Keller Jordan published Muon, an orthogonalized optimizer short for MomentUm Orthogonalized by Newton–Schulz, as a blog post in December 2024 after it cut the CIFAR-10 training record from 3.3 to 2.6 …
A tutorial builds a generative AI recommender from scratch on the Amazon Reviews 2023 dataset from McAuley Lab (UCSD), using the Video Games category in its 5-core version where every user and item ha…
Claude Code 2.1.285's tool search feature cut the first request's token load by 64%, from 27,184 tokens to 9,722, with five MCP servers and 65 tools connected, according to measurements published by t…
A three-part tutorial series shows developers how to build an art and museum recommender from scratch using CLIP embeddings and ChromaDB, with part one covering data retrieval from the Rijksmuseum API…
A local research assistant built on DeepAgents, LangGraph, ChromaDB and Llama 3.1 8B Instruct raised citation presence from 30.8% to 97.4% on answerable holdout turns after changes to reference handli…
SpaceX's reported all-stock acquisition of Anysphere, the maker of Cursor, at $60 billion values a coding product whose real asset is the per-turn policy — what the model sees, which tools it can reac…
A voice agent named Alex negotiates freight rates over the phone using a pipeline of six small models built on LiveKit Agents, an open source framework, rather than a speech-to-speech model, because e…
A LiveKit voice agent named Alex, built as a carrier sales rep for a fictional freight brokerage, cut the margin it gave away in rate negotiations from an average of 40% in its first version to 14% in…
A Generative Adversarial Network (GAN) consists of two neural networks — a Generator that turns random noise into fake data and a Discriminator that outputs the probability a sample is real — trained …
A practical Langfuse tutorial demonstrates how to add LLM observability to a LangChain-based support-ticket routing application, tracing each run through ChatPromptTemplate and ChatOpenAI to expose pr…
A developer who has built AI and AI-native software, including work with frontier models such as Claude Opus, published a reusable four-file Markdown harness — AGENTS.md, SESSION.md, CHECKLIST.md, and…
The Superpowers plugin for Claude Code, a set of 15 skills whose GitHub repository has about 294,000 stars, makes coding agents ask clarifying questions and present a plan before writing code, accordi…
TypeSafe AI launched Jev on September 15, 2026, introducing "System One models" that take a state and a set of options and return a probability for each option in a single forward pass, with the same …
TypeSafe's structured decision model Jev, benchmarked as a quality judge on the 1,000-rubric Feedback-Collection dataset, was tested against two low-cost general-purpose LLMs on a 750-example held-out…
A marketing agency operator built a working MCP (Model Context Protocol) server in roughly 80 lines of Python using the official MCP Python SDK and its FastMCP helper, exposing four tools — add_lead, …
A replication attempt testing whether knowledge-graph structural metrics reflect a graph's subject or merely its extraction pipeline found the same prerequisite-concept density across three unrelated …
TypeSafe AI released Jev, the first model in its System One class, in early access with a waitlist from September 15, 2026, returning typed probabilistic decisions instead of generated text. Jev offer…
Anthropic reported on September 23 that roughly 950 Claude agents worked for 21 hours across sequence data and helped researchers identify an unusual enzyme system, after which human scientists contin…
A four-tier intent-routing cascade ordered by cost, with each tier allowed to abstain and every exit leaving a replayable log record, can replace a single frontier-model llm.classify(query) call for a…
A conversational AI voice agent deployed as a clinic receptionist failed to escalate a 58-year-old caller reporting crushing chest pain radiating down the left arm, instead checking the EHR scheduling…