Subscribe to stay ahead of technology trends. Never miss future editions.
Why I built the coding tool everyone dismissed — then watched it take over the Fortune 500
Michele Catasta, President and Head of AI at Replit, joins NEW ECONOMIES on why vibe coding is no longer just for hobbyists, how Replit went from a side project to having 85% of the Fortune 500 as users, and why 2026 is the year everyone becomes an agent manager.
In our latest podcast episode with Michele, we discuss Replit's origin story — a fifteen-year journey that started as an open source side project in 2011 and spent years building infrastructure in obscurity before the AI unlock that changed everything — and how launching the very first vibe coding agent on the market, before the term even existed, put Replit at the centre of a category it invented.
We also unpack why the SaaS apocalypse is real but overblown, why technical moats don't matter as much as execution moats, and how a crucible encounter with Replit's founder Amjad over a primitive AI demo set the course for what the product would eventually become.
If that’s not enough, we also explore why Replit scrambled an enterprise sales team almost overnight after Fortune 500 IT departments started knocking, the Visa partnership that lets anyone monetize a product they built in a single prompt, and why Michele believes the coding problem is almost solved — and what that means for where Replit goes next.
Watch or listen now on YouTube, Apple Podcasts, and Spotify
Download the transcript
Timestamps
(0:00) Michele Catasta
(1:25) The State of AI
(3:25) The Impact of AI Companies Going Public
(6:17) What Michele Is Most Excited About
(8:50) Replit's Founding Story
(18:40) Where Is Vibe Coding Going Next?
(21:00) The Role for PMs Today
(26:07) What Are Agent Managers?
(33:42) Are Technical Moats Relevant?
(36:31) Visa Partnership
(40:05) Ollie Joining as Chief of Staff
(43:48) How to Stay Disciplined
(46:27) Rapid Fire Round
Our notes from this conversation
1. Launch before the category has a name.
Replit shipped a vibe coding agent months before the term even existed. They didn’t wait for validation — they built, launched, and let users define the market. Lesson to founders is to just launch fast.
2. Sometimes being early means writing the playbook, which is totally okay!
Product–market fit as we know can takes ages to figure out. For a decade, Replit looked technically strong but commercially stuck. Those years built the infrastructure and conviction that made the AI moment possible. Remember, successes very rarely happens overnight.
3. Enterprise wasn’t actually a strategy for the team - their users pulled them there.
When employees started building internal tools, IT followed. Strong PLG can create demand before sales does.
4. Technical moats fade. Execution moats compound.
Features can be copied. Judgment can’t. A decade of learning what breaks, scales, and actually matters becomes the real advantage.
5. Coding is becoming the easy part.
The harder problem is everything around it — integrations, payments, governance, and infrastructure. The product becomes the platform.
6. 2026 is the year of ‘’the agent manager.’‘
Work shifts from creating everything yourself to directing, reviewing, and orchestrating multiple agents. The operating model changes before the job titles do.
Links
Subscribe to NEW ECONOMIES: @NEWECONOMIESPOD
Follow Ollie on X (https://x.com/ollieforsyth)
Follow Michele on X (https://x.com/pirroh)
Sign up to Replit: (https://replit.com)
