Hacker News Reader

mohsen1on Vibe coding and agentic engineering are getting closer than I'd like

Multiple computers and each multiple Claude Code or Codex sessions. It had lots of ups and downs. Now I have a good enough test harness that makes it easier to iterate faster

mohsen1on Vibe coding and agentic engineering are getting closer than I'd like

I am experimenting with writing en entire TypeScript compiler[1] with AI assistant. I've spent 4 months on it already. It might not be successful at the end of the day but my thinking is that if LLMs are going to write a lot of the code I better learn how this can and can not work. I've learned a lot from this project already. I think we're still in charge of design and big ideas even if all of the code is written by AI

[1] https://github.com/mohsen1/tsz

mohsen1on SubQ: a sub-quadratic LLM with 12M-token context

- magic.dev claimed 200M context window and it's been two years since and no real product yet.

- They are admitting that this is built on top of a Chinese model[1]

- They committed a huge chart crime with the Y axis of a chart comparing to Opus on their website that I can't find anymore (Too embarrassing to keep?). The delta between their score (81%) vs. Opus (87%) on SWE bench was hugely minimized

- They named the company subquadratic but in parts they said O(1) linear scaling. At O(1) you could do much more than 12M tokens context window. At O(log n) even.

I hope this is real but I doubt...

mohsen1on Zig → Rust porting guide

Rust is a good choice to let LLMs run without a ton of supervision. In my experience you need to monitor the progress heavily and take ownership of the design of the thing you're building or porting. Test harness is a must. Each iteration should run the test and ensure it doesn't break things in other places.

I am in the middle of porting TypeScript to Rust and learned a ton doing this. You can check out the work in progress here https://github.com/mohsen1/tsz/

Happy to share my learnings on this

mohsen1on Zig → Rust porting guide

I am also porting TypeScript to Rust. With a different design I managed to make it faster than tsgo port. I've made a lot of progress in the last 4 months but needs more work. Contributions are welcome!

https://tsz.dev

mohsen1on Incident with Issues and Webhooks – Resolved

The same company operates the Xbox network. More daily active users and more events per second

mohsen1on DeepClaude – Claude Code agent loop with DeepSeek V4 Pro

This has been my experience working on tsz.dev. Only Opus 4.7 and GPT 5.5 can really be productive for the remaining test cases.

mohsen1on DeepSeek V4–almost on the frontier, a fraction of the price

In my experience V4 is pretty good but for very hard problems it burns way too many tokens that it ends up being not so cheap anymore. I'm working on a compiler and the tasks are very involved. Tests won't pass unless it gets it absolutely right. 5.5 can achieve more in less time compared to V4 for me.

mohsen1on Uber wants to turn its drivers into a sensor grid for self-driving companies

I was working at Lyft 8 years ago and suggested this to the head of AV program then. They didn't listen.

mohsen1on An update on GitHub availability

Try google3

mohsen1on An update on GitHub availability

Another interesting bit: they are hitting performance issues due to the rise of monorepos. GitHub and frankly Git were not designed for monorepos

mohsen1on GitHub Copilot is moving to usage-based billing

This was my solution to very very though compiler tests that would take sometime up to 4 hours to figure out. Some of the time would be spent on running the tests, but still... I was burning so much tokens. I have free Copilot for my open source work so I wasn't even paying the $20.

this is the project that I am working on https://github.com/mohsen1/tsz

mohsen1on DeepSeek v4

"2 minutes ago" https://huggingface.co/unsloth/DeepSeek-V4-Pro

mohsen1on Vision Banana: Image Generators Are Generalist Vision Learners

https://clockbench.ai

Can it read the clock?

mohsen1on SpaceX says it has agreement to acquire Cursor for $60B

I know Cursor is getting economically not so viable compared to OpenAI and Anthropic offerings but with a deal like this they could also offer $200/mo plans that are attractive. Obviously _if_ their models are good. We have to see!

mohsen1on Ask HN: How to solve the cold start problem for a two-sided marketplace?

Uber also paid riders to ride. I was working for Garret Camp at SumbleUpon and we got free Uber Black back then. The number of available drivers even in SF was so low that it was not really useful, even free!

mohsen1on Instant 1.0, a backend for AI-coded apps

Congratulations on this launch. Stepan and his cofounder have been working on this for years and great to see it being launched finally. I should be building something fun with this.

Sending you guys lots of love and best of luck!

mohsen1on We've raised $17M to build what comes after Git

I watched video to see where my prompts etc are stored in a way that makes sense. But no, this is just a nicer git. We need a solution to all these 10k loc PRs.

mohsen1on Wit, unker, Git: The lost medieval pronouns of English intimacy

If you're interested in history of English, I'd highly recommend the History of English podcast. https://historyofenglishpodcast.com

mohsen1on Expanding Swift's IDE Support

I'm building iOS apps without opening Xcode. All in VSCode and CLI build tools. This is excellent!

mohsen1on You can't cancel a JavaScript promise (except sometimes you can)

Back in 2012 I was working on a Windows 8 app. Promises were really only useful on the Windows ecosystem since browser support was close to non existent. I googled "how to cancel a promise" and the first results were Christian blogs about how you can't cancel a promise to god etc. Things haven't changes so much since, still impossible to cancel a promise (I know AbortSignal exists!)

mohsen1on GitHub has DMCA'd nearly all forks of the official Claude-code repo

I haven't time to do it but can someone try to unminify the newer version based on the minified new version + the source of previous version? There's gotta be a way to do this

mohsen1on Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw

They have terms to not allow `claude -p` used like that. However, people can hide this with the leaked source code. What a funny cat-and-mouse game!

mohsen1on Google releases Gemma 4 open models

On LM Studio I'm only seeing models/google/gemma-4-26b-a4b

Where can I download the full model? I have 128GB Mac Studio

mohsen1on Claude Code's source code has been leaked via a map file in their NPM registry

Maybe hard to believe but not everyone is speaking English to Claude

mohsen1on Claude Code's source code has been leaked via a map file in their NPM registry

it's the `runHeadlessStreaming` function btw

mohsen1on Claude Code's source code has been leaked via a map file in their NPM registry

here's another gem. src/ink/termio/osc.ts:192–210

  void execFileNoThrow('wl-copy', [], opts).then(r => {
    if (r.code === 0) { linuxCopy = 'wl-copy'; return }
    void execFileNoThrow('xclip', ...).then(r2 => {
      if (r2.code === 0) { linuxCopy = 'xclip'; return }
      void execFileNoThrow('xsel', ...).then(r3 => {
        linuxCopy = r3.code === 0 ? 'xsel' : null
      })
    })
  })

are we doing async or not?

mohsen1on Claude Code's source code has been leaked via a map file in their NPM registry

src/cli/print.ts

This is the single worst function in the codebase by every metric:

  - 3,167 lines long (the file itself is 5,594 lines)
  - 12 levels of nesting at its deepest
  - ~486 branch points of cyclomatic complexity
  - 12 parameters + an options object with 16 sub-properties
  - Defines 21 inner functions and closures
  - Handles: agent run loop, SIGINT, rate-limits, AWS auth, MCP lifecycle, plugin install/refresh, worktree bridging, team-lead polling (while(true) inside), control message dispatch (dozens of types), model switching, turn interruption
  recovery, and more

This should be at minimum 8–10 separate modules.

mohsen1on Day 1 of ARC-AGI-3

Uses public dataset to evaluate which is not meant for evaluation. Writes super specific prompt[1] and claims eye catching results.

This is the state of "AI" these days I guess...

[1] https://github.com/symbolica-ai/ARC-AGI-3-Agents/blob/symbol...

mohsen1on Tell HN: Litellm 1.82.7 and 1.82.8 on PyPI are compromised

If it was not spinning so many Python processes and not overwhelming the system with those (friends found out this is consuming too much CPU from the fan noise!) it would have been much more successful. So similar to xz attack

it does a lot of CPU intensive work

    spawn background python
    decode embedded stage
    run inner collector
    if data collected:
        write attacker public key
        generate random AES key
        encrypt stolen data with AES
        encrypt AES key with attacker RSA pubkey
        tar both encrypted files
        POST archive to remote host