20 articles
Dario Amodei published Anthropic's stance on open-weights models this week - no total ban, but support for chip export controls, distillation crackdowns, and mandatory safety testing. HN responded with 800+ comments calling it regulatory capture. Here is what the CEO said, what the thread argued, and why the debate matters for every developer deploying AI.
Moonshot AI's Kimi Linear paper introduces KDA, a hybrid linear attention that beats full attention at all scales - 75% less KV cache, 6x decoding at 1M context, and open-source checkpoints.
Moonshot AI released the full Kimi K3 weights on HuggingFace today - 2.8T parameters, 1M context, native MXFP4 quantization, ~1.63TB download. The HN community reaction, what the license really says, and why this matters for the open-weights AI market.
Compare every verified Kimi K3 access route, including Moonshot, Together, Fireworks, Baseten, Modal, Vercel AI Gateway, Cloudflare, RunPod, SiliconFlow, OpenRouter, and OpenCode Go.
Moonshot AI releases Kimi K3 with 2.8 trillion parameters, 1M context window, and Delta Attention architecture. Here's what developers need to know about pricing, performance, and where it fits in the frontier model landscape.
Martin Alderson's argument for why open-weights models like GLM 5.2 will compress frontier lab margins is sparking debate on HN. Here is what the thesis actually says, where HN agrees and disagrees, and why it matters for developers choosing models.
Unsloth's dynamic quantization makes GLM-5.2 runnable on a 256GB Mac or a 24GB GPU with CPU offloading. Here is the hardware math, the quantization tradeoffs, and what the HN community learned from actually running it.
GLM-5.2 ships under an MIT license, so it is hosted everywhere - and a few places run it for free or nearly free right now. Here is every way to access Z.ai's open-weights coding model, from OpenCode Go referral credits and Devin to the cheapest per-token routes on OpenRouter, Fireworks, and DeepInfra, plus local Ollama.
New benchmark data shows GPT-5.5 hallucinates 86% of the time when it does not know the answer - versus 28% for the open-weights GLM-5.2. The numbers challenge the assumption that bigger models equal more reliable output.
No single model wins every task anymore, and the companies that never trained one - Factory, Devin, Perplexity, Cursor, OpenCode - are turning that into a moat. This is how model routing works, why open weights and neoclouds make it cheap, and the honest counter-argument.
JetBrains released Mellum2 on June 2, 2026 - a 12B MoE model with only 2.5B active parameters per token. Here is how to run it locally, when to use it, and where it fits in your AI coding stack.
A $500M accidental Claude bill and an open-weights model beating GPT-5.5 at one-sixth the cost point to the same conclusion: the margin is moving to the layer that decides when to use which model for what. Here is how routing and orchestration differ, and how to cut your model spend.
DeepSeek V4 Pro lands a 63.5 on SWE-bench Verified at $0.435/$0.87 per million tokens, and Flash runs agent inner loops for cents. Here is the worked cost math, the Flash-vs-Pro split, and a clear guide on when to route to DeepSeek instead of a frontier model.
Showing 12 of 19 articles
New tutorials, open-source projects, and deep dives on coding agents - delivered weekly.
Explore 813 topics
Browse All Topics