-
Neural Particle Automata grows self-organizing AI from moving particles
A research project drawing attention on Hacker News, Neural Particle Automata (NPA), asks a simple question: what if the cells of a self-organizing system weren’t pinned to a grid? It comes with an interactive demo, which is part of why it’s circulating. ## What it is NPA is a generalization of Neural Cellular Automata —… Continue reading
-
Terminal Mode puts your coding agent status on AR glasses
Coding agents now run for minutes at a time, and you end up babysitting a terminal to see what they’re doing. Terminal Mode, one of the launches fighting for the Mac desktop this week, takes a different route: it projects your coding agent’s live status onto AR glasses. ## What it does Instead of another… Continue reading
-
Qwen3-Coder-Next hits 70% on SWE-Bench Verified as an ultra-sparse open model
Alibaba’s Qwen team shipped Qwen3-Coder-Next, an open-weight model built for coding agents rather than chat. It’s ultra-sparse — built on the Qwen3-Next-80B-A3B base with hybrid attention and MoE, activating only a fraction of its parameters per token for roughly 10x higher throughput on repo-scale tasks. ## The benchmarks It clears 70% on SWE-Bench Verified using… Continue reading
-
VibeThinker-3B matches frontier reasoning models at 3 billion parameters
Sina Weibo’s AI team put out a claim that set off another benchmark fight: a 3-billion-parameter model reasoning at the level of systems hundreds of times its size. VibeThinker-3B is open, MIT-licensed, and runs on a single consumer GPU. ## The benchmarks On AIME 2026, the US math competition, VibeThinker-3B scores 94.3 — level with… Continue reading
-
GLM-5.2 lands within a point of Claude Opus 4.8 on coding at a sixth of the cost
Z.ai (Zhipu AI) released GLM-5.2 on June 13, and the story is the gap that isn’t there: on FrontierSWE it scores 74.4% against Claude Opus 4.8’s 75.1% — a near-tie with the proprietary leader, from an open-weight model you can download. ## The benchmarks GLM-5.2 is a 753-billion-parameter model under an MIT license with a… Continue reading
-
Oak is a version control system rebuilt for AI coding agents
Git was built for humans typing commit messages. AI coding agents don’t work that way — they branch constantly, snapshot mid-task, and touch huge repos they never fully read. Oak, shown on Hacker News this week, is a version control system rebuilt for how agents actually operate. ## What’s different Oak mounts a repo without… Continue reading
-
HyperFrames turns HTML into MP4 video that AI agents can render
AI agents are already fluent in one thing: writing HTML. HyperFrames, HeyGen’s open-source framework, turns that skill into video — an agent writes an HTML file, HyperFrames renders a deterministic MP4. It’s been surging on GitHub this week as teams wire it into agent pipelines. ## How it works A composition is just an HTML… Continue reading
-
Moebius matches 10B image models for inpainting at 0.2B parameters
Most state-of-the-art image inpainting runs on 10-billion-parameter models like FLUX.1-Fill-Dev. Moebius, open-sourced June 18 by researchers at Huazhong University of Science and Technology and vivo AI Lab, does the same job with 0.22B parameters — under 2% of the size — and in some cases does it better. ## What Moebius does Inpainting means filling… Continue reading
-
GitHub Agent Finder lets Copilot discover MCP servers and skills at runtime
Every Copilot user knows the tax: hand-wiring which MCP servers, skills, and agents each task needs, then watching them eat your context window whether you use them or not. GitHub Agent Finder, live since June 17 on all Copilot plans, hands that job back to Copilot itself. ## How it works Describe a task in… Continue reading
-
Sakana Fugu orchestrates a pool of LLMs behind one API call
The hard part of multi-agent AI was never the models. It was the wiring — routing, planning, verifying, stitching outputs back together. Sakana Fugu, released June 22 by Tokyo’s Sakana AI, buries all of that inside a single model you call like any other. ## What Fugu actually is Fugu isn’t a bigger LLM. It’s… Continue reading
Top AI Product
Every day, hundreds of new AI tools launch across Product Hunt, Hacker News, and GitHub. We dig through the noise so you don't have to — surfacing only the ones worth your attention with honest, no-fluff reviews. Explore our latest picks, deep dives, and curated collections to find your next favorite AI tool.
Top AI Product 