Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A developer built Agent Rigor, an open-source framework that enforces strict empirical discipline on autonomous AI coding agents to prevent the 'Doom Loop' where agents degrade a codebase through repeated errors. The sys…
A lead of the Rust Secure Code Working Group used GPT-5.5 via the Codex for Open Source program to audit Rust crates, finding dozens of memory safety issues. The developer combined the LLM with the miri tool to eliminate…
A developer created a Polytoken Reviewer Skill that orchestrates two parallel reviewer subagents on gpt-5.5(high) for code review. The skill dispatches a standard reviewer and an adversarial reviewer concurrently, merges…
Agent Rigor, a new open-source framework, provides a structured multi-layer progressive disclosure system to prevent AI coding agents from entering 'doom loops' and other failure modes. The framework enforces empirical d…
Developer Juan Garza released Cody, an open-source Neovim plugin that enables voice control of the editor using its native commands, LSP, and plugin integrations. The tool routes spoken commands through a Node.js bridge …
Foxconn Chairman Young Liu estimated that building a 1-gigawatt AI datacenter using NVIDIA's Vera Rubin architecture could cost up to $47 billion, with 3,557 racks at $9.1 million each and an annual electricity bill of $…
A developer rebuilt the 90s deduction game Dream Phone as a two-player race against Sköll, the Norse wolf, in a game called Save the Sun. The game uses Gemini to interpret player questions into structured queries for a d…
A developer at AI Got Ranked compared Cursor, GitHub Copilot, and Windsurf across six metrics without paid rankings. Cursor is an AI-first IDE with deep codebase awareness, GitHub Copilot is the trusted incumbent with a …
A developer argues that token optimization for LLM costs is a context engineering problem, not just prompt shortening. By instrumenting token usage and analyzing cost distributions, teams can reduce bills by 60-80% throu…
Magnitude launched a coding agent that runs entirely on open models, costing 60% less than Claude Code with no drop in performance. The agent uses smart model routing to achieve 50–90% cost savings by assigning subtasks …
A new chess benchmark, chess-bench, ranks AI models by performance, with Gemini 3.5 Flash leading at 61.3%, followed by Grok 4.1 Fast at 58.7% and Gemini 3.1 Pro Preview at 55.3%. The test included Claude Fable 5, which …
Maneshwar, developer of git-lrc, explains the ReAct (Reasoning + Acting) prompting pattern from Yao et al., 2022. ReAct interleaves reasoning traces with actions like search or lookup, allowing language models to ground …
A developer argues that event-driven architecture is often overkill for expensive, stateful jobs that aren't latency-sensitive, advocating instead for a simple polling loop after experiencing silent failures and complexi…
A software engineer catalogs personal AI use-cases, scoring them 1–5 for usefulness, and finds a pattern: rubberducking with AI scores 2/5 because it risks generating noise and losing ownership of thoughts. The engineer …
Senior engineers can now learn the internal architecture of large language models in a new five-part series. The second installment explains why recurrent neural networks have been replaced by transformers, which use att…
SouthBridgeAI released offmute-v2, an open-source multi-step pipeline for timestamp-correct, diarized transcripts with identified speakers, after GLM-5.2 outperformed Opus 4.8 on an AI-resistant backend take-home. The to…
A self-taught delivery rider from China built a multi-expert AI system that runs multiple LLM-powered expert agents in parallel, enabling cross-validation and debate. The MVP, developed after learning to code in 2026, fe…
Sarvam AI raised $234 million at a $1.5 billion valuation, becoming India's newest unicorn as sovereign AI moves from slogan to necessity after US restrictions on Anthropic models highlighted foreign dependency risks. Th…
Google has deployed a new export of Gemini 3.5 Flash after identifying and mitigating a bug that caused unusually high rates of output text looping, disrupting conversations and coding workflows. The update is rolling ou…
A Hacker News user asks the community how to make Claude Code produce deterministic results, sparking debate about the fundamental non-determinism of large language models and whether 100% reliability is achievable or ev…