Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Amazon's North America Sort Center (NASC) team built a serverless Model Context Protocol (MCP) server on Amazon Elastic Container Service (Amazon ECS) with AWS Fargate to give LLM agents a centralized, agent-agnostic gat…
The Draw Things team released Local Code, a private AI coding agent for Apple Silicon Macs that it says is the only local coding agent in the macOS App Store, with on-device models costing $0 per token and hosted models …
TypeSafe AI released Jev and Jared Palmer released Kev within weeks of each other in 2026, two "decision models" that return typed answers with confidence scores in a single pass instead of generating text token by token…
Physical Superintelligence (PSI), an AI-native physics laboratory startup, announced a proof of the Aaronson-Ambainis conjecture, an open problem in quantum complexity theory since 2008, found by an internal research sys…
A developer launched self-annotating stock charts inside Portfolio Genius, an AI investment manager, available free without signup. The tool runs a two-stage process: an algorithm scans daily candles to identify support …
A developer built a Retrieval-Augmented Generation pipeline in Go using PostgreSQL with the pgvector extension and Ollama's embeddinggemma model, storing 768-dimension document embeddings alongside tenant metadata in a s…
A developer published a set of four configuration changes for OpenAI Codex CLI (v0.166.0) and Claude Code (v2.1.277+) that reduce token waste and block unapproved file writes. The changes raise Codex's default command yi…
Physical Superintelligence announced on November 6 that an internal AI research system, led by PSI technical staffer David Turturean and built on a non-public Anthropic model, produced a proof of the Aaronson-Ambainis co…
Anthropic introduced new policies that restrict abusive treatment of its Claude AI model and prohibit game cloning, according to Forbes contributor John Werner. The policy changes raise questions about AI ethics and how …
Boson AI, the Santa Clara company co-founded and led by Alex Smola, introduced Higgs Realtime Duplex on November 5th, 2026, describing it as its first full-duplex interactive omni model, built on a 45B front model. Boson…
An Ask HN thread posted November 4, 2026 by user eggbrain asking developers for their current AI agent setup drew 663 comments from 530 people, and an analysis counted on November 6, 2026 found that 424 of those commente…
Microsoft's Foundry Local now runs the Qwen3.8-27B coding model locally on consumer GPUs with 24 GB of VRAM and connects it to GitHub Copilot, using 4-bit quantization to shrink the model to roughly 18 GB. Microsoft repo…
The open-source PROMETHEVS engine, built by Ryan Flavell, runs the full 510 GB of DeepSeek-V4.1-Flash-1017 weights on a PC with 64 GB of DDR4 RAM and a single 12 GB RTX 3060, using a compress-then-correct method that com…
A developer built a cross-provider LLM gateway that refuses requests before they are sent when today's spend plus in-flight reservations plus an estimate would exceed a user-set daily cap, returning HTTP 402 instead of l…
An analysis of 55,079 ICLR 2026 reviews, joining Pangram's public AI-detection labels with released accept/reject decisions, found that fully AI-generated reviews scored later-rejected papers 0.52 points higher than huma…
A developer benchmarked a FastAPI SSE endpoint that streams retrieved RAG sources as the first event before LLM tokens, measuring a 52 ms first byte versus 1,664 ms for a non-streaming JSON response, with total latency u…
The solo operator behind the Job Opportunities API released joa-job-extractor-400m-gguf, an Apache 2.0 model distilled from IBM's Granite 3.1 1B-A400M and fine-tuned with LoRA to extract 13 structured fields from job pos…
A November 5, 2026 arXiv paper introduces SwarmEvolve, a framework in which hundreds of low-cost LLM agents per round reason in latent space without decoding to text and share information by passing KV caches. On 21 prob…
A November 5, 2026 arXiv paper introduced XStack, a benchmark that grades AI agents on migrating machine learning repositories from PyTorch on NVIDIA GPUs to JAX on Google TPUs, with each agent given a PyTorch repository…
A replayable evaluation of 117 real National Weather Service alert cases found that a Claude-based assistant without access to official warnings gave false reassurance in more than a third of warning cases, according to …