A developer built nano-harness, a coding agent in ~970 lines of Python that scored 59.6% (53/89) on the full Terminal-Bench 2.0 suite using Claude Opus 4.8. The harness uses only three tools—bash, rea…
A developer released Autoretrieval, a framework that uses an AI agent to automatically optimize RAG pipelines by running experiments against an evaluation dataset. The agent iteratively edits retrieva…
In November 2025, Anthropic released Claude Opus 4.5, a model that made AI agents capable of autonomous task execution, leading to a dramatic shift in software development and personal productivity. N…
Context engineering has emerged as a critical discipline for building effective AI agents powered by large language models (LLMs), governing everything an LLM sees at inference time including system i…
LangChain released OpenWiki Brains, a memory enhancement for large language models that builds a live-updating wiki from external sources such as Gmail, Notion, and X, nine days after the initial Open…
A developer created a /council skill for Claude Code that convenes a five-advisor council to pressure-test decisions, plans, or ideas. Based on Andrej Karpathy's LLM Council, the skill uses five subag…
Taleeb Shahbaz argues that AI will not replace developers in 2026, despite the rise of 'vibe coding'—a workflow popularized by Andrej Karpathy where developers describe what they want and let AI gener…
A developer built a discrete diffusion model for character-level text generation using Andrej Karpathy's Shakespeare dataset, implementing a continuous-time Markov chain (CTMC) with a rate matrix to h…
A Hacker News user asks the community for use-cases and success stories with LLM-Wikis, inspired by Andrej Karpathy's concept, noting they have not yet found value in the approach. Applications for th…
Anthropic reversed its policy of covertly degrading Claude Fable 5 for AI-research-related use cases within roughly a day after public backlash, with critics arguing that opaque sandbagging violates t…
A Google employee wrote a dependency-ordered series of over 80 posts walking through the full LLM stack from linear algebra to agent protocols, motivated by a fear of not understanding the layers bene…
A third-party repository inspired by Andrej Karpathy's observations about AI coding failures has amassed over 189,000 GitHub stars. The project, created by Jiayuan Zhang under the account multica-ai, …
Philosopher John Haugeland suggested 40 years ago that artificial intelligence should be called synthetic intelligence, and business leaders now need to understand the distinction as five key signs ma…
Andrej Karpathy released nanochat 2026, an open-source full LLM pipeline in 8,000 lines that he calls a 'ChatGPT for $100', while AgentMemory became the top persistent memory system for AI coding agen…
NVIDIA has demonstrated an autoresearch workflow using RL agent skills and the NeMo framework, where a frontier coding agent (Codex with GPT 5.5) autonomously set up a NeMo RL and NeMo Gym stack, crea…
A growing number of accomplished founders and executives, including Monzo co-founder Tom Blomfield, Instagram co-founder Mike Krieger, former OpenAI founding member Andrej Karpathy, and Workday CTO Pe…
Tom Blomfield, co-founder of GoCardless and Monzo, announced Monday he is taking a leave of absence from Y Combinator to join Anthropic's compute team as a member of technical staff, joining a wave of…
Andrej Karpathy's LLM Wiki method uses Obsidian and Claude Code to build an AI-maintained knowledge base, turning passive notes into a living document that a language model can read, update, and reaso…
Anthropic has hired Tom Blomfield, cofounder of British fintech Monzo, as a member of technical staff on its compute team, the AI startup announced Monday. Blomfield is taking a leave of absence from …
A team built a persistent LLM Wiki to stop their agent from re-deriving the same answers from scratch, using a directory of markdown files on a dedicated repo branch. The wiki classifies questions int…