# $300M for a CLI?

The former CEO of GitHub just raised $60m at a $300m valuation in the largest seed round in developer tools history. Their debut product is nothing more than a CLI for git hooks.

Confused? I was too.

Clearly, this isn't the full story.

The product is called Entire (@entireio). You install it, it plugs into your git workflow, and from that point on every AI agent session you run gets captured in parallel with the code it created — the full transcript, the prompts, and the reasoning. The code + the thinking that produced it, linked permanently.

Thomas Dohmke (@ashtom), who spent four years as GitHub's CEO watching developers operate at a scale hardly anyone has seen, calls this the first step toward "a semantic reasoning layer over the lifecycle of a software project." Translated: Git tracks the change. Entire tracks the conversation that led there.

It's a start.

## Dev environments are called Terminals for a reason

By the time a task reaches a developer, it's the end of a long chain of reasoning, not the beginning. Most tasks originate in one-off calls, quarterly reviews, and Slack threads nobody bothered to save. By the time you start building, the agent session sees tactics, not strategy.

The full decision context is distributed across micro-interactions in dozens of systems. None of it is linked to the code it eventually affects. And none of it is recoverable 8 months later when someone asks why things are the way they are.

And AI is making it worse. Agents generate code faster than anyone can read, much less document. As throughput goes up, context gets left behind.

## Gossip

In distributed systems, a gossip protocol is how nodes share state without a central coordinator. Each node whispers what it knows to a few neighbors, those neighbors whisper it onward, and eventually the whole system converges organically toward a consistent view. Structured eavesdropping, propagating the right information to wherever it's needed.

The answer certainly isn't a knowledge base or a wiki. Those are libraries, and libraries sit on a shelf until read. What teams actually need is a protocol that eavesdrops on conversations wherever they happen and whispers the relevant context to whoever needs it, at the moment they need it.

Say you're working on a PRD and it surfaces a support ticket describing that exact feature. Or you're about to merge a PR and it tells you the database column you're using is about to get deprecated. Or you're in a pricing meeting and it recaps the last two times sales pitched this and what the prospects said. Not because you searched for it — because the protocol recognized you needed it.

Getting this right requires modeling decision state: what kind of call is being made, who has standing in the decision, what prior work is relevant vs. merely adjacent. It's a really hard, genuinely unsolved, and fascinating problem.

Entire hasn't said any of this publicly.

They've shared a CLI, a GitHub repo, and a promise that bigger things are coming.

Nobody outside that room knows exactly why investors wrote that check. But if this product grows into what I think it could, Entire may be planning the biggest shake-up in organizational dynamics we've seen in a long time. That may have been the plan all along.

And if it was, they'll have the commits to prove it.