
OpenAI's Astra Cracks 10 Math Problems That Stumped Humans for Decades
OpenAI's unreleased Astra model solved 10 decade-old open problems in math and CS, verified in Lean 4, for roughly $2,000 in compute
20 stories

OpenAI's unreleased Astra model solved 10 decade-old open problems in math and CS, verified in Lean 4, for roughly $2,000 in compute

Parlor runs voice and vision conversations entirely on a laptop, using Gemma 4 E2B and Kokoro to hit sub-3-second replies with 3 GB of RAM.

LiveKit's open-source Python framework for realtime voice agents crossed 13K stars, packing telephony, MCP, and semantic turn detection into one stack.

Epoch AI's MirrorCode leaderboard puts Claude Fable 5 at 64% and GPT-5.6 Sol at 20% on full-program reimplementation tasks that can run for weeks and cost thousands of dollars per attempt.
We ran Claude Opus 5 through 260 debugging attempts at Low, Medium, High, and XHigh effort, then compared its reliability, cost, latency, token use, and patch behavior with nine other models.

A community contributor packaged MiniMax H3, the open-weights omni-modal video and audio model, into GGUF quants that fit on consumer GPUs.

Alibaba's 2.4T-parameter Qwen3.8-Max arrives with open weights promised, frontier coding claims, and a stealth arena debut that the community unmasked

The containers/kubernetes-mcp-server gives AI agents a native, dependency-free interface to Kubernetes clusters, turning natural language into real cluster operations.

Sakana AI opens its Japanese-specialized Namazu LLM as an API, built on Kimi K2.6 with Japan-specific fine-tuning, built-in web search, and OpenAI-compatible endpoints.
A 14B ternary Qwen2.5-based model refactored with BitNet-style weights runs on plain CPUs with 8-12 GB of RAM and native tool-call plus robotics tokens.

OpenHanako is a free, open-source desktop AI agent with persistent memory, a configurable personality, and full computer access — no CLI required

An open-source marketplace of 471 plugins and 3,000+ skills gives Claude Code an npm-like package manager for AI agent behaviors

Liquid AI's 2.6B parameter agent model runs entirely on-device at 220 tokens per second, fits in 2.5 GB, and beats models four times its size on tool use.

Agent Zero gives your AI a full Linux desktop, a browser with DOM annotation, and a 100+ plugin ecosystem — all self-hosted in Docker
How Viktor replaces isolated solo bots with shared memory, secure access, and team-wide coordination.

Mureka V9 debuts at #2 on both Artificial Analysis leaderboards, sitting within 1 Elo point of Suno V5.5 for instrumentals — and its DAW-native workflow sets it apart from every rival.
Why “open” now means four different engineering trade-offs, not one model category

NVIDIA Research's Spatial-IQ benchmark reveals that top AI models hit 3D object-counting scores through shortcuts, not real spatial reasoning — and shows how to fix it.

Google's agent evaluation suite hits GA with 20+ metrics, adaptive rubrics, and live production monitoring -- all on one consistent engine

Google's July Gemini Drops brings faster Flash models, system-wide macOS voice control, Spark's global rollout, and personalized image generation for all U.S. users.