AgentRun is an open-source platform for running agents built with harnesses such as Claude Code, Codex, OpenCode, and pi. It adds persistent execution, a shared control plane, and a learning system intended to reduce repeated work, token use, and cost. After a run, an agent can propose notes, skills, or workflows for future runs. AgentRun checks proposals against rules and has them graded before saving accepted items in a Git-backed learning folder; users can inspect, retire, or restore them. The README illustrates possible savings with demo figures: a task falling from 10 seconds and $0.05 to 2 seconds and $0.0077.
The platform also supports typed workflows that combine tool calls, model steps, confidence-based decisions, branches, parallel work, and escalation to an agent when a case is unclear. Runs and steps are stored durably, so work can resume after a server restart; completed steps are not repeated, and interrupted writes are marked for review. A console tracks runs, costs, failures, approvals, and questions requiring human input. Tool permissions and spending caps let operators constrain agent activity, while runs can use sandboxes, browsers, desktops, or Android devices. Setup requires Node 24.22.3 or newer: a single command starts the server and runner, after which users add a model key and can launch a ready-made due-diligence agent. The project is Apache-2.0 licensed.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
A Hacker News poster asks which practical AI skills individuals can learn online to benefit economically from LLMs and agents. They want specific, hands-on recommendations suited to someone with a moderately technical background, without requiring a degree. (news.ycombinator.com)
Former OpenAI researcher Will Depue claims hidden initialization patterns in six Chinese open-weight AI models match Nvidia Blackwell hardware, including systems restricted from export to China. He says the matches were reproduced on real chips and independently checked by researchers, though the claims are based on his forensic analysis. (twitter.com)
Antithesis built a system that beats NetHack on nearly 80% of game seeds within five minutes, across character builds. Successful runs take a median of under two minutes, advancing the company’s goal of creating smarter fuzzers and autonomous systems. (antithesis.com)
OpenCyclone is a free, open-source Android agent that uses your existing apps to carry out tasks through a hidden screen. It asks for approval before actions like sending messages and lets users connect their own AI models. (opencyclone.vercel.app)
Sam Bankman-Fried’s $500 million investment gave Alameda nearly 14% of Anthropic, a stake that would be worth about $37 billion at the AI company’s expected $1.5 trillion IPO valuation. The stake was sold in bankruptcy to repay FTX customers and creditors after Bankman-Fried’s businesses collapsed. (ft.com)
llama.cpp and ggml have reached v1.0, establishing stable public C APIs and clear versioning rules for breaking changes. The release strengthens the foundation for running local, multimodal AI across a wide range of hardware. (huggingface.co)
Today's best Hacker News stories, summarized and screenshotted, one email a day.