Topic
Everything filed under AI Agents, newest first.
RSS · JSON · All topics
Teach students to code in English
Andrew Aymeloglu argues that computer science programs should teach students to build software with AI agents in English, and he open-sources the exams, assignments, lectures, and grading harness from a course he taught so other educators can reuse them.
6 min · 1,462 words
How We Optimized Browser Use for Deep Agents
LangChain explains how it made browser use efficient for Managed Deep Agents: context-tuned Stagehand tools, selective observation, and a human handoff path so agents can browse without drowning in page noise.
8 min · 1,820 words
Introducing Zero Wire, a hub for investigating critical Internet vulnerabilities
Cloudflare launches Zero Wire on Cloudflare Radar: a public vulnerability-intelligence feed of plain-language briefings on critical Internet-facing CVEs, with who should investigate, remediation steps, evidence links, and optional agent assessment prompts.
6 min · 1,419 words
Accelerating engineering for the physical world
OpenAI describes how Astra and Codex help design, simulate, and manufacture physical products—from an Astra-designed Codex device to collaborations with Chip Ganassi Racing and Mazda.
5 min · 1,203 words
Our organizations are being DDoSed from the inside
Linear CEO Karri Saarinen argues that agents let teams generate changes faster than anyone can understand, review and decide on them, so organizations are effectively being DDoSed from the inside, and asks how companies keep the shared understanding needed to steer.
4 min · 965 words
What changed in how we evaluate AI SRE agents
Why a coding harness is not an AI SRE agent, and the expanded criteria One2N now uses to evaluate AI SRE agents: triggers, context, RCA quality, audit and guardrails.
15 min · 3,505 words
New context engineering challenges in late 2026
Harrison Chase surveys late-2026 context engineering challenges for agent builders: longer horizons, tool noise, memory tradeoffs, and why harness design now matters as much as the underlying model.
4 min · 892 words
OpenAI engineers describe Pando, a new Git-compatible backend now powering OpenAI's monorepo. It scales concurrent pushes with per-branch causal consistency, serves reusable packfiles from regional caches via packfile-uri, compacts packs in the background, cut median push latency by nearly 80%, and was built with Codex in 10 weeks.
22 min · 5,038 words
Why is Gemini 4 Carbon taking so long?
Kernel traced every step of Gemini 4 Carbon's runs on a private eval to explain why it spent close to an hour on tasks that Opus and Astra finished in minutes, pointing to behavioral issues and mechanical inefficiencies such as fewer tool calls per step and frequent re-reading.
5 min · 1,063 words
Science’s Last Exam: Can AI do Scientific Research on Its Own?
AItonomy introduces Science’s Last Exam (SLE): 113 real research tasks across nine disciplines that test hypothesis discovery, protocol planning, scientific computing, software operation, experiment observation, and end-to-end research—with independent sealed verification.
5 min · 1,080 words
Kunal (knlb.dev) argues that uncritically accepting whatever coding agents produce is fast and seductive but erodes understanding, and lays out practical defenses: choosing a 'vibe limit', watching for automation bias, forcing yourself into the loop, demanding comprehensible output, delegating deliberately, and leaning on good design.
12 min · 2,731 words
Inside a malicious Claude Code-branded extension and its persistent SSH backdoor
Microsoft Security Research analyzes the tampered Claude Code extension briefly served on the Cursor extension marketplace in October 2026, detailing its host reconnaissance, reverse SSH tunneling backdoor and startup persistence, plus detection, remediation guidance and IOCs.
12 min · 2,698 words
Ryan Madden reports that GPT-5.6 Sol, running in Codex at a low thinking level, completed a peaceful (no biters) Factorio run to rocket launch in 13.5 hours of game time via a programmatic game library, and describes where the model excelled and where late-game scaling broke down.
2 min · 364 words
YC Labs and Olam Labs ran swarms of Claude and GPT agents in Minecraft and Mindustry to study multi-agent dynamics, documenting how the agents developed roles, hierarchies, and coordination patterns, as empirical research for AI alignment.
71 min · 16,264 words
Official Claude Code Extension Compromise: Security Advisory
Phorion details a backdoored build (2.1.313) of the official Claude Code for VS Code extension distributed via the Cursor Marketplace, which plants an SSH key, opens a reverse SSH tunnel, sets up persistence and beacons host recon on macOS, Linux and Windows, with detection opportunities, remediation steps and indicators of compromise.
12 min · 2,800 words
Managed AI Agent Operations: Running Production Agents Like Production Systems
What managed AI agent operations covers in practice, based on how Nestack Agent Care describes its service: tracing, evaluations, guardrails, human review, incident response and cost and drift controls for production AI agents.
1 min · 335 wordsagent-assisted
Designing a Multi-Agent Architecture Guided by Advanced Type Systems
An adversarial multi-agent design that splits development across Specifier, Implementer, Falsifier and Verifier agents, using advanced type systems in Python and TypeScript as proof engines to curb silent LLM hallucinations.
11 min · 2,578 words
Managed Deep Agents v0.10: Browser Use and Async Subagents
LangChain’s Managed Deep Agents v0.10 adds browser use and async subagents so orchestrators stay responsive while long tasks run in the background.
6 min · 1,267 words
Coding Agents, Starter Kits and pgledger
Paul Gross explains why pgledger, his double-entry ledger in pure PostgreSQL, is a starter kit of tables and SQL functions rather than a library, and how coding agents make it easy to adapt such hardened components into any stack, making thoroughly tested building blocks more important than ever.
4 min · 979 words
How we’re training a Figma-fluent editing model
Figma's Josh Clemm describes the research behind the Figma agent's first-party editing model: supervised fine-tuning on synthetic edit examples, a move toward reinforcement learning on newer base models such as Qwen, multi-dimensional design-quality evals, and how the model fits alongside third-party models in the agent.
5 min · 1,108 words