frontpage.

I am experimenting with the model and then model spends all its output tokens while thinking making no room left for final output. I have even set thinking budget, but still does not work, anybody has any workarounds or something I am missing?

Wellformed: Validation Schemas as JSON for TypeScript and Rust

DeepSeek V4 official release coming in mid-July with 2x peak-hour API pricing

Universal agents require universal memory

The state of the AI economy from bottom up

Open Hardware and Free Software: Teufel Mynd, a Case Study of a BT Loudspeaker

QDBP: Explicit depth markers as an alternative to indentation and parentheses

AI Policy Update

Reward hacking is swamping model intelligence gains

Vega: Zero-knowledge proofs for digital identity in the age of AI

Gemma 4 on Cerebras - The Fastest Inference Is Now Multimodal

Show HN: Bored People Chat – Anonymous global chat room

I built 25 executable skills for my AI agent � all open source

Another Semiquincentennial

Ask HN: Which is the best local model under 3B parameters today?

The op log was peer-to-peer the whole time

I built a free invoice generator for freelancers – no login, no subscription

Operation RYaN

We built a P2P app with no servers. 1M users didn't miss them [Video]

Tangled CI runs on microVMs

Manifest-Driven Development

Meshtryoshka: Differentiable Mesh Rendering for Unbounded Scenes

OGN 3D Viewer – glider flights replayed in 3D in the browser

PostgreSQL Management Platform

There's this mystery of what, actually, is this thing?: DeepMind's philosopher

Can Facial Recognition tools track you online?

EU Commission HQ forced to shut down air-conditioning amid heatwave

Vāgdhenu: Open-source meter-aware text-to-speech for Sanskrit

What Is Binning? A Basic Definition (2022)

Zero Trace AI – private AI chat with nothing to subpoena

I have open-sourced gojaja, a CLI for local multi-agent collaboration

Qwen3.5 2B burns all the output tokens while thinking