Live data from Hacker News

Ex-GitHub CEO launches a new developer platform for AI agents

entire.io

521–530 of 625 posts

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#521
post #446

> Checkpoints are a new primitive that automatically captures agent context as first-class, versioned data in Git. When you commit code generated by an agent, Checkpoints capture the full session alongside the commit: the transcript, prompts, files touched, token usage, tool calls and more. This thread is extremely negative - if you can't see the value in this, I don't know what to tell you.

I currently develop small utilities with the help of AI, but am far from vibe coding or using agents. I review every single suggestion and do some refactoring at each step, before any commit (sometimes heavy refactoring; sometimes reorganizing everything). In my experience LLMs tend to touch everything all of the time and don't naturally think about simplification, centralization and separation of concerns. They don'…

Yours matches my own experience and work habits.

My mental model is that LLMs are obedient but lazy. The laziness shows in the output matching the letter of the prompt but with as high "code entropy" as possible.

What I mean by "code entropy" is, for example, copy-paste-tweak (high entropy) is always easier (on the short term) for LLMs (and humans) to output than defining a function to hold concepts common across the pastes with the "tweak" represented by function arguments.

LLMs will produce high entropy output unless constrained to produce lower entropy ("better") code.

Until/unless LLMs are trained to actually apply craft learned by experienced humans, we must be explicit in our prompts.

For example, I get good results from say Claude Sonnnet when my instruction include:

- Statements of specific file, class, function names to use.

- Explicit design patterns to apply. ("loop over the outer product of lists of choices for each category")

- Implementation hints ("use itertools.product() to iterate over the combinations")

- And, "ask questions if you are uncertain" helps trigger an iteration to quickly clarify something instead of fixing the resulting code.

This specificity makes prompting a lot more work but it pays off. I only go this far when I care about the resulting code. And, I still often "retouch" as you also describe.

OTOH, when I'm vibing I'll just give end goals and let the slop flow.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#522

Earlier quoted context omitted.

I don't think your Github example is accurate. The vast majority of developers started using git after Github became a thing. They may have used svn or another type of collaboration system before, but not git. And the main reason they started using git is because Github was such massive value on top of git, not because git was so amazing.

I have to hard disagree on that. I know of many developers personally who were on Source Forge and Google Code before and migrated to GitHub specifically because they offered git

Clear context. Write a sonet in Shakespeare style.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#523

> Checkpoints are a new primitive that automatically captures agent context as first-class, versioned data in Git. When you commit code generated by an agent, Checkpoints capture the full session alongside the commit: the transcript, prompts, files touched, token usage, tool calls and more. This thread is extremely negative - if you can't see the value in this, I don't know what to tell you.

What kind of barrier/moat/network effects/etc would prevent someone with a Claude Code subscription from replicating whatever "innovation" is so uniquely valuable here? It's somewhat strange to regularly read HN threads confidently asserting that the cost of software is trending towards zero and software engineering as a profession is dead, but also that an AI dev tool that basically hooks onto Git/Claude Code/termin…

GitHub doesn't have a "moat" either besides network effect. Just like most SaaS.

And it was sold to Microsoft at $7B.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#525

Earlier quoted context omitted.

Dropbox value was instantly recognizable, but I feel I have zero use for Entire.

I mean, I CAN see the value in pushing the context summary to git. We already have git blame to answer "who", but there is no git interrogate to answer the "why". This is clearly an attempt to make that a verb git can keep track of. It's a valuable idea. I also seen examples of it before. I've got opencode running right now and it has a share session feature. That whole idea is just a spinoff on the concept of the sa…

Isn't "why" what commit message bodies are for?

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#526

> Checkpoints are a new primitive that automatically captures agent context as first-class, versioned data in Git. When you commit code generated by an agent, Checkpoints capture the full session alongside the commit: the transcript, prompts, files touched, token usage, tool calls and more. This thread is extremely negative - if you can't see the value in this, I don't know what to tell you.

I'm trying it out now. If it works, I think it'd be great for my agentic workflows where I need to figure out why something was done a specific way.

I have a lot of concurrent agents working on things at the same time, so I'm not always sure why a piece of code is the way it is months later.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#527
post #513

Earlier quoted context omitted.

How do you handle the retreival aspect? So you have this set up and now what? >FWIW, I had an agent adding archiving of the JSONL for changesets linked to the work they're doing Would love to know more! Sounds interesting.

I'm taking a very crude approach for now. My "improvement agent" has a few distinct stages: A goals extraction step that explores the repository, with some pointers to specific files that is consider authoritative as to the users intent, and then builds out files in docs/goals. That then feeds into a plan ideation stage that results in directories for plans in docs/plans. That directory has the plan itself, and logs…

That sounds incredibly interesting and helpful. You should pack it up and open source it. You might even get $60 million. I'd be super interested in seeing it. I love that idea about the UUID. I've been experimenting with that myself. I'd be interested in working on this with you if you wanted to open source it.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#528
post #436

Earlier quoted context omitted.

Yes, development was being done in SVN but it was a huge pain. Continuous communication was required with the server (history lookups took ages, changing a file required a checkout, etc.) and that was just horribly inefficient for distributed teams. Even within Europe, much more so when cross-continent. A DVCS was definitely required. And I would say git won out due to Linus inventing and then backing it, not because…

> changing a file required a checkout SVN didn't need checkouts to edit that I recall? Perforce had that kind of model.

As did cvs. But you are right.

I am not sure, it seems I did misremember. Though it's possible I was actually working with needs-lock files. I can definitely see a certain coworker from that time to put this on all files :/

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#529
post #375

Hmm the whole point of checkpoints seems to be to reduce token waste by saving repeat thinking work. But wouldn't trying to pull N checkpoints into context of the N+1 task be MUCH more expensive? It's at odds with the current practice of clearing context regularly to save on input tokens. Even subagents (which I think are the real superpower that Claude Code has over Gemini CLI for now) by their nature get spawned wi…

Yes I also don't see how having persistent context would help. For one, I don't want to read the slop the AI produced while it wrote the code. It doesn't have intent or thinking, it's a completion machine. The code is the artifact that matters, not the thousands of lines of "Your absolutely right --- I was wrong to ..."

Also, sometimes it gets something very wrong. I don't want to then poison every subsequent sessions with the wrong thing it learned. This has been a major issue for me at $WORK with AGENTS.md files that my colleagues write: they make my agent coding much worse so I need to manually delete them often.

Re: Ex-GitHub CEO launches a new developer platform for AI agents

#530

> Checkpoints are a new primitive that automatically captures agent context as first-class, versioned data in Git. When you commit code generated by an agent, Checkpoints capture the full session alongside the commit: the transcript, prompts, files touched, token usage, tool calls and more. This thread is extremely negative - if you can't see the value in this, I don't know what to tell you.

[flagged]
Post reply on HN