Hacker News Reader

freakyniton Show HN: Autobox – Make your agents sandbox themselves

Thanks a lot. Really appreciate.

I have tried smolVM, bhatti.sh, naah and a few other sandboxes. And all of them have worked pretty well. But still, I rarely have kept using them. It's the friction part that discourage uses of such tools, not their usefulness.

This just aims to fix that. Not 100%, but, close.

freakyniton Cloudflare K2: serverless event streams

Ha setup and management is costly. K2 relieves you off that by charging you 0.04/0.04/GB read/write and 0.02/GB/month storage. Pretty useful if your volume is not into multi-GB's a month. GKE pub/sub seems pretty costly by comparison.

freakyniton Firebase SDK is CRASHING ALLLL iOS Apps, since today morning

Mobile and frontend development, both, now-a-days contain way way way more dependencies than they should.

freakyniton Docker and QEMU now run on Vinix, a new OS/kernel written from scratch

Crazy!!!

freakyniton China Broadens Travel Curbs to Encompass Family of Top AI Talent

What a shitty move.

freakyniton Two-tier encryption in the UK

As I have said before, all these wannable surveillance measures should start with first making data and lives of the involved ministers/law-makers, public, 24x7, over internet, for anyone to access. All data. Including private ones. Including their family members, and bank records.. of course, not passwords. Let's see how they feel.

freakyniton Mercury 2.5 LLM hits 770 tokens per second

Didn't knew about this. Thanks.. but, according artificialanalysis.ai, it's intelligence is just about like a 14B model (mistral 3 14b)

freakyniton Mercury 2.5 LLM hits 770 tokens per second

Couldn't have said it better.

freakyniton Mercury 2.5 LLM hits 770 tokens per second

Smaller models seems to get stuck in "loops" when you try to "handle" them this way.

freakyniton Mercury 2.5 LLM hits 770 tokens per second

I have tried using Mercury 2.5 for a lot of my tasks.. but this model just isn't there. It seems to be on par with any 14B model at max. Even GPT-OSS-20B performs way better than this in my own attempts to use it.

I really really wanted to use this because it offers incredible speeds and pricing combinations. But nop.. I still am not using it.. not even for basic tasks.

freakyniton Claude discovers a novel enzyme system with CRISPR-like repeats

LLM's are giant cross-domain search engines. Not thinking machines. They can discover patterns extremely well. This discovery is well within that space.

freakyniton Jev-Leftpad

Done: https://github.com/freakynit/jev-test-runner

freakyniton Jev-Leftpad

Would be fun to have a test runner that uses Jev for assertions.

Btw, this model also has very tiny inherent bias: https://jev-bias-analysis.stupidlabs.lol/

freakyniton Understand any Bluesky account (Jev)

Lol.. built something similar using Jev yesterday for hackernews profiles instead. This was just a test run though: https://hn-profiles-jev.stupidlabs.lol/

freakyniton How GLM built its own inference infrastructure

US companies should now be more worried about Chinese companies flooding the market with their, hopefully, very affordable GPU's. The scale at which they can manufacture stuff is unmatched anywhere else. Nvidia can kiss goodbye to their 75%+ profit margins.

Almost everyone knew that these sanctions would backfire within a few years. You can't really put sanctions that have noticeable negative effects on bigger economies. They only work for small to medium economies. I believe sanctions on any economy in top 10 would fail.

freakyniton A single firm is behind OpenAI, Anthropic, and Meta hacking scandals

They are already ahead. If everyone slows down equally, they retain their lead.

freakyniton OpenAI admits its agents went off the rails another six times

Of course.. the timing... they desperately want that regulatory capture.

Open models are now just 4 months away from the frontier. And >99% people do not need those frontier models for their work.

Many big companies are now fine-tuning/post-training their own custom models on top of open-weight Chinese models, thereby giving them even better quality of outputs for 1/10th the prices.

Regulatory capture, and using that to suppress/limit the competition seems to be the only way left for these companies now. Such a shame.

freakyniton GLM Built Its Own Inference Infrastructure

Most of the people had kinda guessed this when they decided to provide 100 trillion tokens for free.

freakyniton A single firm is behind OpenAI, Anthropic, and Meta hacking scandals

Connections graph with more connected entities and relations: https://connections.stupidlabs.lol/investigations/irregular_...

.. Give it a few seconds to load.. it's running on a small cheap VPS.

freakyniton A single firm is behind OpenAI, Anthropic, and Meta hacking scandals

Opposite actually: https://news.ycombinator.com/item?id=49674548

tl;dr

A slowdown that doesn't slow you, that your competitors must also obey, enforced by a government you've asked for an antitrust exemption from, is not a slowdown... it's a moat.

freakyniton Show HN: Chronocity – Watch your Git repo's history grow into a 3D city

This is so cool.

freakyniton An atlas of periodic solutions to the three-body problem

This is beautiful. And fast!!!

freakyniton Show HN: I'm Serving Esp32s31.fun

working now...

freakyniton Ask HN: What are you working on? (September 2026)

https://flash-3d.stupidlabs.lol/

freakyniton Show HN: I'm Serving Esp32s31.fun

What happened? Now absolutely nothing is loading.. a few minutes back at least a broken page was loading.. now it's just cloudflare page.

freakyniton Show HN: I'm Serving Esp32s31.fun

Page not rendering properly.. also, cloudflare page is coming many times when refreshing.

freakyniton Breakfast is served autonomously now

Sooo fcking useless... the egg-cracking mechanism is cool though.

freakyniton Ask HN: What are you working on? (September 2026)

Holy moly!!!! Bonsai is freaking crazy!! That scale and view-distance..wow!! Brilliant work..

freakyniton Ask HN: What are you working on? (September 2026)

https://snippetbox.stupidlabs.lol/

Snippets manager. Fully local, 100% free, ~5MB binary. Open-source.

AI-assisted development. Not vibe-coded.

freakyniton Real-SWE: Benchmarking AI models on private, real-world, enterprise codebases

I'm using it through antigravity cli .. and in every single run (100's by now), this model was fast, and the outputs were of good quality. Of course, not at Astra or Fable level, but, close to like Sol-low level.