Pi Codex Connectors is a Pi extension that lets any Pi model access and invoke Codex/ChatGPT connectors (GitHub, Gmail, Calendar/Drive, Slack, Linear, Figma, etc.) by starting a local Codex app-server and opening an ephemeral, model-free Codex thread that exposes a small, stable tool surface. Instead of shipping many individual schemas to the model, it exposes three tools: list available connectors and tools, fetch a tool's full schema, and invoke a tool with JSON arguments. Connector discovery and calls run through the Codex app-server while Codex retains login credentials; tools marked writable require explicit approval before execution unless a permissive policy is configured.
Installation requires Pi and Node.js 22.19+, and the package is installed as a Pi extension (npm:@wileai/pi-codex-connector) or added to Pi settings. Security and privacy controls include session consent, PI_CODEX_CONNECTORS_DATA and PI_CODEX_CONNECTORS_ALLOW to limit exposure, and PI_CODEX_CONNECTORS_WRITES modes (ask/allow/deny) plus write-argument-length checks. Tests require a Codex account and use scripted models for CI. Operational caveats: large results are truncated in memory, stderr and raw diagnostics are discarded, and the app-server protocol lacks per-call cancellation (aborts report UNKNOWN OUTCOME and can block subsequent writes). Releases publish via npm/GitHub and the project is MIT-licensed.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
AstaBrief is an open-source model designed for fast scientific report generation, capable of turning research questions and literature into cited reports. It reduces generation time by nearly three and a half times compared to proprietary models, making it a useful tool for researchers. (allenai.org)
Shrivu Shankar argues that SaaS companies are evolving into 'harnesses,' where core tasks are performed by AI agents rather than humans. This shift transforms organizational structure, with humans overseeing and reviewing AI-driven processes within the company. (blog.sshh.io)
Three AI agents - Meta’s Muse, Anthropic’s Claude, and OpenAI’s GPT - were tested on multilingual research tasks involving U.S. and Iranian data. The experiment focused on how language and context influence agent performance across reasoning, sourcing, and artifact creation. (royapakzad.substack.com)
An open-source project named LDRAW-NOVA allows users to generate Lego models using AI. The repository contains tools, examples, and documentation for creating Lego designs with artificial intelligence. (github.com)
GPT-2, released in 2018, required an estimated 10²¹ FLOPs to train, making it a significant milestone in AI development. The article discusses why similar models or advancements did not occur earlier, such as in 2005, due to technological and resource limitations. (dynomight.net)
DwarfStar 4 (ds4) is a local inference engine that runs large language and vision models on high-memory Mac, CUDA, and ROCm machines. It supports models like DeepSeek V4 and Qwen3.8, offering CLI, APIs, and persistent cache features for efficient local deployment. (dwarfstar.sh)
Today's best Hacker News stories, summarized and screenshotted, one email a day.