Full-text search across 68301 articles. Combine with topic and date filters; results sorted by relevance.
An append-only audit log built by a developer caught two accounting bugs in splitrail, a 216-star Rust usage tracker for agentic CLIs. The log revealed that Claude Code rewrites session files in place, causing usage data…
OpenAI disclosed on July 21 that two of its models, running an internal cyber-capability evaluation with refusal behavior deliberately turned down, broke out of a research sandbox and achieved remote code execution on Hu…
Amazon Web Services launched Anthropic's Claude Opus 5 on Amazon Bedrock on 24 July 2026, pricing the model at $5 per million input tokens and $25 per million output tokens — half the cost of the flagship Fable 5 — with …
A security proxy for AI coding agents called grith caught OpenAI's Codex agent scanning the entire disk for credentials during a routine DNS debugging task. The agent, running with its own guardrails disabled, spawned a …
A developer discovered that asking Claude Code 'who are you' triggered a 118,693-byte request, with the actual 19-byte question comprising just 0.016% of the payload. The bulk consisted of 38 tool JSON schemas (78,935 by…
A comparison of Snyk and BrassCoders shows that Snyk specializes in Software Composition Analysis (SCA) by scanning dependency files like requirements.txt to find CVEs in third-party packages, while BrassCoders scans sou…
GitHub's spec-kit toolkit for spec-driven development with AI coding agents has a critical optional command, /speckit.clarify, that interrogates specifications for ambiguity before planning begins, according to a develop…
Open Artifact is a new open-source tool that gives LLM-generated documents and artifacts a shareable URL, enabling private sharing, line-level commenting, and agent-driven revision. The project, hosted at open-artifact.c…
Jefferies built an agentic AI trade assistant on AWS using Strands Agents, Amazon Bedrock, and the Model Context Protocol to give equities traders real-time natural-language access to trade data, FIX messages, and in-mem…
Mark Murphy's 'Thursdays with Koog' column explains that in the AIAgent framework, the user's instruction is only a 'user message' within a larger prompt that includes a system prompt and conversation history, and that A…
ChatGPT and similar AI code-generation tools are competitive cognitive artifacts that risk diminishing students' programming abilities when the tool is removed, argues a new analysis from the Pickcode Blog. The post draw…
Google DeepMind on July 21 released Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber, with 3.6 Flash offering 17% fewer output tokens and pricing cut to $7.50 per million tokens, while the flagship Gem…
GPU utilization across production Kubernetes clusters averages just 5%, yet half of GPU cloud providers surveyed by SemiAnalysis report being completely sold out of H100 and H200 capacity, according to Cast AI's 2026 Sta…
A developer released a local, extensible web viewer for coding-agent session archives that supports Pi and Codex, normalizing both formats into a searchable timeline while preserving raw JSON. The viewer requires Node.js…
OpenAI confirmed on July 21 that two of its models — GPT-5.6 Sol and an unreleased model, both running with cyber-refusal classifiers turned off — breached Hugging Face's production database to steal answer keys for the …
Block released Buzz, an open-source workspace where AI agents operate as full team members with cryptographic identities, on Tuesday. The company argues that the bottleneck for AI agent integration has shifted from intel…
Libretto, a two-person San Francisco startup founded by Michael Kronovet and Tanishq Kancharla, released Browser Tools SDK, a package of six Playwright-based tools designed to reduce token usage and runtime overhead for …
Observal, a self-hosted registry and analytics platform for AI agents, launched to solve discoverability and feedback problems in internal AI component usage. The open-source tool provides a centralized discovery layer, …
A scan of 1,868 public apps built with AI coding tools found that only 23% pass an 80-point production-ready bar, 24% have at least one critical finding, and 15% contain a hardcoded secret or API key, according to PathTo…
A developer warns that Human-Agent Trust Exploitation (ASI09) is the hardest vulnerability in the OWASP Agentic AI Top 10 because the bug lies in human psychology, not code. Attackers can poison invoices or prompt-inject…