TL;DR
- Current software development lifecycles were built for humans, not AI agents — and it's showing
- AI agent sessions are volatile: the reasoning behind code changes disappears the moment the session ends
- Former GitHub CEO Thomas Dohmke launched Entire, raising $60M in seed funding (post-money valuation: $300M)
- The open-source Entire CLI automatically saves AI agent context — prompts, transcripts, tool calls — directly into Git
- The goal: a platform where AI agents and humans co-develop software with full traceability
Background: The Memory Problem No One Is Talking About
Claude Code. GPT-5.3-Codex. Cursor Composer 1.5.
AI coding agents are changing how we build software. Engineers now
run multiple terminal windows, fire prompts at agents, and let them
generate, evaluate, and iterate on code autonomously.
But there's a problem hiding in plain sight.
Every AI agent session is volatile.
The prompt lives in your terminal. The reasoning lives in the context
window. The decision-making process — the why behind each change —
evaporates the moment the session closes.
You retrace steps. Re-enter decisions. Burn tokens re-explaining
context. The background behind a judgment made hours or days ago?
Gone.
Git tells you what changed.
Git has never been able to tell you why.
That's the gap Thomas Dohmke — former GitHub CEO — is building
Entire to close.
What Just Happened: $60M and a New Company
On February 10, 2026, Entire announced a $60 million seed round,
with a post-money valuation of $300 million.
The thesis is straightforward: today's development tools were designed
for human-to-human collaboration, built before the cloud — let alone
before AI agents existed. They weren't designed for a world where
AI agents are primary contributors.
Entire wants to rebuild the software development lifecycle from the
ground up, with AI and humans as co-authors.
Entire CLI: Making AI Reasoning a First-Class Git Citizen
Entire CLI is an open-source tool available on GitHub. Its core
feature is called Checkpoints.
Every time an AI agent generates code and commits it, Entire CLI
automatically captures the full session context and attaches it to
the commit:
| What Gets Recorded | Details |
|---|---|
| Full transcript | The entire agent conversation log |
| Prompts | Every instruction sent to the agent |
| Changed files | Code diffs |
| Token usage | How many tokens were consumed |
| Tool calls | Every tool the agent invoked |
The result: not just a record of what changed, but a traceable
history of the reasoning that produced the change.
The Platform Architecture: 3 Layers
Entire CLI's platform is built on three components:
① Git-Compatible Database
Code, intent, decisions, and reasoning — all managed in a single
version control system.
② Universal Semantic Reasoning Layer
A context graph (a network mapping relationships between contexts)
enables multi-agent collaboration across sessions.
③ AI-Native Software Development Lifecycle
The entire dev lifecycle rebuilt around the assumption that agents
and humans work together — not humans alone.
Currently supports Claude Code and Google Gemini CLI, with
OpenAI Codex, Cursor CLI, OpenCode, and GitHub Copilot CLI support
coming soon.
A Brief History of Version Control
| Era | What We Could Track | What We Couldn't |
|---|---|---|
| Before Git | Manual file backups | Change history |
| Git era | Code changes & diffs | The reasoning behind them |
| Entire CLI era | Code changes + AI reasoning & context | — |
Git revolutionized how we record changes.
Entire CLI wants to revolutionize how we record decisions.
A Hardware Engineer's Perspective
As a semiconductor engineer, traceability isn't optional — it's a
compliance requirement. When something goes wrong on a production
line, you need to explain every design decision, every change, and
every judgment call. "The AI suggested it" is not an acceptable
answer.
This is what makes Entire CLI's approach genuinely interesting to me.
The problem it's solving isn't unique to software development.
Anywhere AI is being used to make or assist with technical decisions,
the same gap exists: the output survives, but the reasoning
doesn't.
In semiconductor and manufacturing environments, I see this playing
out already:
- AI-assisted design verification produces results, but the diagnostic path that led there isn't preserved
- Failure analysis with AI support surfaces insights, but the chain of reasoning disappears after the session
- When auditors or quality reviewers ask "why was this decision made?", the honest answer is often: "we don't know anymore"
Treating AI as a disposable tool — use it, get the output, move on —
works until it doesn't. The moment traceability matters (and in
engineering, it always eventually matters), you'll wish you'd
recorded the why.
Entire CLI is a software-first answer to a problem that exists
across all of engineering.
Takeaways & Next Actions
Bottom line: The next frontier isn't just AI-generated code.
It's AI-generated code with traceable reasoning.
Here's what to do now:
- Try Entire CLI locally — it's open source and available on GitHub right now
- Audit your current AI agent workflow — can you reconstruct why a decision was made a week later? If not, that's the gap
- Add "context preservation" to your team's AI usage guidelines — before it becomes a problem
- If you use Claude Code or Gemini CLI, Entire CLI integrates today — no waiting required
Git changed how we collaborate on what we build.
Entire CLI might change how we understand why we built it that way.
Sources:
- @it — "Git Can't Record 'Why' — Former GitHub CEO Launches Entire CLI" (March 18, 2026)
- Entire Official Blog — "Hello Entire World"
- Entire Blog — Dispatch series (Feb–Mar 2026)
Top comments (0)