Daily AI news, sourcedUpdates / All
Updates Unconfirmed · 662 items · newest issue first · page 12 of 34
AreaAllModelsAgents & codingApps & productsChips & infraResearchPricing & limitsBusiness & fundingPolicy & safetyOther
- Developer launches a benchmark site that ranks only open-weight models, with coding and agentic testsUnconfirmedr/LocalLLaMAOct 10Models
- User runs a Qwen3.8-Flash-Next variant at 20-30 tokens per second on 12GB VRAM with a custom Strata forkUnconfirmedr/LocalLLaMAOct 10Models
- Maintainer of a widely used open-source local MCP server releases version 2Unconfirmedr/LocalLLaMAOct 10Agents & coding
- Developer shares a 24KB stand-alone HTML language model that generates short stories in the browserUnconfirmedr/LocalLLaMAOct 10Models
- User tests Mellum2.1-12B-A2.5B as a local coding agent and finds it usable but weak on one-shot projectsUnconfirmedr/LocalLLaMAOct 10Agents & coding
- Developer says Aplomb 1 tops Hugging Face's Typed Decisions benchmark for probabilistic decisionsUnconfirmedr/LocalLLaMAOct 10Models
- Open-source CLI lets AI coding agents draw arrows, boxes and text on a Mac screenUnconfirmedHacker NewsOct 9Agents & coding
- AI company executives are privately planning for a public backlash after a catastrophic AI event, a post saysUnconfirmedr/singularityOct 9
- A Reddit user says they miss reading AI models' chain of thought as they reasonUnconfirmedr/singularityOct 9Apps & products
- Seattle bars grocery stores from setting prices with AI based on personal data, a post citing the mayor's office saysUnconfirmedr/singularityOct 9Policy & safety
- Essay "What Are the Monkeys Typing?" argues we can see what capable AI systems do but not reliably whyUnconfirmedr/singularityOct 9Research
- Paper tests whether AI agents can do open-ended scientific discovery in the Station multi-agent environmentUnconfirmedr/MachineLearningOct 9Research
- Developer shows ALHR, a tree-based sparse attention method claiming 35x KV-cache read reductionUnconfirmedr/MachineLearningOct 9Research
- Developer shares Memory Journal, a portable text-file format for carrying AI context and skills between chatsUnconfirmedr/LocalLLaMAOct 9Agents & coding
- Developer releases MaRN, a PyTorch library that trains networks through compact low-dimensional parameter mappingsUnconfirmedr/MachineLearningOct 9Models
- A 100-object text-to-3D-game-object test ranks GPT-6 Astra most reliable while Opus 5.5's results look better, a Reddit post saysUnconfirmedr/singularityOct 9Models
- DLoop paper proposes looped speculative decoding for faster LLM generationUnconfirmedr/LocalLLaMAOct 9Research
- Stepped MoE paper proposes segment-level routing with configurable inference costUnconfirmedr/LocalLLaMAOct 9Research
- Free in-browser calculator estimates the electricity cost of running a local LLM server, developer saysUnconfirmedr/LocalLLaMAOct 9Apps & products
- NInfer runs uncensored Qwen3.8-27B on an RTX 4090 at 262K context and about 130 tok/s, developer saysUnconfirmedr/LocalLLaMAOct 9Models