A 500k line codebase for an agent CLI proves one thing: making a probabilistic LLM behave deterministically is a massive state-management nightmare. Right now, they're great for prompting simple sites/platforms but they break at large enterprise repos.
If you don't have a rigid, external state machine governing the workflow, you have to brute-force reliability. That codebase bloat is likely 90% defensive programming; frustration regexes, context sanitizers, tool-retry loops, and state rollbacks just to stop the agent from drifting or silently breaking things.
The visual map is great, but from an architectural perspective, we're still herding cats with massive code volume instead of actually governing the agents at the system level.
autocracy101
Author here. I built this in a few hours after the Claude Code leak.
I've been working on my own coding agent setup for a while. I mostly use pi [0] because it's minimal and easy to extend. When the leak happened, I wanted to study how Anthropic structured things: the tool system, how the agent loop flows, A 500K line codebase is a lot to navigate, so I mapped it visually to give myself a quick reference I could come back to while adapting ideas into my own harness and workflow.
I'm actively updating the site based on feedback from this thread. If anything looks off, or you find something I missed, lmk.
Isn't it a simple REPL with some tools and integrations, written in a very high level language? How the hell is it so big? Is it because it's vibecoded and LLMs strive for bloat, or is it meaningful complexity?
If it was 2020, it would be hard to imagine that after some hours/days you getting a visual representation of the leak with such detailed stats lol
show comments
brauhaus
Even today, I'm still astounded that there are people capable of building a gorgeous and interesting site like this in less than 2 days...
show comments
jedisct1
I'm developing an agent focused on A2A, support for small models, and privacy (https://swival.dev).
I looked at the leaked code expecting some "secret sauce", but honestly didn't found anything interesting.
I don't get the hype around Claude Code. There's nothing new or unique. The real strength are the models.
stingraycharles
I guess they really do eat their own dogfood and vibe code their way through it without care for technical debt? In a way, it’s a good challenge, but it’s fairly painful to watch the current state of the project (which is about a year old now, so it should be in prime shape).
Thanks to Claude Code, we got such a beautifully polished and dazzling website that gives a complete introduction to itself the very moment the leak happened :)
brandensilva
ccleaks.com seems to be "temporarily paused" from Vercel.
Here is another one that goes in depth as well: www.markdown.engineering for anyone going deep on learning.
restlessforge
Okay those "hidden features" are amazing, especially the cross-session referencing. I hope we can look forward to that in the future
Also I definitely want a Claude Code spirit animal
There's this weird thing about AI generated content where it has the perfect presentation but conveys very little.
For example the whole animation on this website, what does it say beyond that you make a request to backend and get a response that may have some tool call?
show comments
blueTiger33
its April fools joke. this has really gone wide
sourcecodeplz
Nice presentation. The reality is there is nothing really special about the claude code harness?
jen729w
Is it just me or do I not find the Claude Code application that fascinating?
I use it all day and love it. Don't get me wrong. But it's a terminal-based app that talks to an LLM and calls local functions. Ooookay…
show comments
sibtain1997
Kairos and auto-dream are more interesting than anything in the agent loop section. Memory consolidation between sessions is the actual unsolved problem. The rest is just plumbing tbh
show comments
AJRF
This is AI slop.
First command I looked at:
/stickers:
Displays earned achievement stickers for milestones like first commit, 100 tool calls, or marathon sessions. Stickers are stored in the user profile and rendered as ASCII art in the terminal.
That is not what it does at all - it takes you to a stickermule website.
What is the motivation for someone to put out junk like this?
show comments
vivzkestrel
would be nice if the transformers code for one of these frontier LLM models got leaked, HN will have a field day with a reveal like that
show comments
kinnth
this claude code leak is such a fuck up...
The fact that now every agent designer knows what was already built is a huge shot of steroids to their codebase!
cubefox
I think is unethical, and "everyone else is also doing it" is not a valid excuse.
lanbin
However, excellent development practices involve modularizing code based on functional domains or responsibilities.
The utils directory should only contain truly generic, business-agnostic utilities (such as date retrieval, simple string manipulation, etc.).
We can see that the code produced by Vibe is not what a professional engineer would write. This may be due to the engineers using the Vibe tool.
show comments
rhofield
Really nice visualisation of this, makes understanding the flow at a high levle pretty clear. Also the tool system and command catalog, particularly the gated ones are super interesting.
Hannah_Adam
Is that safe to use?
nitnelave
Ah, good well-architected code, finally... With most of the code in utils/other :D
fersarr
why do people care so much? it's just an agentic loop
519K lines of code for something that is using the baseline *nix tools for pretty much everything important, how do they even manage to bloat it this much? I mean I know how technically, but it's still depressing.
Can't they ask CC to make it good, instead of asking it to make it bigger?
dominotw
what is so fascinating about claude code. we have codex that is open source already. is there something special to learn from claude code?
m132
I mean, I get it: vibe-coded software deserves vibe-coded coverage. But I would at least appreciate it if the main part of it, the animation, went at a speed that at least makes it possible to follow along and didn't glitch out with elements randomly disappearing in Firefox...
How is this on the front page?
show comments
ramon156
I expect dozens more "research articles" that
- find nothing
- still manage to fill entire lages
- somehow have a similar structure
- are boring as fuck
At least this one is 3/4, the previous one had BINGO.
fsniper
Source leak or free code review? I can say that there is no bad publicity.
Thanks, I'll use this for teaching next week (on what not to do). BashTool.ts :D But, in general, I guess it just shows yet again that the emperor has no clothes.
A 500k line codebase for an agent CLI proves one thing: making a probabilistic LLM behave deterministically is a massive state-management nightmare. Right now, they're great for prompting simple sites/platforms but they break at large enterprise repos.
If you don't have a rigid, external state machine governing the workflow, you have to brute-force reliability. That codebase bloat is likely 90% defensive programming; frustration regexes, context sanitizers, tool-retry loops, and state rollbacks just to stop the agent from drifting or silently breaking things.
The visual map is great, but from an architectural perspective, we're still herding cats with massive code volume instead of actually governing the agents at the system level.
Author here. I built this in a few hours after the Claude Code leak.
I've been working on my own coding agent setup for a while. I mostly use pi [0] because it's minimal and easy to extend. When the leak happened, I wanted to study how Anthropic structured things: the tool system, how the agent loop flows, A 500K line codebase is a lot to navigate, so I mapped it visually to give myself a quick reference I could come back to while adapting ideas into my own harness and workflow.
I'm actively updating the site based on feedback from this thread. If anything looks off, or you find something I missed, lmk.
[0] https://pi.dev/
> 500k lines of code
Isn't it a simple REPL with some tools and integrations, written in a very high level language? How the hell is it so big? Is it because it's vibecoded and LLMs strive for bloat, or is it meaningful complexity?
Here's a codeberg repo with the leaked source: https://codeberg.org/wklm/claude-code
If it was 2020, it would be hard to imagine that after some hours/days you getting a visual representation of the leak with such detailed stats lol
Even today, I'm still astounded that there are people capable of building a gorgeous and interesting site like this in less than 2 days...
I'm developing an agent focused on A2A, support for small models, and privacy (https://swival.dev).
I looked at the leaked code expecting some "secret sauce", but honestly didn't found anything interesting.
I don't get the hype around Claude Code. There's nothing new or unique. The real strength are the models.
I guess they really do eat their own dogfood and vibe code their way through it without care for technical debt? In a way, it’s a good challenge, but it’s fairly painful to watch the current state of the project (which is about a year old now, so it should be in prime shape).
Feel free to add this to Awesome Claude code. https://github.com/rosaboyle/awesome-cc-oss
Thanks to Claude Code, we got such a beautifully polished and dazzling website that gives a complete introduction to itself the very moment the leak happened :)
ccleaks.com seems to be "temporarily paused" from Vercel.
Here is another one that goes in depth as well: www.markdown.engineering for anyone going deep on learning.
Okay those "hidden features" are amazing, especially the cross-session referencing. I hope we can look forward to that in the future
Also I definitely want a Claude Code spirit animal
> also related: https://www.ccleaks.com
This deployment is temporarily paused
There's this weird thing about AI generated content where it has the perfect presentation but conveys very little.
For example the whole animation on this website, what does it say beyond that you make a request to backend and get a response that may have some tool call?
its April fools joke. this has really gone wide
Nice presentation. The reality is there is nothing really special about the claude code harness?
Is it just me or do I not find the Claude Code application that fascinating?
I use it all day and love it. Don't get me wrong. But it's a terminal-based app that talks to an LLM and calls local functions. Ooookay…
Kairos and auto-dream are more interesting than anything in the agent loop section. Memory consolidation between sessions is the actual unsolved problem. The rest is just plumbing tbh
This is AI slop.
First command I looked at:
That is not what it does at all - it takes you to a stickermule website.What is the motivation for someone to put out junk like this?
would be nice if the transformers code for one of these frontier LLM models got leaked, HN will have a field day with a reveal like that
this claude code leak is such a fuck up...
The fact that now every agent designer knows what was already built is a huge shot of steroids to their codebase!
I think is unethical, and "everyone else is also doing it" is not a valid excuse.
However, excellent development practices involve modularizing code based on functional domains or responsibilities.
The utils directory should only contain truly generic, business-agnostic utilities (such as date retrieval, simple string manipulation, etc.).
We can see that the code produced by Vibe is not what a professional engineer would write. This may be due to the engineers using the Vibe tool.
Really nice visualisation of this, makes understanding the flow at a high levle pretty clear. Also the tool system and command catalog, particularly the gated ones are super interesting.
Is that safe to use?
Ah, good well-architected code, finally... With most of the code in utils/other :D
why do people care so much? it's just an agentic loop
I prefer this mapping from Nikita @ CosmoGraph: https://run.cosmograph.app/public/dfb673fc-bdb9-4713-a6d6-20...
Nice site. I might suggest moving SendMessage to the Hidden Features as they don't appear to have implemented a ReadMessage or ListMessages tools.
So it does use ripgrep and not unix grep. [0] I knew it from some other commenters here on HN, but it's nice to see it in the source as well.
0 - https://github.com/zackautocracy/claude-code/blob/main/src/u...
I hope /Buddy is ported across to OpenCode.
nice example: Find all TODO spin the AI machine
i do shift ctrl F
519K lines of code for something that is using the baseline *nix tools for pretty much everything important, how do they even manage to bloat it this much? I mean I know how technically, but it's still depressing. Can't they ask CC to make it good, instead of asking it to make it bigger?
what is so fascinating about claude code. we have codex that is open source already. is there something special to learn from claude code?
I mean, I get it: vibe-coded software deserves vibe-coded coverage. But I would at least appreciate it if the main part of it, the animation, went at a speed that at least makes it possible to follow along and didn't glitch out with elements randomly disappearing in Firefox...
How is this on the front page?
I expect dozens more "research articles" that
- find nothing - still manage to fill entire lages - somehow have a similar structure - are boring as fuck
At least this one is 3/4, the previous one had BINGO.
Source leak or free code review? I can say that there is no bad publicity.
How the hell is it 500k lines?
I got a goose
War flashbacks to genshin
Ccleaks is down?
cool Archaeologization Collection Output
Hey, nice job! Next time tell calude to add some explosions, car crashes and stuntment into the design! Who cares about content anyway ... https://speculumx.at/blogpost/getting-sick-of-ai-slop
Thanks, I'll use this for teaching next week (on what not to do). BashTool.ts :D But, in general, I guess it just shows yet again that the emperor has no clothes.