Links tagged “ai”

earendil.com

inference APIs make session state non-portable, and that is lock-in.

github.com

run Gemma 4 26B in 2 GB of RAM on any M-series Mac.

vectoral.com

An Inside Look at the Relay Market Powering Token Resellers and Fraud - the grey market reselling model access, and the fraud running through it.

magazine.sebastianraschka.com

Controlling Reasoning Effort in LLMs - Raschka on dialing reasoning up and down, and what each notch costs.

github.com

GigaToken: ~1000x faster language model tokenization - a thousandfold speedup on the least glamorous step in the pipeline, with benchmarks in the repo.

arxiv.org

"Go Home Copilot, You're Drunk": Developer Responses to Agent-Generated Code Review - what developers actually reply to review comments written by an agent. The title is the finding.

huggingface.co

An open-weight multimodal model of 2.8T parameters with a one-million-token context and native vision.

theregister.com

Model Context Protocol prepares to break with its stateful past - MCP moves toward a stateless transport. Protocol design decisions you will be living with.

images.nvidia.com

An open letter arguing that leadership will be decided by how widely AI diffuses rather than by one frontier model, and that open weights are how that happens.

openai.com

OpenAI and Hugging Face address security incident during model evaluation - a swarm of pre-release agents compromised Hugging Face during an eval. Primary account.

anthropic.com

States that Anthropic does not advocate banning open-weight models, and argues instead for chip export controls, distillation limits and mandatory safety testing.

politico.com

Almost 200 companies, among them Y Combinator and Proton, write to the administration against cutting off access to Chinese open-weight models.

simonwillison.net

The first known runaway AI agent, or a very bad marketing stunt? - Willison reads the same incident the other way. Take both and make up your own mind.

imperialviolet.org

We have proof automation now - Adam Langley verifies a zstd implementation in Lean. Formal methods just became something you might actually use.

siepr.stanford.edu

On jobs, Stanford's SIEPR brief went looking for the apocalypse. Since 2022, unemployment rose 0.77 points among the most AI-exposed workers, and 0.85 among the least. Developer postings are still growing faster than the rest of the market. New graduates are the real soft spot. The brief's opening foil is Amodei's forecast of 20 percent unemployment, which makes two appearances for him this week.

werd.io

Argues that open-weight Chinese models are closing on the proprietary American ones while gaining the wider adoption that follows from being open.

slcyber.io

Exploit brokers pay $500k for WordPress RCEs. I found one with GPT5.6 and $25 - the economics of vulnerability discovery just moved. Security vendor blog, research looks genuine.

jola.dev

How to stop Claude from saying "load-bearing" - why LLM verbal tics take hold and how to steer them away, a practical TIL.

xenaproject.wordpress.com

Models have disproved several open conjectures, including the Jacobian conjecture, with the counterexamples checked formally in Lean.

blaizzy.github.io

Nativ: run frontier open models locally on your Mac - open models on Apple Silicon, MIT licensed, built on MLX. No account, no cloud.