Two ex-GitHub folks, including the former CEO, are rethinking version control for AI agents. You can see what the agent did. The conversation that got you there? Gone.
Their fix: make the conversation part of the record, not just the code.
One of my favorite things about @EntireHQ is that code no longer appears out of nowhere.
You can trace a commit back through the conversations, prompts, iterations, and decisions that produced it.
I've learned a ton from seeing how teammates use Codex and Claude Code in their
One of my favorite things about @EntireHQ is that code no longer appears out of nowhere.
You can trace a commit back through the conversations, prompts, iterations, and decisions that produced it.
I've learned a ton from seeing how teammates use Codex and Claude Code in their
@EntireHQ seems like a great candidate to solve this exact problem. Tried it yesterday and it is kinda intuitive. Will have to experiment for a couple of weeks to see if it is the right solution.
One thing I’d love coding agents to have in their DNA is to preserve a system of record:
The context behind why something was built a certain way.
Sometimes I’ll have a long back-and-forth with an agent:
- debate tradeoffs
- reject a few tempting but bad options
- push back on
Beep, boop. Entire Dispatch 0x0012 is here.
From Tokyo summit week: new ways to trace code changes back to the checkpoints and sessions behind them.
Highlights:
🧪 Entire CLI 0.7.6 adds experimental `entire blame` and `entire why` commands in labs
↩️ `entire checkpoint rewind`
From returning Git to its original promise—a truly distributed system where repositories are replicated across many hosts—to extending Git to version sessions, prompts, chat logs, and more as the new provenance layer of software development, our CEO, @ashtom discussed how we
Agents shouldn't have amnesia.
I recently hit a wall with an expired auth token in an undocumented, internal workflow. I knew my agent and I had spent an hour agonizingly debugging this exact setup two months ago.
But because that terminal session vanished into the ether, I
me and @socratesom were selected from 100+ applicants as one of 25 teams admitted to @FlorentVC's €25k #hackathon at panathēnea in athens and we made it to the top 3 finalist stage 🏆
the photo with @ashtom was worth more than the apple prizes 😎
thanks @EntireHQ beep boop
Beep, boop. Entire Dispatch 0x0011 is here.
Highlights:
🔐 Entire CLI auth is clearer, with streamlined contexts and keyring timeouts instead of frozen logins
🧠 Claude and Pi sessions now capture skill metadata for richer transcripts
🥧 Pi setup is more consistent across
So excited to be cohosting AI Engineer Trivia Night w/ @EntireHQ, @sanity_io, @temporalio, && @OpenAI as a side event for @aiDotEngineer!
📆Thursday, July 2
📍Standard Deviant Brewing Pier 70
Hope to see you there!
I’m seeing a lot of negative reactions to this. I think some of it is because the idea feels uncomfortable, but isn’t this close to what people were already doing with Ralph Loops?
- Give the agent a detailed PRD
- Write a loop
- Let it self-check progress and completion
-
Same hosts. New night. Bigger room. @jreyesdev and @MelGoesTech are back on July 2nd in SF with AI Engineer Trivia Night — a new take on the format with rounds like "Are you smarter than AI?" and "Vibe-coding gone wrong," plus the crowd favorites that kept selling out last year.