Articles, releases and code from Hacker News, Reddit, GitHub and the people building RPA, workflow automation and AI agents — plus what the community pushed to the top today.
AIBigarrow is an open-source macOS CLI tool that lets AI agents display click-through screen overlays to guide humans during interactive approval steps.
AIOpenAI fired three safety researchers over alleged information mishandling, signaling stricter internal information controls while raising concerns about external oversight and collaboration for autonomous agent safety.
AIEdi Life OS provides an open-source personal dashboard that includes an MCP server, enabling AI agents to directly manage personal habits, tasks, and productivity workflows.
AIDevelopers should route AI agent context retrieval across vectors, keyword search, graphs, and live queries based on task shape rather than relying on a single retrieval method.
AIGoogle transferred Docsy to the Linux Foundation and added llms.txt and Markdown generation, helping AI agents more easily retrieve technical documentation for automated coding workflows.
AIAPIblaze launched a serverless gateway that converts OpenAPI specs into Model Context Protocol servers and proxies with built-in authentication and authorization for AI agents.
AIIntegrum is an open-source library that turns any Python module or class into a Model Context Protocol server, enabling AI agents to use existing code as tools.
AIThunk.AI released an open benchmark for pharmacovigilance intake, giving automation teams test data to measure how reliably AI agents handle regulatory triage and adverse-event reporting workflows.
AIGoogle released Android Bench 2.0, adding long-horizon tasks, agentic evaluation, and continuous scoring to help developers assess AI agents on complex, multi-day mobile engineering workflows.
AIGoogle launched an enterprise agentic AI for Gemini that integrates with internal business systems and Model Context Protocol servers to execute multi-step workflows.
AIResearchers introduced AIDA, a multi-agent framework that uses adversarial challenges to reduce LLM false-negative rates in automated alert triage from over 40% to 3.1%.
AIGoogle launched an enterprise Gemini agent that orchestrates multi-step workflows across systems and can route tasks to external models like Anthropic's Claude.
AIA new framework equips multi-agent systems with explicit partner mental models, improving coordination and decision-making by inferring the hidden beliefs and intentions of other agents.
AIRisotto launched an AI agent that diagnoses Tier-2 IT tickets and automatically converts resolved edge cases into reusable, self-improving operational runbooks.
AIResearchers introduced a multi-agent RAG framework for conversational shopping that decomposes tracking, retrieval, and reasoning to maintain consistent user preferences across long, multi-turn interactions.
AISimon Willison released ttok 1.0, updating
AIConventionPlay trains AI agents to probe partner capabilities and steer them toward optimal shared conventions, improving ad-hoc collaboration in multi-agent workflows.
AIGitHub migrated its Copilot runtime from Node.js to Rust, enabling developers building automation tools to embed Copilot in-process with sub-second startup times and lower memory overhead.
AIQuandri released AI Workflows, expanding its platform beyond renewals to automate commercial policy check-ins, client servicing, and revenue tasks across P&C insurance agency books.