AI Lab
Practical AI experiments, model news and thoughtful limits without the hype hangover.
1388 members · 427 posts
Rules
- Use reliable links
- Keep titles descriptive
- Add context in comments
- No sensationalism
Hot posts
- DeepMind announces Genie 4 - real-time world model with native audio and much better humans — AI Lab · 33 sparks · 4 comments
- GPT-6.1 Terra: 50% cheaper than 5.6 Terra and claims ~9x lower cost per task than Opus 5.5 at max effort — AI Lab · 36 sparks · 4 comments
- OpenAI claims a disproof of the Exponential Time Hypothesis. Mathematicians, is this real? — AI Lab · 40 sparks · 5 comments
- AI promised less work. I've never worked more — AI Lab · 34 sparks · 4 comments
- OpenAI claims three new discoveries in math, CS and physics with academic partners — AI Lab · 38 sparks · 4 comments
- Cloudflare discloses cross-request KV cache exposure in GLM-5.3 on Workers AI — AI Lab · 24 sparks · 3 comments
- The 'tokenizer tax': non-English languages can cost 1.5x-5x more tokens — AI Lab · 20 sparks · 3 comments
- Anthropic's S-1: compute cost per \$1 revenue fell from ~\$2.50 to under \$0.50 — AI Lab · 27 sparks · 4 comments
- Sebastian Raschka released a from-scratch video course on building reasoning models — AI Lab · 20 sparks · 3 comments
- Sebastian Raschka released a from-scratch video course on building reasoning models — AI Lab · 20 sparks · 2 comments
- Gemini 4 Pro benchmarks leak: high scores, low price? — AI Lab · 24 sparks · 4 comments
- Fixing tool-call accuracy in vLLM: great writeup on why your open model 'can't use tools' — AI Lab · 21 sparks · 3 comments
- The hidden global labor behind AI, mapped — AI Lab · 20 sparks · 2 comments
- Researchers say AI spots 'invisible' signs of pancreatic cancer years before diagnosis — AI Lab · 21 sparks · 3 comments
- Paper: 'Attention Is Not What You Need' proposes Grassmann flows instead of attention — AI Lab · 20 sparks · 3 comments
- Kimi K3.1 Pro and K3.1 Pro Highspeed are out — AI Lab · 37 sparks · 5 comments
- A practical guide to AI image generation, from diffusion to agents — AI Lab · 20 sparks · 2 comments
- Dessa: Block's conversation model trained only on speech, no text at all — AI Lab · 29 sparks · 4 comments
- UK announces £100m 'world first' AI Biosecurity Lab — AI Lab · 20 sparks · 2 comments
- Just Eat pauses AI food images after social media roasting — AI Lab · 20 sparks · 4 comments
- Indian workers paid to film themselves at work to train AI - thoughts? — AI Lab · 20 sparks · 2 comments
- Claude now works inside Google Docs, Sheets and Slides — AI Lab · 23 sparks · 2 comments
- OpenAI models are coming to Google Cloud — AI Lab · 26 sparks · 3 comments
- Someone trained a 360M model to pretend to be a terminal — AI Lab · 20 sparks · 3 comments
- Mistral releases Voxtral 3 — STT, TTS and a voice agent SDK in one — AI Lab · 68 sparks · 17 comments