AO agent, which pushes content and schema changes to improve a client’s visibility on answer engines like ChatGPT, then measures and repeats.

8. Pricing debates (subscription vs. seat vs. consumption vs. outcome)
Her contrarian take: none of these pricing models are wrong, they’re just successive attempts to get closer to charging for the actual value delivered. AI just makes it possible to meter something closer to real value than ever before.

Links

Follow Ollie on X - https://x.com/ollieforsyth

Follow Linda on X - https://x.com/YayLT

Vist Webflow - https://webflow.com

Episode Partner - Discover Harmonic, your go-to startup database - https://harmonic.ai

Previous episodes include

If you enjoyed this episode, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

This Week In Startups

Ollie Forsyth — Sun, 13 Sep 2026 12:16:43 GMT

Hey readers,

Welcome back to your weekly tech digest, a quick look at the biggest stories shaping the tech world this week.

This past week: the AI funding race showed no signs of slowing. Poolside raised $6B led by Nvidia, Mistral closed a $3.5B round at a $24.4B valuation, and Cognition raised $2B at a $48B valuation. Harvey meanwhile, hit a $15.5B valuation just months after its last raise. On the consolidation side, Bending Spoons acquired Miro for $1.3B, a fraction of its 2022 $17.5B valuation, another sign that even well-funded SaaS companies are getting bought for a few multiples of revenue rather than the premiums they once commanded a few years ago.

Y Combinator also held its Summer 2026 Demo Day earlier this week, with a couple of stats worth noting: 52% of startups are B2B, with founders increasingly building the infrastructure behind AI agents rather than the agents themselves, from routing and context management to evaluation and automation. See the full batch breakdown below.

This edition covers:

👀 The latest AI news
🗝️ Inside the YC Summer 2026 batch
💰 Top startup news & funding rounds (sourced using Harmonic)
🤝 Acquisitions including a look at the Bending Spoons portfolio
🎤 Must-watch interviews from our podcast show

and much more.

Inside the current state of AI

Harvey hits $15.5B valuation, months after reaching $11B having just raised another $550M!
Instacart launched an AI grocery shopping assistant called Clementine.
Anthropic CEO Dario Amodei called for AI labs to slow the pace of frontier-model development, proposing independent evaluators, coordination between labs, and international cooperation.
OpenAI CEO Sam Altman also said he doesn’t expect OpenAI to IPO in 2026 and described even a 10% risk of AI-driven human extinction as unacceptable.

Top startup news & funding rounds

We take a closer look at some of the hottest startup rounds of the week:

Poolside, the AI coding agent suite for development environments, raised $6B led by NVIDIA.
Mistral, the enterprise and government-focused frontier AI model developer, raised $3.5B Series D at a $24.4B valuation co-led by Samsung Electronics, EQT Group, European Innovation Council, and PSG.
Cognition, the autonomous software engineer behind Devin, raised $2B Series E at a $48B valuation co-led by Avenir, Andreessen Horowitz, Founders Fund, General Catalyst, and Accel.
Motive, the AI-powered fleet management and driver safety platform, raised $1.3B led by General Catalyst.
Stoke Space, the fully reusable rocket system developer, raised $1B Series E at a $10B valuation co-led by Point72 Ventures and Spark Capital, with Y Combinator, Glade Brook Capital, and Woven Capital.
Positron AI, the specialized AI inference hardware maker, raised $875M Series C at a $5B valuation co-led by Atreides Management, SemiAnalysis, NEA, Andra, and Valor Equity Partners, with Qatar Investment Authority, Cisco Investments, and Hudson River Trading.
Mach Industries, the ML-driven defense manufacturing company, raised $600M Series C extension at a $3.7B valuation, with Sequoia Capital, Ribbit Capital, Bedrock Capital, and Infinite Capital.
The Exploration Company, the European reusable space capsule and rocket engine developer, raised $450M Series C co-led by EQT Group, Bessemer Venture Partners, European Innovation Council, and Atomico, with Balderton and Plural participating.
Tripo AI, the AI platform that generates 3D models from text, images, or sketches, raised $447M Series B extension, with DreamWorks Animation, ThunderSoft, CMC Capital Partners, and BlueFocus.
Celero Communications, the coherent optical DSP chipmaker for AI data center interconnects, raised $275M Series C at a $3B valuation co-led by CapitalG, Valor Equity Partners, and Atreides Management, with Sutter Hill Ventures.
Encoded Therapeutics, the gene therapy developer targeting neurological disorders, raised $275M Series F led by GV, with ARCH Venture Partners, SoftBank, Venrock, RTW Investments, and Illumina Ventures.
Covenant, the long-range precision strike weapons system developer, raised $250M at a $1B valuation, with Andreessen Horowitz, Founders Fund, Lux Capital, Lightspeed, 8VC, Altimeter, and Aleph.
Cylake, the AI-native on-prem cybersecurity platform, raised $245M convertible note co-led by Redpoint, Lightspeed, and Picture Capital.
Solstice Oncology, the clinical-stage immuno-oncology company focused on neoadjuvant treatment, raised $225M Series A led by RA Capital Management, with Canaan and Forbion.
Forus, the AI-powered medication access and prior authorization network, raised $150M Series C at a $3B valuation led by Bain Capital Ventures, with Thrive Capital, Accel, General Catalyst, Redpoint, and BoxGroup.
Ayar Labs, the co-packaged optics chipmaker connecting GPUs across racks, raised $150M with NVIDIA, Intel, AMD, MediaTek, and Wiwynn.
Beep, the autonomous shuttle and driverless mobility network, raised $130M Series B led by Mobileye.
Kapital, the Mexican financial platform for business payments, credit, and international transfers, raised $125M Series C extension co-led by Tru Arrow Partners and Fasanara Capital, with Cervin Ventures and Niya Partners.
TAR, the American energy infrastructure builder, raised $120M Series A at a $1B valuation led by Spark Capital, with Buckley Ventures and Align.
Carrum, the Indian fleet operator received a $10M investment from Uber at a $168M valuation

(sourced using Harmonic)

Acquisitions

Meta acquired Swedish artificial intelligence startup Stilla AI.
Bending Spoons acquired collaboration tool Miro for $1.3B, 90% below its $17.5B 2022 valuation, despite Miro having raised a $400M Series C. It's another case study in the current state of SaaS: companies with hundreds of millions in revenue and cash-rich balance sheets are still only fetching 2-3x revenue on exit. Who's next on Bending Spoons' list?

Here’s their acquisitions to date 👇

2018: Splice (Price N/A)

2021: Remini (Price N/A)

2022: Filmic (N/A), Evernote ($200M)

2024: Mosaic Group ($100M), Meetup (N/A), Issuu ($100M), WeTransfer ($500M), StreamYard + Vimeo ($1.38B)

2025: Brightcove ($233M), Komoot ($350M), Harvest (N/A), AOL (N/A)

2026: Tractive ($781M), Eventbrite ($500M), Airtable ($1.3B), Miro ($1.3B).

Inside the YC Summer 2026 batch from this week’s demo day

Y Combinator’s Summer ‘26 Demo Day took place on Thursday so we’re back with a full overview of the latest startups in the batch, what they’re building, the categories they’re focused on and the trends behind them, as well as the partners who coached them.

So what have we learned about this batch so far?

52% of the batch are B2B startups, with founders building across productivity, the agent supply chain, and infrastructure tools for other companies rather than end users.
Consumer stays a minority category at around 5% of the batch, consistent with the last several cohorts.
The agent story is maturing: YC’s previous batch was defined by agent applications, while this batch of companies are building the infrastructure underneath them, routing, context management, evaluation, and automation tooling that makes an agent usable in production.
Batch size: 235 companies are in this cohort, a 20% increase from the previous batch.

Let’s explore the latest trends shaping this batch in full.

Elizabeth Holmes, founder of Theranos is back in the news: A24 has revealed, You Can See Everything, a documentary from Nathan Fielder and Lance Oppenheim covering the 34 days before she reported to prison in 2023. It premiered as a secret screening at Telluride and hits theaters October 16.

So, if she made a comeback, what would it look like?

Episode 3 of Billion Dollar Tea, our AI microdrama on fiction and non-fiction stories tells one possible comeback scenario!

Must-watch interviews from our podcast show

One of our latest podcast guests is one incredible founder: Michael Birch, co-founder of Bebo, who shares how he started and scaled the social network into one of the most popular sites of its time before selling it to AOL for $850 million in 2008, how he reinvented himself after Bebo, and what he’s building now that Bebo is closed for good after 21 years.

Previous episodes include

Superhuman CEO: The Strategy to Turn Standalone Apps Into a $13 Billion Suite | Shishir Mehrotra.
Europe Is Quietly Building the New Silicon Valley | Saul Klein.
$2 Billion Founder: The Secret to Building a High-Growth Startup in 24 Months | Tanay Kothari.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Why most people are underusing AI and why the winners won’t just be OpenAI and Anthropic | Reid Hoffman.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).

Watch and subscribe now!

If you enjoyed this edition, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Pegasystems

Ollie Forsyth — Fri, 11 Sep 2026 13:32:50 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

About Pegasystems
Pega is an enterprise workflow automation company Alan founded in 1983 on the tagline “build for change.” The company has been public for 30 years, does close to $2 billion in annual revenue with over 5,000 employees, and serves roughly 800 of the world’s largest enterprises, including four decades of continuous relationships with the same banks it started with.

Watch the episode now

Throughout this episode, we also cover why Pega put a literal “no token cost” badge on stage at Pega World and how it can actually afford that promise, the vector database strategy Pega chose over building its own foundation model so it can move freely between OpenAI, Claude, and Gemini, why Alan thinks the “cost cap on tokens” debate misses the point because competition, not government, is what brings prices down, and the math behind why a workflow running on a CPU is thousands of times cheaper than a reasoning session on a GPU.

As the SaaS apocalypse debate rages on, we asked Alan directly whether AI coding tools are about to eat Pega’s forty-year business. We learn why he thinks the real moat was never the code, “they can compete on code, but they can’t compete on trust” - why his succession plan is basically “I’m not leaving” (his words: “my exit strategy is going to be a pine box”), and much more.

Available everywhere you listen to podcasts: YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Alan Trefler (Pega)

300KB ∙ PDF file

Timestamps

(0:00) Meet Alan Trefler
(2:23) What Is Pegasystems?
(5:26) How to Build Trust Today?
(6:32) Being Public for 30 Years
(7:59) The First Year of Going Public
(11:56) Navigating the Frothy AI Market (?)
(14:00) Stock Price Ups and Downs
(20:00) Measuring the Cost of AI Compute
(22:30) Why Absorb the Cost of Tokens for Customers?
(26:57) Will There Be a Cost Cap on Tokens?
(28:02) Treating AI Models with Fungibility
(31:14) The SaaS Apocalypse
(34:35) Does AI Replace Trust?
(35:37) Inside Pega's Opportunities & Challenges
(39:14) New Tools Outpacing Entire Revenue Streams
(41:50) AI Talent & Agent Managers
(43:46) New Roles Being Hired Today
(45:46) Succession Planning
(47:04) Ollie Joins Alan as Chief of Staff
(48:45) Preparing for an Earnings Call
(50:05) Being a Hands-On Leader
(52:28) What Alan Does Outside of Work
(55:03) One Board Member You Would Have on Your Board
(56:11) Alan's Legacy

Lessons from this episode with Alan

1. Customers do not need to pay for tokens
Alan’s explanation for why Pega doesn’t charge customers for tokens: design the recipe once in the test kitchen, serve it a million times, no need to reinvent the dish every order. Concrete framing of design-time vs. runtime AI cost.

2. “A token is just an example of the BS that’s going on”
The token-vs-words rant, why calling it “tokens” instead of “words” obscures cost on purpose.

3. Bubblicious behavior
"I think the reality is that a bunch of what we're seeing is I would describe as bubblicious behavior. And so there is going to need to be some reallocation and there will be corrections."

WATCH THE FULL EPISODE

4. “My exit strategy is going to be a pine box”
Some founders just keep going until the end of life and Alan said he is one of those.

5. The 27-slide self-congratulatory deck
His take on organizational culture and why he actively discourages “brilliant group” presentations.

6. Don’t go public too early
Pega was so advanced but also very early in their journey when they decided to go public. Looking back, Alan says - ‘‘Don’t go public too early’’

7. Trust vs. transactional relationships
“The people in transactional environments only show up when there’s a transaction.”

Links

Follow Ollie on LinkedIn: https://www.linkedin.com/in/ollieforsyth

Follow Alan on LinkedIn: https://www.linkedin.com/in/alantrefler

Visit Pegasystems: https://www.pega.com

Episode Partner - Discover Harmonic, your go-to startup database: https://harmonic.ai

Previous episodes include

If you enjoyed this episode, support our work by clicking ❤️ and 🔄 at the top of this post.

Felicis Ventures

Ollie Forsyth — Tue, 08 Sep 2026 13:50:45 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

About Felicis
Felicis is a venture capital firm that backs iconic founders starting at Seed, with early bets on Notion, Canva, Shopify, Adyen, n8n, Supabase, and Mercor. Over 20 years, Felicis-backed companies have driven more than $300 billion in market value, and Aydin has appeared on the Forbes Midas List for 13 consecutive years.

Watch Now - Investing In 50 Unicorns

Throughout this episode, we also cover the origin of Felicis’s 1% founders pledge, a co-CEO’s idea borrowed from tennis mental coaching that now covers health therapy and coaching for 100+ founders with no strings attached; why Felicis is obsessed with “global resilience” and the four markets (space, defense, manufacturing, energy) big enough that 1% share still returns a fund; and the stat from Felicis’s own data showing the top 1% of exits have doubled in value every five years while the bottom 90% have stagnated.

As venture faces its own AI challenges, we asked directly whether AI is coming for the junior analysts and associates learning the trade underneath Aydin. We learn why he thinks trust, not analysis, is the one thing AI can’t replicate, and why founders keep telling him they picked Felicis for the person across the table, not the term sheet, and much more.

Available everywhere you listen to podcasts: YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Aydin Senkut (Felicis)

403KB ∙ PDF file

Timestamps

(0:00) Meet Aydin Senkut
(2:21) The State of Venture Today
(3:49) About Felicis Ventures
(6:35) Google's First Early Employees
(8:09) Lessons from Larry and Sergey
(10:11) Building Google from Nothing
(13:10) Great Investors: Operators or Founders?
(15:57) 13x on the Forbes Midas List
(17:28) What Makes Felicis Successful
(20:11) Navigating Crucible Bets
(22:16) Characteristics of Unicorn Companies
(25:22) Coaching Founders to Stay Disciplined
(28:30) Longevity and Stewardship
(32:56) The 1% Founders Pledge
(40:21) Founders' Hardest Challenges Today
(43:07) Building an Enduring Fund
(46:52) From Meeting to Term Sheet
(49:02) Inside Felicis
(51:32) Are LPs Concerned About Venture?
(57:28) AI and the Future of Young Venture Talent
(1:00:29) The Opportunity for Emerging Managers
(1:03:39) How Aydin Spends Time Outside Venture

Lessons from this episode with Aydin

1. Being ruthlessly focus: “Cut nine things off your list”
The core operating principle: it feels productive to work on 10 things, but real focus means having the courage to kill nine of them and go all-in on one.

2. “Your weakness can be your strength”
Aydin wasn’t an engineer, never worked at a big tech company, and built a philosophy around not needing technical depth, just needing to understand what makes a company succeed.

3. Authenticity, trust, and doing the homework: “The personality of the investor really matters”
Three concrete, teachable behaviors for building trust fast: no hidden layers, a track record of not going against people, and showing up with a prepared, original point of view.

4. Market-sizing lesson for founders and investors: “Chase markets where 1% share still wins”
The math behind why Felicis is focused on “global resilience”: pick markets big enough that even a small share of them clears the bar for a great outcome.

5. Where does AI leave young talent in venture?
Why he believes relationships, not analysis, are becoming the scarce resource in venture, and the “orchestra conductor” framing for how humans and AI should actually divide labor.

Previous episodes include

If you enjoyed this episode, support our work by clicking ❤️ and 🔄 at the top of this post.

This Week In Startups

Ollie Forsyth — Sun, 06 Sep 2026 15:14:56 GMT

Hey readers,

Welcome back to your weekly tech digest, a quick look at the biggest stories shaping the tech world this week.

This past week: the AI infrastructure race continued to accelerate. Crusoe raised $3B at a $30B valuation, while Wonderful raised $550M at a $5B valuation and Instinct raised $250M at $2.5B. At the same time, Nvidia agreed to acquire Hugging Face for $12.9B, bringing one of the world’s largest AI infrastructure companies closer to one of the biggest open-source AI ecosystems.

The trend is becoming increasingly visible at the startup level, too. In YC’s latest batch, 52% of startups are B2B, with many founders moving beyond building AI agents themselves to focus on the infrastructure that powers them, from routing and context management to evaluation and automation. See the latest batch in full below.

The trend is overall hard to miss: the next phase of AI may be less about building another application and more about building the infrastructure that makes AI useful at scale. As models themselves become increasingly capable, the opportunity may increasingly sit in everything around them, the compute, data, tooling, security and physical infrastructure needed to turn intelligence into something businesses can actually use.

This edition covers:

👀 The latest AI news
🗝️ Trend of the week: Inside the YC Summer 2026 batch
💰 Top startup news & funding rounds (sourced using Harmonic)
🤝 Acquisitions
🎤 Must-watch interviews from our podcast show

and much more.

A message from today’s partner: Mostly Metrics

The CFOs of Robinhood, Figma, Plaid, and Major League Baseball all read the same free newsletter every week.
Mostly Metrics is trusted by 75,000 readers and regularly breaks down metrics and monetization models the best CFOs in the world actually use to run their companies. Fundraising. Headcount planning. Pricing. The stuff nobody teaches you until you’re already in the seat.
Subscribe to the Mostly Metrics newsletter to learn the latest strategies.

Inside the current state of AI

Oura has filed to go public according to its filing.
Adobe CEO, Shantanu Narayen steps down, being replaced by Anil Chakravarthy.
OpenAI launched its latest model, Astra.
Google launched AI voice features in Gmail, Docs, and Keep.
John Ternus officially became Apple’s CEO on September 1, 2026, succeeding Tim Cook, who transitioned to executive chairman of Apple’s board of directors after 15 years leading the company.
NYC banned generative AI for students in grades 2-K through 8, about 600,000 kids, while high schoolers keep limited access to approved tools. Companion chatbots are barred at every grade.
Wayve and Uber launched the first-ever autonomous rides in the UK.
Network School Kazakhstan is now officially open, with a packed crowd for the ribbon cutting, and government officials welcoming them.
Isar Aerospace made history on their second flight, becoming the first commercial space company from Europe to deliver satellites into orbit. Mission 'Onward and Upward' successfully lifted off and deployed payloads into orbit. A new chapter for European spaceflight is here!

Meet the latest YC startups ahead of demo day next week

Y Combinator’s Summer ‘26 Demo Day is under a week away. We’re back with a full overview of the latest startups in the batch, what they’re building, the categories they’re focused on and the trends behind them, as well as the partners who coached them.

So what have we learned about this batch so far?

52% of the batch are B2B startups, with founders building across productivity, the agent supply chain, and infrastructure tools for other companies rather than end users.
Consumer stays a minority category at around 5% of the batch, consistent with the last several cohorts.
The agent story is maturing: YC’s previous batch was defined by agent applications, while this batch of companies are building the infrastructure underneath them, routing, context management, evaluation, and automation tooling that makes an agent usable in production.
Batch size: 235 companies are in this cohort, a 20% increase from the previous batch.

Let’s explore the latest trends shaping this batch in full.

Top startup news & funding rounds

We take a closer look at some of the hottest startup rounds of the week:

Crusoe, the AI cloud and data center developer powering OpenAI, Meta, and Oracle, raised $3B at a $30B valuation co-led by Atreides Management and Valor Equity Partners, with Mubadala Capital.
Wonderful, the enterprise AI operating system, raised $550M Series C at a $5B valuation led by Insight Partners, with Salesforce, Index Ventures, IVP, and Bessemer Venture Partners.
Instinct, a proactive AI assistant that texts, calls, and handles everyday tasks, raised $250M in a Series B co-led by Index Ventures and Benchmark at a $2.5B valuation.
REGENT, developer of electric Seaglider vessels for coastal transport, raised $240M in a Series B co-led by Mare Liberum and AE Ventures, with Founders Fund, Lockheed Martin Ventures, Japan Airlines, and DCVC.
Félix, a WhatsApp-native remittance platform, raised $200M in a Series C led by Andreessen Horowitz, with QED Investors and General Catalyst.
Blank Street, the coffee and matcha chain with 100+ compact storefronts, raised $105M led by General Atlantic at a $650M valuation, with Left Lane Capital, General Catalyst, and Tiger Global.
HiddenLayer, an AI security company protecting agentic and generative AI apps, raised $100M in a Series B led by Delta-v Capital, with Ten Eleven Ventures, Morgan Stanley, Microsoft’s M12, and Booz Allen Ventures.
Scan.com, a medical imaging network for booking diagnostic scans, raised $90M in a Series C led by Noteus Partners, with Aviva, Concord Health Partners, YZR Capital, and Oxford Capital.
Odyssey Energy, financing distributed renewable energy in emerging markets, raised $74M from Broadscale Group, FMO, Al Mada Ventures, Union Square Ventures, and FJ Labs.
Conveo, an AI consumer research platform running qualitative interviews at scale, raised $50M in a Series A from DST Global Partners, Balderton Capital, 6 Degrees Capital, and Y Combinator.
AIR Security, an Israeli AI cybersecurity startup, raised $50M across two seed rounds backed by Sequoia and Greenoaks.
Deep Cogito, a post-training research lab building AI that teaches itself to improve, raised $43M in a Series A led by TQ Ventures, with Benchmark, Nexus Venture Partners, Atreides, and Zscaler.
Empirik.ai, an AI infrastructure reliability platform out of stealth, raised $21M in a seed backed by Sequoia, Canapi Ventures, and Alumni Ventures.
XDOF, a robotics data startup collecting real-world teleoperation data to train general-purpose robots, is in talks for a Series B at ~$1.2B valuation led by 8VC, just three months after exiting stealth.
Ultrahuman, the Indian smart ring maker expanding into health intelligence, raised $70M led by Qualcomm Ventures, a bet to turn smart rings into wrist-free wearable computers.

(sourced using Harmonic).

Acquisitions

Nvidia confirms the Hugging Face acquisiton for $12.9 billion.
Action camera maker GoPro to be acquired by Starman Optical.
Palo Alto Networks acquired Console for $500M.
Adobe acquired Indian AI marketing startup Rilo.

Must-watch interviews from our podcast show

Will GPUs become a new asset class? Our latest podcast guest may have some answers.

Stephen Balaban, co-founder and CTO of Lambda, joins us to explain why a $40,000 monthly AWS bill was the best thing that ever happened to the company in its early days, why he recently handed over the CEO title after fourteen years at the helm, and what is really going on in the GPU market, and whether we should be concerned or excited about the opportunity.

Previous episodes include

Superhuman CEO: The Strategy to Turn Standalone Apps Into a $13 Billion Suite | Shishir Mehrotra.
Europe Is Quietly Building the New Silicon Valley | Saul Klein.
$2 Billion Founder: The Secret to Building a High-Growth Startup in 24 Months | Tanay Kothari.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Why most people are underusing AI and why the winners won’t just be OpenAI and Anthropic | Reid Hoffman.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).

Watch and subscribe now!

NEW ECONOMIES Subscriber Offers

Hyperagent: Build AI agents in minutes. Get $1,000 in free credits here. Enter code NEWECONOMIES.
Attio: Tired of trying multiple CRM tools? Attio is the favorite, used by thousands of startups. Start using it here.
Harmonic is your go-to startup database provider.

If you enjoyed this edition, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Bebo

Thu, 03 Sep 2026 17:27:40 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Our latest podcast guest is one incredible founder: Michael Birch, co-founder of Bebo, who shares how he started and scaled the social network into one of the most popular sites of its time before selling it to AOL for $850 million in 2008, how he reinvented himself after Bebo, and what he’s building now that Bebo is closed for good after 21 years.

Watch now - Bebo Co-Founder: How to Build an $850 Million Social Network

During the episode, we learn about Michael’s first entrepreneurial attempt, Ringo.com (another social network), which landed 400,000 members in three months on a $6,000 database server; how Bebo grew to a million users in nine days before going quiet for six months, until a basic quiz feature cracked engagement; and how The Battery, a members’ club in San Francisco, became his next act after the acquisition.

We also explore an interesting question: what happens to a founder’s sense of identity when the company he and his wife, Xochi, spent years building is no longer theirs to run? Michael also shared that, although he’s technical at heart, he hasn’t written a line of code since January, building his latest venture almost entirely with AI, a glimpse of what building without a team could look like for the next generation of founders.

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Michael Birch (Bebo)

390KB ∙ PDF file

Timestamps

(0:00) Meet Michael Birch
(2:58) The Founding Story of Bebo
(8:55) Why Bebo Went Viral
(11:19) Bebo's First 12 Months
(21:40) How Much Has Social Actually Changed?
(23:58) Selling to AOL for $850M
(29:50) Reinventing Your Identity Post-Acquisition
(31:12) Starting The Battery
(38:47) Technology or Hospitality?
(39:54) Michael's Take on AI's Future
(45:00) Is IRL Being Affected by AI?
(47:55) Michael's Latest Venture: Bluebell
(1:02:08) Building With Your Spouse

Our notes from this episode with Michael

1. Virality gets people in, engagement keeps them.

Bebo stalled after launch despite a working social graph. A single “how well do you know me” quiz, which required non-members to create an account to see their results, fixed that and pushed growth to hundreds of thousands of users a day.

2. ‘‘We killed virality on purpose outside English-speaking markets.’’

Bebo blocked non-English IP addresses from viral features so it wouldn’t blow up in markets he couldn’t support, staying deliberately focused on the UK, US, and other English-language countries.

3. A password scraper added 40 million users a year.

Michael built a tool that logged into people’s Hotmail and Yahoo using their own passwords, pulled their address books, and mass-invited every contact. One overnight run alone added 100,000 members.

4. The $850M sale.

Although a life changing amount of money to retire on, Michael never actually met anyone from AOL until the sale was officially closed.

5. The day the sale closed, he was unemployed.

Michael and his wife were the only two people at the company not offered a job when AOL took over, despite three and a half years of building it together.

6. He thinks the technical moat he’s built his career on is gone by next year.

He hasn’t written a line of code since January 1st, building his new app Bluebell entirely by directing AI, and expects deep technical understanding to stop mattering for shipping real products within the year.

Links

Follow Ollie on X: https://x.com/ollieforsyth

Follow Michael on X: https://x.com/mickbirch

Sign-up to Michael’s Newsletter - The Long Way Back to Friends:

Sign-up to Bluebell - Michael’s latest venture: https://bluebell.social

About Bluebell
Bluebell is the social and messaging app Michael and Xochi Birch built as Bebo’s successor. It fuses DMs, group chats, and a feed into shared spaces called “pods,” with no public timeline, no follower count, and no feed of strangers to perform for.

Previous episodes include

If you enjoyed this episode, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Y Combinator Summer 2026 Batch

Ollie Forsyth — Tue, 01 Sep 2026 15:38:04 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Y Combinator’s Summer ‘26 Demo Day is just over a week away. We’re back with a full overview of the latest startups in the batch, what they’re building, the categories they’re focused on and the trends behind them, as well as the partners who coached them.

So what have we learned about this batch so far?

52% of the batch are B2B startups, with founders building across productivity, the agent supply chain, and infrastructure tools for other companies rather than end users.
Consumer stays a minority category at around 5% of the batch, consistent with the last several cohorts.
The agent story is maturing: YC’s previous batch was defined by agent applications, while this batch of companies are building the infrastructure underneath them, routing, context management, evaluation, and automation tooling that makes an agent usable in production.
Batch size: 235 companies are in this cohort, a 20% increase from the previous batch.

Let’s explore the latest trends shaping this batch in full.

Brought to you by ElevenLabs

NEW ECONOMIES subscribers get 12 months and 33 million characters of free access to ElevenLabs’ AI audio, equivalent to over 600 hours, through the ElevenLabs Grants Program. Built for early-stage startups and small teams, the grant removes cost as a blocker to building on the platform. Whether you’re creating an AI tutor, support assistant, or voice-powered game characters, ElevenLabs gives you the complete conversational agents platform to build it.
Sign up for a grant here, select NEW ECONOMIES and use code: 11neweconomies.
Eligibility: One grant per company. For full-time founders with startups with less than 25 employees. Existing enterprise customers are not eligible for grants. We are unable to award grants to companies that are building projects for children aged 18 or under.

Want to learn more? Join us in a few weeks: Meet the Eleven Labs team | Why You Should Be Using ElevenLabs.

The history of Y Combinator

In just over a week, 230+ startups from the latest Y Combinator batch will present to investors at the Summer ‘26 Demo Day, and this may be the strongest batch yet. After meeting several founders in recent weeks, one trend is very clear: the founders are doubling down on AI, focusing on B2B as well as categories covering industrials such as robotics and defense.

Ask any founder or investor, and they have undoubtedly heard of Y Combinator, one of the world’s most renowned startup accelerators, with a mighty track record. Founders who raise from them are 45% more likely to raise a Series A and the accelerator turned 21 years old earlier this year. Throughout that time, they have achieved some incredible stats, including:

$1.3T+ in combined market value has been created.
6.5% of YC startups become unicorns.
A quarter of all YC-backed unicorns achieve a valuation of over $10 billion.
50% of YC startups are still operating after 10 years.

So with the latest batch wrapping up, many of these startups are poised to rise, let’s meet them!

Meet the Summer ‘26 Batch

With a lot of eyes on YC companies, we wanted to showcase them through a wider infographic to help you navigate the areas that most interest you. Meet the 235 startups from YC’s latest batch who are building across the following categories:

B2B
General B2B Software, Infrastructure, Engineering & Product, Security, Supply Chain & Logistics, Analytics, Operations, Finance & Accounting, Productivity, Sales, Recruiting & Talent, Marketing, Legal, and Retail.

FinTech
General FinTech, Insurance, Asset Management, Banking & Exchange, and Consumer Finance.

Consumer
General Consumer, Gaming, Content, Social, and Job & Career Services.

Healthcare
General Healthcare, Drug Discovery, Consumer Health & Wellness, and Healthcare Services.

Industrials
General Industrials, Manufacturing and Robotics, Defense, Aviation and Space, as well as Energy and Climate.

Disclaimer: Most startups are listed, but companies still in stealth without a public logo may have been excluded.

*This market map was fully designed by a human.

Access the companies in full

Discover the batch in full below!

Discover the startups

Trends from this batch

We’re tracking the latest trends across batches, comparing Spring 2026 (the previous batch) with the current Summer 2026 batch. Here’s a quick look at which categories the startups are building in:

Batch size: 196 → ~235 companies (+20%)
B2B: 59.2% → 52.3% (−6.9%)
Industrials: 12.8% → 23.0% (+10.2%, the biggest shift)
Fintech: 10.2% → 6.8% (−3.4%)
Healthcare: 8.7% → 8.5% (essentially flat)
Consumer: 6.6% → 5.5% (−1.1%)
Real Estate & Construction: 2.0% → 2.6% (+0.6%)
Defense: 6 → 8 companies (+33%)
US-headquartered: ~94% → ~90% (slightly more international)

Biggest trends shaping this batch

1. Physical AI is the headline story
Industrials is now 23% of the batch, the second-largest category after B2B, up from just 12.8% last batch. This isn’t just robotics-as-a-vertical; it’s split across humanoid/home robots (Nori, OS3, NORI), warehouse and data-center robotics (Manifold, Proprio Robotics), and robot evaluation/training infrastructure (Robocurve, Instance, Markov). The through-line: AI agents are moving off screens and into physical labor.

2. Defense is a real, sizeable opportunity
At least 8 companies are building precision munitions, drone swarms, counter-UAS systems, and tactical situational-awareness hardware (ISENGARD, Earendil Robotics, Greypoint Industries, GUILD, Hop Aero). This is no longer a small bet these companies and founders are taking, it’s a consistent, well-represented vertical which many founders and investors are paying attention to.

3. “Agent-native” back-office replacement is everywhere
Dozens of companies pitch themselves as replacing entire functions with AI, not just augmenting them: agent-native accounting firms (Last Accounting Company, Rational, Billow AI), agent-native law firms (Osmaura, Erinys), agent-native ERP implementation (Trope), agent-native insurance carriers. The framing has shifted from “AI copilot for X” to “AI-run X.”

4. Infrastructure for AI agents themselves is maturing into a stack
Identity (Inkbox, OneCLI), payments (Agentcard), memory (Glen, Egoist Machines), evals (Archal, Instance), runtime debugging (Hyperprobe), and credential gateways are all distinct, well-populated niches, suggesting the “agent economy” has enough agents in production now to need real plumbing, not just experiments.

5. Compute economics as its own category
GPU marketplaces and liquidity products (Stoa), inference routing/cost-cutting (Conifer, Understudy Labs, OpenRelay), and specialized AI chips (Baud, Standard Machines, Lamb Labs), all responding directly to the cost of running frontier-scale AI at production volume.

6. Vertical-specific “AI-native” replacement of professional services
Beyond accounting and law: AI-native government affairs firms (Locke), AI-native McKinsey-style consultancies, AI-native radiology and clinical operations (Insurf). The pattern spans regulated, white-collar industries specifically, the ones with the most repetitive document/compliance work.

7. Solo founders and very small teams building hardware
Several one or two-person teams (Osseus, hardware intelligence, screenpipe) are shipping hardware-adjacent or infra-heavy products that historically required larger teams, suggesting that AI tooling is dramatically expanding what tiny teams can build.

Graphs - Categories YC’s Summer ’26 Batch Is Building In: B2B

Industrials

Healthcare

Fintech

Consumer

Discover previous batches

NEW ECONOMIES Subscriber Offers

Hyperagent: Build AI agents in minutes. Get $1,000 in free credits here. Enter code NEWECONOMIES.
Hostinger: Buy a domain, hosting service or build any website within minutes. Start building with 10% off here.
Attio: Tired of trying multiple CRM tools? Attio is the favorite, used by thousands of startups. Start using it here.
Harmonic is your go-to startup database provider.

If you enjoyed this edition, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

This Week In Startups

Ollie Forsyth — Sun, 30 Aug 2026 17:01:47 GMT

Hey readers

Welcome back to your regular tech digest, a quick snapshot of the most recent tech news worth paying attention to.

This week, NVIDIA agreed to acquire Hugging Face for $12.9B, bringing together the infrastructure powering AI with one of the largest open-source model ecosystems. At the same time, Anthropic introduced a hardware standard designed to let AI agents interact directly with physical lab and factory equipment.

Fundraising for startups is following the same direction. AI accounted for 86% of venture capital dollars in Q2, while a16z launched a $1.1B fund specifically aimed at accelerating the physical buildout of AI.

The pattern is becoming hard to miss: AI is shifting from something that lives inside a browser or app into infrastructure that runs businesses, controls machines, and increasingly touches the physical world. Perhaps the next phase of the AI race may not be about who builds the smartest chatbot, it may well be determined by who owns the infrastructure around the intelligence.

This edition covers:

🗝️ Trend of the week: Inside The Current State of AI
💰 Top startup news & funding rounds (sourced using Harmonic)
🤝 Acquisitions
👀 New VC funds launched
🎤 Must-watch interviews from our podcast show

and much more.

A message from today’s partner: Mostly Metrics

The CFOs of Robinhood, Figma, Plaid, and Major League Baseball all read the same free newsletter every week.
Mostly Metrics is trusted by 75,000 readers and regularly breaks down metrics and monetization models the best CFOs in the world actually use to run their companies. Fundraising. Headcount planning. Pricing. The stuff nobody teaches you until you’re already in the seat.
Subscribe to the Mostly Metrics newsletter to learn the latest strategies.

Inside the current state of AI

OpenAI loses exec, Chris Malone, former Head of Data Centers who has left the company alongside Kevin Weil (Former Chief Product Officer), Kate Rouch (Former Head of Marketing), Fidji Simo (Former CEO of Applications), Brad Lightcap (Former COO) & Denise Dresser (Former Chief Revenue Officer) who have all left the company in the last few moths.
Anthropic and Salesforce partner to launch ClaudeForce.
Nvidia acquired open source model repository Hugging Face for $12.9B.
a16z launched a $1.1B ‘Machine Age’ fund to ‘accelerate the physical buildout of AI.’
Anthropic launched Model Hardware Standard, a new protocol that lets AI agents directly control physical lab and factory hardware, a foundational move toward agents operating in the real world, not just software.
AI accounted for 86% of all venture dollars in Q2 2026, with three firms (a16z, Thrive Capital, Founders Fund) capturing ~48% of all capital raised by VC funds. First-time fund formation is on pace for its lowest year since 2016.

Top startup news & funding rounds

Stability AI, the generative AI studio building creative tools for music, gaming, and entertainment, raised $76M in a Series B backed by EA, Sony Music, Universal, Warner, Coatue, and Eric Schmidt.
Repodo, the world’s first AI-native audit firm raised $9.5M led by Hedosophia - one of the largest pre-seed rounds in Nordic history.
Instinct, a proactive AI assistant that texts, calls, and handles everyday tasks, raised $250M in a Series B co-led by Index Ventures and Benchmark at a $2.5B valuation, reaching unicorn status four months after incorporation.
Pennylane, the Paris-based accounting and financial management platform for European SMEs, raised $205M led by TCV and Blackstone Growth at a $4.25B valuation, one of Europe’s largest fintech rounds of the year.
Airbound, a Bangalore-based drone logistics company building blended-wing tailsitter aircraft for last-mile delivery, raised $37M in a Series A led by Greenoaks with Lightspeed, DoorDash, and Lachy Groom participating.
Wrtn technologies, a Seoul-based generative AI platform for everyday productivity and entertainment used across South Korea, raised $72M in a Series C backed by Goodwater Capital, Korea Development Bank, and Antler.
Pearl, an AI diagnostic tool that analyzes dental X-rays and CBCT imagery to detect pathologies and automate clinical documentation, raised $58M in a Series B led by Left Lane Capital.
Adaptyv, a Swiss protein engineering platform that automates expression, binding, and thermostability analysis to validate AI-designed proteins, raised $40M in a Series A led by Highland Europe with Y Combinator.
Transfyr, an API layer that makes scientific lab experiments observable and reproducible for scale-up, raised $25M in a Seed led by General Catalyst with Lux Capital and SV Angel.
Keenable, building a web-scale knowledge index purpose-built for AI model retrieval, raised $26M in a Seed led by Accel with Google, Amazon, Databricks, Snowflake, and ClickHouse.
Metriport, a healthcare data exchange platform that converts messy medical records into clean FHIR-native data for care teams, raised $26M in a Series A led by Matrix with Y Combinator.
Ringg AI, a Bangalore-based AI voice and chat agent platform for enterprise customer interactions, raised $15M in a Series A led by Peak XV Partners.
TerraBlaster, using laser-induced breakdown spectroscopy to give farmers real-time soil nutrient readings in the field, raised $13M in a Seed led by Ulu Ventures with Khosla Ventures.
Primero, building enterprise AI systems for finance, HR, and operations across Latin America, raised $12M in a Seed co-led by General Catalyst and KASZEK with 8VC, Conviction, FEMSA, Grupo Bimbo, and Aeromexico.
Arga Labs, a sandbox that lets teams test AI agents against realistic digital twins of third-party services before going live, raised $10M in a Seed led by General Catalyst with Emergence Capital.
Atorie, connecting luxury factories directly to consumers using AI for trend prediction and demand testing, raised $9.5M in a Seed co-led by a16z speedrun and Lightspeed.
Neno, a full-stack AI financial services company embedding payroll, accounting, and tax for entrepreneurs in Amsterdam, raised $8M in a Seed led by AlleyCorp with Coinbase, Deel, PayPal, Mollie, and Hugging Face.
Mara, building autonomous counter-drone systems that detect and neutralize FPV threats in GPS-denied environments, raised $7M in a Pre-Seed led by Khosla Ventures with a16z Speedrun.
Willow, an AI dictation app that converts speech to text in any text field across platforms, raised $4M in a Seed led by BoxGroup with Goodwater Capital.

🤝 Acquisitions

NVIDIA’s Hugging Face deal is the headline of the week, and arguably the month. At $12.9B, NVIDIA is buying the distribution layer for open-source AI models, giving it control over both the hardware (GPUs) and the software ecosystem where developers discover, share, and deploy models.
Vanguard + Altruist signals wealthtech consolidation, the $12T asset manager is paying a significant premium to embed itself in the daily workflow of independent advisors, moving from investment product manufacturer to infrastructure owner.

OpenAI launched a $400M second venture fund, this time funded entirely from its own balance sheet, more than double its $175M 2021 debut that was backed by outside investors including Microsoft.
Navisalma, the new deep tech firm from Einride co-founders Robert Falck and Linnéa Kornehed Falck, launched targeting €450M for European deep tech with an operator-led, hands-on model.
MassMutual Ventures launched a $150M Climate Tech Fund II, doubling down and pushing its total climate commitment to $300M.
Capital F raised a $17M debut fund with goal to back the future of the ‘female economy’
Fiat Ventures raised $35M for Fund II.
African venture fund, Ventures Platform raised an oversubscribed $84M for Fund II.
Wave Ventures, run entirely by investors under 25, closed a €10M Fund III (5x its last) to lead Nordic pre-seed rounds and launched a €240K founder grant.

🎤 Must-watch interviews from our podcast: Superhuman CEO: The Strategy to Turn Standalone Apps Into a $13 Billion Suite

Shishir Mehrotra, CEO of Superhuman, joins NEW ECONOMIES to explain why he renamed a 16-year-old company mid-flight instead of just adding a new label on top, why he built his own board by ranking every past boss he’s ever had instead of chasing famous names, and why the real threat to a legacy SaaS category isn’t a faster competitor but the coordination problem AI agents are about to make bigger, not smaller.

Previous episodes include

$2 Billion Founder: The Secret to Building a High-Growth Startup in 24 Months | Tanay Kothari.
Canva Co-Founder: AI Is Useless Without This 1 Skill | Cam Adams.
Hollywood is Secretly Replacing Films With AI | Justine Moore.
Why most people are underusing AI and why the winners won’t just be OpenAI and Anthropic | Reid Hoffman.
Base44 Founder: The Strategy Behind Building a Million Dollar Company Solo | Maor Shlomo (Base44) & Mark Tluszcz (Wix).

Watch and subscribe now!

NEW ECONOMIES Subscriber Offers

Hyperagent: Build AI agents in minutes. Get $1,000 in free credits here. Enter code NEWECONOMIES.
Hostinger: Buy a domain, hosting service or build any website within minutes. Start building with 10% off here.
Attio: Tired of trying multiple CRM tools? Attio is the favorite, used by thousands of startups. Start using it here.
Harmonic is your go-to startup database provider.

If you enjoyed this edition, help sustain our work by clicking ❤️ and 🔄 at the top of this post.

Lambda

Ollie Forsyth — Thu, 27 Aug 2026 13:02:34 GMT

Subscribe to stay ahead of technology trends. Never miss future editions.

Will GPUs become a new asset class? Our latest podcast guest may have some answers.

About Lambda
Lambda, branded “The Superintelligence Cloud,” is an AI infrastructure company founded in 2012 that builds and operates gigawatt-scale AI factories and GPU supercomputers for training and inference, serving AI researchers, enterprises, and hyperscalers.

Watch now - Stephen Balaban, co-founder at Lambda

During our latest episode, we also learn how Lambda pivoted from a facial-recognition startup, to a data-collection hardware product called Lambda Hat, to an AI consulting shop nobody would fund, to the largest generator of Deep Dream images on the internet, before finally landing on an opportunity that very few saw coming: GPU cloud businesses.

We also explore an interesting trend and opportunity: could GPUs become a new institutional asset class? We also discuss the lessons Lambda learned from partnering with one of the world’s most talked-about companies: Nvidia.

Watch or listen now across YouTube, Apple Podcasts, Spotify, and X

Download the transcript 👇

NEW ECONOMIES Podcast: Stephen Balaban (Lambda)

278KB ∙ PDF file

Watch the episode in full

Timestamps

(0:00) Meet Stephen Balaban
(2:13) The State of GPUs
(4:40) 2026: Is This the Breakthrough Year?
(6:27) Starting Lambda in 2012
(11:14) Spotting the GPU Opportunity
(16:42) How Stephen Stays Focused
(18:47) Starting a Company with Your Brother
(21:44) Stepping Down as CEO
(27:52) How to Find a New CEO
(29:34) Not Raising from Traditional VCs
(31:30) Where Does the GPU Opportunity Go Next?
(34:35) GPUs Becoming an Asset Class
(36:15) What Keeps Stephen Up at Night
(39:33) How Many AI Models Will There Be?
(41:15) Partnering with Jensen Huang
(46:21) How Technology Has Changed

Our notes from this conversation

1. Pivoting isn’t a failure mode for Lambda, it’s the operating model.
Since 2012 the company has run through roughly half a dozen pivots: facial recognition software, an AR data-collection hardware product called Lambda Hat, a failed “Accenture for AI” consulting play no VC would fund, and Dreamscope, an app that became the largest generator of Deep Dream images on the internet before the fad died. Stephen’s rule: try something, if it works keep doing it, if it doesn’t move on.

2. An expensive AWS bill accidentally invented the company’s real business.
At the peak of the Deep Dream craze, Lambda was paying $40,000 a month to Amazon, nearly enough to sink it. An early investor pushed the team to build their own servers instead. The $60,000 CapEx bet to build workstations wiped out that $40,000 monthly OpEx completely. Workstation sales went from $35,000 in March to $140,000 in May, on the way to $3 million in revenue that first year. The GPU cloud business was the byproduct of a cash crunch, not the plan.

3. GPUs are becoming an asset class, not a depreciating expense.
Short sellers have argued GPUs carry a three-year usable life. Lambda’s counterevidence: V100s launched in 2017 are still nearly fully sold out in its cloud today, generating cash flow almost a decade later. Balaban’s comparison is insurance companies parking their float in power plants and toll roads, stable infrastructure that institutional and eventually retail capital moves toward as a category matures.

4. Starting a company with your brother for moral support.
As a solo founder, your mood on any given day is the company’s mood that day. A co-founder averages two signals instead of riding one. Stephen credits this as much as anything for surviving Lambda’s early years, run out of a small Chinatown apartment with his brother and co-founder, Michael.

5. The requirement to study computer science to build software is gone.
The clearest signal: anyone can now ship real tools built with Claude, people he describes as "meant to be programmers" who majored in something else. The real gate was never a CS credential, it was structured thinking, and that gate is now open to anyone.

Links

Follow Ollie on X: https://x.com/ollieforsyth

Follow Stephen on X: https://x.com/stephenbalaban

Visit Lambda: https://lambda.ai

Previous episodes include