OpenCyclone is a free, open-source Android app that aims to make an existing phone work like an AI-operated phone, without buying new hardware. Its agent reads and interacts with apps on a hidden screen while the owner continues using the device. Users can give it tasks such as sending a message or setting an alarm; before actions that send, spend, or delete, it displays what it plans to do and requests approval. The project emphasizes that this approval mechanism is implemented in readable, modifiable code rather than relying only on instructions to the AI. It also supports different model providers, including ChatGPT, OpenRouter, API keys, and locally hosted models, and offers tools through the Model Context Protocol.
The project is presented as an early, independent experiment, not a finished product or an official Cyclon offering. Short tasks within one app work best, while longer tasks spanning multiple apps fail more often; it cannot answer calls because stock Android does not permit the necessary access. The developers say the software was built quickly by one person with AI coding agents, and benchmark results and safety testing have not yet been completed. Installation requires Android 12 or later, a computer for a one-time setup, and access to a model provider, though a local model can cost nothing. The Apache-2.0-licensed code is available for others to inspect and adapt. The central argument is that AI phone features can be delivered as software, so users should try their existing Android devices rather than purchase new hardware.
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)
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)
Gen Z is spending more on premium AI subscriptions to boost productivity, run businesses, and manage everyday tasks. Financial advisers say paid tools can be worthwhile when they deliver clear value, but users should check costs and prioritize essential savings. (reuters.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.