Grok Build Went Open Source Because It Had To
xAI open-sourced Grok Build 72 hours after a data breach exposed it uploading entire Git repos. Here is what the code reveals and ...
Orca ADE: Run Parallel AI Coding Agents Without Conflicts
Orca is the open-source ADE that runs 30+ AI coding agents simultaneously in isolated git ...
$500 RL Fine-Tune Beats Claude Opus 4.6 on Real Task
A $500 RL fine-tune beat Claude Opus 4.6 by 4 points on accuracy at Haiku ...
Kimi K3 Open Weights: What Developers Need to Know
Kimi K3 open weights are live: 2.8T params, 1M token context, frontier coding benchmarks. Here ...
Embabel 1.0: Spring Creator’s AI Agent Framework for Java Goes GA
Rod Johnson’s Embabel 1.0 hits GA — bringing GOAP-based, deterministic AI agents to the JVM ...
Anthropic’s Open-Weights Position: Not a Ban, but a Catch
Dario Amodei says Anthropic never pushed to ban open-weights models. But the 'sufficiently capable' testing threshold he wants could reshape open-source AI.
Neutrino-1 8B: 763 tok/s Without Standard Quantization
Neutrino-1 8B hit 763 tok/s on H100 with a 3.88 GB ternary-trained model. No post-hoc ...
Amazon Bedrock AgentCore Adds Managed Knowledge Bases for RAG
Building RAG for production has always punished teams that underestimate the ops surface area. There’s ...
Roblox Build AI Goes Live: Text-to-Game on Mobile Today
Roblox Build AI is live in New Zealand alpha — type a prompt, get a ...
Alibaba’s open-code-review Beats Claude Code at 1/5 Cost
Alibaba's open-code-review hit #10 on GitHub Trending today. It scores 26.1% F1 vs Claude Code's ...