Weave.
AI that
counts the
real work.

Engineering teams have been flying blind since the first commit.

For decades, engineering managers measured the wrong things. Lines of code rewarded bloat. Story points became a negotiation. Commit counts treated a typo fix and a distributed system redesign as equivalent. Nobody ever solved this cleanly - and then AI arrived and made everything worse.

When Cursor and GitHub Copilot and Claude Code started writing significant portions of every codebase, the old metrics didn't just become misleading. They became actively deceptive. A junior engineer running AI tools could out-commit a senior engineer doing architecture work. The numbers said one thing. Reality said something else entirely.

LLMs read every pull request. Then math happens.

Weave integrates with GitHub (and other tools - Jira, Slack, Linear, project management systems) in around 30 seconds. From there, it runs its own trained models on every PR: analyzing what changed, why it was complex, how much cognitive load it represented, and whether AI was involved. The output is a dashboard that tells you, per engineer and per team, how much work is actually getting done.

The central unit is the Weave Hour - Weave's own metric, designed to be resistant to gaming. Unlike story points, which teams set themselves, a Weave Hour represents what an experienced engineer would need to replicate a given change. It's calibrated to your codebase, not some abstract standard.

Two founders, one question: where did the work go?

Adam Cohen is a salesman turned operator. He grew up in Toronto, ran a lemonade stand on a wagon as a kid, and spent years at Top Hat building revenue operations from the ground up. His father worked in venture capital. Before Weave, he was VP of Operations at Causal - a spreadsheet-meets-financial-modeling startup that was acquired by Lucanet in 2024.

Andrew Churchill is the technical counterpart. MIT computer science and math. Employee number one at Causal, where he built the spreadsheet interface, the access control system, and the AI onboarding engine. He also spent time at Microsoft and Salesforce, and did research at MIT's CSAIL lab. When Adam and Andrew started Weave, Andrew already knew what it took to build AI-native products from scratch.

They interviewed 134 engineers before building the first model. Not 10 customers. Not a handful of design partners. 134 interviews. The pattern was consistent: engineering leaders knew AI tools were being used. They had no way to know if the investment was paying off. Weave was built to answer that question specifically.

They ran through Y Combinator's Winter 2025 batch in San Francisco. By the time the seed round closed - $4.2M, led by Moonfire and Burst Capital with YC participating - 25% of new YC companies were already running Weave on their repositories.

Stop guessing. Start seeing.

If you manage an engineering team and you're spending money on Cursor, GitHub Copilot, or Claude Code - Weave tells you whether those tools are actually accelerating delivery or just generating code that sits in review queues and creates new technical debt. You connect your repository and get your first dashboard in roughly 30 seconds.

For engineering leaders at startups, Weave surfaces who your best AI adopters are, so they can share practices with the rest of the team. For enterprise teams, it provides the DORA metrics and financial framing that engineering organizations need to make budget arguments to non-technical stakeholders. For individual engineers, Wooly - Weave's new AI agent - lets you ask questions about your own output trends in plain language.

Engineering analytics is a real category. Weave is the AI-native entrant.

The engineering intelligence space has incumbents. LinearB, Jellyfish, and Waydev all measure engineering work and have done so for several years. Pluralsight Flow (formerly GitPrime) has been in this space since before AI was a factor. The gap Weave is targeting is specific: none of those tools were built for a world where a meaningful percentage of each codebase is AI-generated.

Every company is buying AI tools. Almost none can tell if they're working.

The trend Weave is riding is straightforward: AI coding tools have gone from novelty to standard infrastructure in under two years. Most engineering organizations are now paying for Cursor, GitHub Copilot, Claude Code, or some combination. The procurement decision was often made quickly. The ROI question was left unanswered.

Weave is betting that engineering observability - actually understanding what your team is producing - becomes a category in its own right, the way security monitoring or performance monitoring did before it. The company's culture page describes a preference for tackling the hardest problems first, pricing month-to-month to keep themselves accountable, and maintaining a shared Slack channel with every customer to stay close to feedback. That last part is not typical at a seed-stage company.