frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Show HN: Fedijar – The Search Mastodon servers meant to have

https://fedijar.com
1•saneefansari•1m ago•0 comments

Show HN: Aitho – slides that advance when you say the words, not when you click

https://present.aitho.app/try
1•aumar2•4m ago•0 comments

Are you clinging to the TUI to feel like a developer?

1•TomEleff•6m ago•0 comments

Ask HN: Could AI's RH proof have used human work found in its training data?

1•nomilk•6m ago•0 comments

Vector and Similarity Search, by Hand

https://www.huy.rocks/everyday/11-21-2026-math-linear-algebra-notes-vector-and-similarity-search-...
1•huydotnet•6m ago•0 comments

Nospeaker: Make your MacBook's trackpad talk using its haptic motor

https://github.com/os6/nospeaker
1•vikrum•7m ago•0 comments

Tor Community Fellowship

https://forum.torproject.org/t/call-for-applications-tor-community-fellowship/22490
1•farhan291•7m ago•0 comments

When services fail, we see what they are made of

https://good.services/writing/when-services-fail-we-see-what-they-are-made-of
1•adrianhoward•7m ago•0 comments

Show HN: Exploitwatch – Alerts for vulnerable Docker containers (e.g. PHP2Shell)

https://github.com/marioleal/exploitwatch
1•ympb121•8m ago•1 comments

Inside an AI prophet's secretive visions for humanity's future

https://www.washingtonpost.com/technology/2026/11/22/inside-an-ai-prophets-secretive-plans-humani...
1•pretext•9m ago•0 comments

Ask HN: Why do these LLM models seem to improve like clockwork over time?

2•iknownthing•13m ago•2 comments

Liseur: Android ePub Reader

https://github.com/chmouel/liseur
1•swapnilmj•13m ago•1 comments

Enrico Fermi Doesn't Blow Up Chicago

https://colab.research.google.com/drive/1LNey4n-V8QciZLhcUSR7IUOKeId_58KP?usp=sharing
1•Eridanus2•13m ago•1 comments

A fast image processing library with low memory needs

https://github.com/libvips/libvips
1•SweetSoftPillow•13m ago•0 comments

Cybermacs 3D Window Manager

https://kibotronics.net/projects/cybermacs/
1•Kiboneu•15m ago•0 comments

Show HN: GTA6 Wiki – Reference Site for Grand Theft Auto VI

https://gtaviwiki.org/
1•ludydev•15m ago•0 comments

Apple says it will shut down its Apple Support Community forums after 28 years

https://mashable.com/tech/apple-shutting-down-support-community-forums
2•taubek•16m ago•0 comments

Perplexity and Intelligence

https://openai.com/index/perplexity-and-intelligence/
3•acossta•20m ago•1 comments

Software Rendering the Mandelbrot Fractal in Rust

https://antr.dev/blog/software-rendering-the-mandelbrot-fractal-in-rust/
1•lr0•22m ago•0 comments

Show HN: TEB RUN – a pseudo-3D arcade racer in TypeScript and Canvas, no engine

https://tebrun.io/
2•tebrun•23m ago•0 comments

Saving Lives Before the Singularity

https://clattubato.substack.com/p/saving-lives-before-the-singularity
1•paulpauper•23m ago•0 comments

Show HN: Type prompts directly in terminal shell to kickoff Claude and codex

https://github.com/mquan/llmao
1•quan•23m ago•0 comments

Banning RSI Is Mere Stalling

https://emethirsch.neocities.org/banning-rsi-is-mere-stalling
2•paulpauper•24m ago•0 comments

Boring: An Idiot in Berkeley

https://henryaj.substack.com/p/boring
1•paulpauper•24m ago•0 comments

How to make Three.js support real-time path tracing?

1•winstongu20•25m ago•0 comments

The Riemann Hypothesis (Post by Marty Ross of Mathologer)

https://mathematicalcrap.com/2026/11/22/the-riemann-hypothesis/
1•joebig•25m ago•0 comments

Connectome for Drones

https://huggingface.co/SingularityEdge/FlyDrone-SNN
1•janmirank•25m ago•0 comments

Proving Rust Web Application Correctness with Lean 4

https://www.reddit.com/r/rust/comments/1wz7q4o/proving_rust_web_application_correctness_with/
1•syumei•26m ago•0 comments

Amazon nearing $7.5B deal to buy AI startup Decart: FT

https://seekingalpha.com/news/4675396-amazon-nears-75b-deal-buy-ai-start-up-decart
1•rainhacker•27m ago•0 comments

The Quality Is Still on Me

https://edisoncode.com/articles/the-quality-is-still-on-me/
1•jeddawson•27m ago•0 comments
Open in hackernews

Ask HN: Why do these LLM models seem to improve like clockwork over time?

2•iknownthing•13m ago
It seems like these AI labs are able to guarantee more advanced models over time. How are they able to do this? Are they continually finding better architectures and learning algorithms or maybe even better data? Or are they simply making the models bigger?

Comments

docshock•7m ago
Who knows? They rent us a black box, and we can either choose to pay or not. The black box is not open for inspection.
docshock•6m ago
To your point, though, I think mostly what we are seeing is bigger. Everything is getting bigger.