The Q/A Layer for the AI Coding Era
TL;DR
Momentic is pitching itself as the verification layer for AI-generated software — Weiwayi and Jeff say they process over 1 million test runs a day for companies like Notion, Bolt, Quora, and Xero, and just raised a $50 million Series A from Standard Capital to scale engineering and go-to-market.
Their core bet is that code is becoming a commodity, while specs and validation become the real source of truth — Jeff argues that in the next 3 to 9 months, engineers will care less about reviewing TypeScript or React and more about defining plain-English requirements, edge cases, and success criteria.
AI coding increases the need for testing because verification is now the bottleneck — linters, human code review, and AI code review can check patterns, but they don’t prove the app works live; Momentic focuses on functional testing by acting like a user in a real browser.
Momentic’s differentiator is not just generating tests, but maintaining them over time — Jeff contrasts their approach with asking Cursor to generate Playwright code, which leaves teams with tens of thousands of brittle lines to update whenever the product changes.
Notion became a flagship customer through a late-night Twitter DM and now runs nearly 500,000 Momentic tests per day — after Simon from Notion posted that he wanted to describe tests in plain English, Weiwayi sent a Loom at 10 p.m., onboarded him that night, and eventually replaced a flaky Selenium-heavy workflow.
The founders think the engineer’s job is shifting from writing code to being a product-minded truth-finder — they still value technical depth, but increasingly emphasize adaptability, taste, ownership, and the ability to specify what should be built rather than hand-author every implementation detail.
The Breakdown
From YC to a $50M Series A
The conversation opens with the headline: Momentic, a W24 YC company, just raised a $50 million Series A from Standard Capital. Weiwayi says the timing was simple — they had a repeatable sales motion and wanted fuel to scale engineering and go-to-market. He also highlights Standard’s unusual structure: instead of a traditional board seat, Momentic joins a peer group of similar-stage founders for board meetings and shared learning.
Why testing is hated — and why that’s dangerous
Jeff gives the most vivid early anecdote from his Robinhood days, where the company grew from 300 to over 1,000 engineers and his team of eight was supposed to convince everyone else to write and maintain tests. The target was 80% code coverage with a 90% pass rate, and he says it was basically impossible because nobody felt rewarded for doing work customers never see. That’s the emotional core of the problem: testing feels like a drag until quality and reliability fail.
AI codegen is exploding, and verification becomes the choke point
As Har notes, the tooling carousel keeps spinning — Cursor, Claude Code, Codex — but the constant is that the volume of code is growing fast. Weiwayi says that creates a new bottleneck: not writing code, but proving it actually works in production. Linters can check style and patterns, and code review — human or AI — can inspect logic, but neither replaces the painful reality that many teams still do bug bashes by logging in and clicking around manually before release.
Where Momentic fits in the stack
Momentic lives at the functional testing layer: it impersonates users, clicks through the app, and verifies that key flows still work after a code change. The founders say customers are increasingly using Momentic through MCP integrations inside Cursor or Claude Code so coding agents can write features and immediately verify them in a real browser. Jeff says general-purpose agents are too slow and too clumsy for hard UI surfaces like rich text editors, drag-and-drop apps, and canvas-based interfaces, while Momentic’s own steps run in under 300 milliseconds and are built for debuggability.
The bigger thesis: truth-driven development
This is the philosophical center of the interview. Jeff argues that code should not be treated as the source of truth, because code contains bugs and implementation mistakes; instead, the source of truth should be a spec describing user journeys, edge cases, and success criteria. In that world, engineers become less obsessed with the generated code itself and more like product-minded requirement gatherers, while Momentic acts as the external source of truth that closes the feedback loop for coding agents.
Why Notion bought in
The Notion story has startup-movie energy: Simon from Notion tweeted that he wanted to describe tests in plain English, people tagged Momentic, and Weiwayi DM’d him at 10 p.m. from San Francisco with a Loom showing the product on his own Notion workspace. That quick demo turned into a broader proof-of-concept with the team. Before Momentic, Notion was juggling manual testing and a large Selenium suite that kept breaking on a product full of notoriously hard-to-test interactions; now they run almost half a million tests a day, and Momentic tests must pass before PRs can merge.
What they’re building next — and how they hire
On the roadmap, the founders call out Android, iOS, and desktop app support, but say the broader focus has actually narrowed around speed, integration, and making adoption feel like a “pit of success.” For hiring, Jeff says AI tools don’t magically turn weak engineers into great ones — they amplify people who are already adaptable, curious, and strong in ambiguity. Momentic is still only 13 people, and Weiwayi says the culture they care about is radical candor without being a jerk: clear feedback, broad product input, and being a pleasure to work with.
Founder backstories, early YC doubts, and pure competitive fire
Both founders came to engineering by rejecting other tracks — Weiwayi abandoned pharmacy after a brutally boring pharmacy camp, and Jeff walked away from chemistry at Cambridge after deciding isolated lab work didn’t fit how he wanted to build. They met through Dan Robinson, quickly realized they were building in the same space, and after Weiwayi literally stayed on Jeff’s couch for a week, they joined forces. By the end, their motivations split in an interesting way: Jeff talks about unlocking global productivity by solving code validation, while Weiwayi lands on a much more feral note — not just wanting to win, but wanting to destroy every competitor in the category.