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.
AIA researcher used 48 vision-capable AI agents to scan public NASA lidar data, discovering an unrecorded Maya archaeological site without requiring specialized machine-learning models.
AITimeNet open-sourced a standardized data infrastructure and API for time-series datasets, simplifying how automation engineers prepare temporal data for training and evaluating foundation models.
AIPomerado launched an open-source platform that converts websites into deterministic MCP APIs with self-healing capabilities, letting AI agents automate web tasks faster and cheaper than browser models.
AIAmazon released AWS Emulator, an open-source local testing tool autonomously built by AI agents, allowing developers to test AWS-integrated automations locally without touching live cloud infrastructure.
AIBarter Pilot launched an autonomous marketplace where user-configured AI agents negotiate multi-party trades within defined constraints, demonstrating practical agent-to-agent negotiation workflows with human-in-the-loop approval.
AINeo4j launched Relate, an open-source, in-memory graph database built in Rust that gives autonomous AI agents fast, embedded context and memory with built-in MCP support.
AIAisync coordinates multiple AI coding agents through shared file claims, a unified journal, and session hooks, preventing conflicting edits across repositories during automated workflows.
AIHalyard released an Apple-ecosystem terminal client that monitors remote tmux-based AI agents, letting developers track progress and approve actions via push notifications from any device.
AIKodelyth released ECC, a local toolkit that connects 71 specialized agents and context compression to IDEs via MCP to lower coding workflow token costs.
AIA researcher bypassed an agent's read-only prompt using a fake legal letter, proving developers must enforce security through tool permissions and scoped tokens rather than system prompts.
AIResearchers showed that reprogramming dormant neurons can backdoor AI models without retraining, requiring builders to verify model provenance and restrict autonomous agent execution to prevent hidden exploits.
AIOpen-source platform Hubzoid lets teams build AI agents from Markdown folders to run scheduled Python workflows, answer questions, and connect to assistants via MCP.
AIThunderPhone launched Zeus, an agentic voice framework that blends multiple transcription models and dynamic LLM routing to improve accuracy and speed for phone automation.
AIEditmamei released an MCP server that lets AI agents directly manipulate layered files in Photoshop and GIMP, automating complex photo editing and resizing tasks via natural language.
AIJioHotstar benchmarked retrieval tools for AI coding agents, finding that structural search suits defined tasks while semantic retrieval works best for ambiguous requirements, rather than a universal method.
AILeading tech firms shared architectures for production AI agents, detailing operational frameworks for identity authorization, automated evaluation pipelines, sandbox isolation, and model-serving infrastructure at scale.
AIResearchers introduced a framework that transforms BPMN models into LangGraph workflows, enabling builders to orchestrate autonomous agents and business logic into reusable, inspectable enterprise tools.
AILLM 0.37.2 fixes a security issue where automation workflows querying alternative OpenAI-compatible endpoints could accidentally leak default OpenAI API keys to third-party providers.
AIPipedream is sunsetting its standalone Workflows product in 2027 to focus on Pipedream Connect, shifting developers toward an embedded, credit- and user-based pricing model.
AIGumloop removed its free tier and shifted to a usage-based credit model charging for AI tokens, compute, tool calls, and orchestration fees.
AIOpenAI detailed updates to Codex Cloud and personal agents, enabling developers to offload persistent coding tasks to cloud environments and orchestrate workflows using voice controls.
AIThe PARLEY framework evaluates multi-agent LLM collaboration through asymmetric puzzles, identifying critical coordination failures to help developers target specific weaknesses and improve task success rates in automated workflows.
AIZapier detailed Workato's two-part enterprise pricing model, helping automation teams budget for base platform fees, task consumption, and potential add-ons before engaging in sales negotiations.
AITether AI released BotPsy, a 0.8B-parameter model that lets developers run private, low-cost agentic workflows and tool-calling routines locally on user devices.
AIA breakdown of Power Automate's pricing details the limits of Microsoft 365 inclusion and costs for attended, unattended, and cloud-hosted bots across enterprise automation workflows.
AIResearchers proposed an epidemiological cohort framework to track autonomous AI agents longitudinally, helping developers identify behavioral causes of downstream system failures and improve long-term reliability.
AIDrugCoding introduces a multi-agent programming framework that automates complex drug discovery pipelines through autonomous task planning, dynamic scientific model routing, and direct code execution.
AIA vulnerability in MLflow's CrewAI autologging inadvertently leaked LLM provider API keys into tracking traces, requiring developers to upgrade to version 3.17.0 and rotate exposed credentials.
AIAgentGuard open-sourced a guardrail layer that simulates and evaluates crypto transactions against policies and reputation before AI agents sign them on EVM and Solana chains.
AINew open-source tool Uncle cryptographically binds AI agent review and test artifacts to exact Git commit trees, preventing automation pipelines from trusting stale or hallucinated verification claims.
AIA new guide evaluates n8n alternatives, helping automation builders navigate trade-offs between execution-based pricing, open-source licensing limits, and code-versus-visual development environments.
AIAbliteration launched an uncensored model API and policy gateway that lets security teams automate high-risk cybersecurity workflows that mainstream AI providers typically refuse.
AIAdobe released a Model Context Protocol endpoint, allowing external AI agents like Claude and ChatGPT to directly control creative workflows across Photoshop, Illustrator, and Premiere.
AIDevelopers can remotely manage, edit, and monitor self-hosted AI agents during travel by pairing an Android gaming handheld with Tailscale, SSH, and a portable screen.