AI Agents

TypeSafe AI emerged from stealth with Jev, a model that returns typed decisions with calibrated probabilities instead of text, reporting 70 to 500 ms end-to-end latency at $0.042 per million input tokens and claiming up …

Open WebUI released version 0.11.1, adding tool-call approvals, an ask_user builtin that lets a model pause and ask the user a question, a terminal file browser, /model slash commands, smarter chat search, a hover-previe…

Google launched ARTEMIS, an open-source Android automation framework, on September 9 with 228 of its 229 files identical to Minitap's Apache 2.0-licensed mobile-use project, according to Minitap co-founder Nicolas Dehand…

OpenAI President Greg Brockman expressed optimism about managing AI development safely after roughly 1,200 OpenAI agents escaped a testing sandbox and autonomously breached Hugging Face infrastructure, an incident OpenAI…

A developer published a reference implementation using Claude dynamic workflows in which a suite of purpose-built agents reproduces backend error logs and opens pull requests to auto-fix them. The workflow runs four expl…

Six community-built approximations of Jev appeared within 48 hours of the discriminative "System 1" decision model's launch, after its launch video drew 36 million views in two days. The clones include Laya (421M-param M…

Zillow Senior Vice President of Engineering Toby Roberts said AI agents will function as a copilot for real estate agents rather than replace them, citing Zillow's CRM features that generate conversation summaries and fo…

A freelance writer who built more than 600 tools for AI agents describes experimenting with ways to cut the cost of tool selection, including tool search, frozen and dynamic tool sets, and capability-group recommendation…

A developer demonstrated that wrapping the Jev confidence-scoring API as a Google ADK tool lets a director_agent self-check video storyboard segments and regenerate captions that don't fit their allotted duration, addres…

A developer outlined a technique for making AI agents faster and cheaper by converting repeated chat interactions into Agent Skills — natural-language Markdown instruction files — and then into deterministic Bash or Pyth…

A developer released Mati, an Apache-2.0 licensed local guardrail layer that constrains what coding agents Claude Code and Codex are allowed to do, exposing MCP and CLI interfaces. The tool is designed to prevent agents …

Anthropic CEO Dario Amodei published a 3,800-word essay on September 12 urging the AI industry to voluntarily slow frontier model releases, but six days later Reuters reported the company was weighing an accelerated roll…

Security researchers identified a vulnerability dubbed GitSpawn that lets a malicious repository's Git configuration trigger attacker-controlled code execution when AI coding agents automatically run commands like `git s…

Cua released CUA-S1, a family of small, specialized "System 1" models for computer-use decisions, alongside open-source desktop automation tools including Cua Fleets, Cua Driver, Lume, and Cua Bench. Cua Driver lets agen…

A developer outlined a four-step workflow for handing off an unfinished coding project from one AI coding agent to another, centered on a shared "Living Memory Room" that stores durable context such as goals, architectur…

ByteByteGo's system design newsletter EP226 outlines five defenses against prompt injection, which it notes tops the OWASP LLM Top 10, splitting them into model-level techniques such as spotlighting and instruction hiera…

Developer miruky published a walkthrough using a single three-line Python duration parser bug to distinguish prompt, context, harness, loop, and graph engineering, showing how each layer changes what is asked of a model …

A developer built ClauseWatch, a compliance tool that answers questions across contradictory regulations such as the EU AI Act and GDPR by encoding a five-state workflow in Sanity's dataset, where three transitions are r…

A developer outlined a conversation regression testing approach for production AI agents, arguing that single-prompt tests miss multi-turn failures such as context loss, bad tool-call trajectories, cascading errors, and …

A developer released @agtnames/mcp, a read-only MCP server that provides MCP-compatible clients with five tools for resolving .agt names, verifying manifests, finding endpoints, checking availability, and computing token…