Anthropic’s Claude services hit a brief authentication-related snag on August 16, with an incident affecting claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork starting around 22:02 UTC. A fix was deployed roughly 30 minutes later, by approximately 22:34 UTC, and the incident was marked resolved. Claude for Government was unaffected throughout. What Happened According to Anthropic’s official status page, the incident (tracked internally as qt14v73myyy5) was authentication-related — meaning users across multiple Claude surfaces would have hit login or session-validation failures rather than, say, degraded model output quality or slow response times. The affected surface list is broad: it spans the consumer-facing claude.ai web app, the developer-facing Claude Console at platform.claude.com, the Claude API that powers third-party integrations, Claude Code (Anthropic’s terminal-based coding agent), and Claude Cowork. ...

On August 13, OpenAI began previewing “Ultrafast,” a new inference tier that runs GPT-5.6 Sol on Cerebras Wafer-Scale Engine hardware at up to 14x the speed of standard GPU-based processing — up to 750 output tokens per second, according to OpenAI’s own announcement. Cerebras’ corroborating technical blog cites comparative figures too: roughly 11x faster than Claude Fable 5 and about 5x faster than Claude Opus 4.8 Fast mode on Artificial Analysis benchmarks. Access is currently limited to a small preview group, but it’s worth planning ahead now if your agent workloads live or die on response latency. Here’s how to think about the migration. ...

OpenAI’s official ChatGPT desktop app landed on Linux in public preview on August 11, roughly a month after Anthropic shipped its own Claude desktop app for Linux. If you run an agentic coding workflow on a Linux box — a dev workstation, a homelab server, or a cloud VM you SSH into — this is the first time you can get the full ChatGPT + Codex desktop experience natively, rather than through the web or a terminal-only Codex CLI. Here’s exactly how to get it running, straight from OpenAI’s own documentation. ...

Strix, an open-source autonomous AI penetration-testing agent from usestrix, has surged past 50,000 GitHub stars amid a broader wave of interest in agentic security tooling. It’s picked up strong community momentum lately — enough that it’s been showing up prominently in trending-repo coverage — though we can’t independently confirm claims about hitting a specific #1 trending rank on any given day, since trending snapshots aren’t archived or reproducible after the fact. What we can confirm directly from the repo: it’s a real, actively maintained tool with a genuinely useful setup for security teams. Here’s how to get it running. ...

If you’ve seen headlines this week saying OpenAI’s “Daybreak” cyber models just hit “general availability” on AWS, hold that thought — that’s not quite what happened. What actually launched on August 11-12 is more interesting for security practitioners: a vetted, application-gated access program for two purpose-built model tiers, now reachable through Amazon Bedrock’s existing infrastructure. If you run a security team and want in, here’s what the process actually looks like. ...

The debate over how — or whether — to regulate frontier AI just got a pointed rebuttal from one of the industry’s most prominent voices. Anthropic CEO Dario Amodei has publicly pushed back against the argument that AI regulation inevitably concentrates power in the hands of a few companies and governments, calling it a “false choice.” What Happened Per reporting from ETEnterpriseAI, Amodei’s comments came in direct response to investor Gavin Baker, who had argued that AI’s risks should potentially be addressed by distributing capabilities widely rather than concentrating them through regulation. Baker also suggested that Amodei’s own public warnings about AI risk could contribute to regulatory pressure and restrictions on data centers — ultimately reducing the odds of AI delivering broad societal benefits. ...

OpenClaw’s pre-release train just shipped one of its more quietly important features: a first-class, built-in way to snapshot and restore your agent database. If you’ve ever worried about what happens to your OpenClaw state — sessions, memory, config — during an upgrade, a bad migration, or just a disk going sideways, this is the release to pay attention to. What Landed According to the official GitHub release notes for OpenClaw 2026.8.1-beta.2, published August 15, the new SQLite snapshot system adds: ...

If your team is building or evaluating an AI agent security tool — a prompt-injection scanner, a firewall for tool calls, an anomaly detector for agent behavior — you now have a genuinely open, reproducible way to measure it. A new benchmark corpus called Contemporary Agent Attacks just landed on GitHub, and it’s built specifically for this kind of testing. What’s in the Corpus Per the official GitHub repository, the corpus contains: ...

Most writing about “harness engineering” — the design of the loop, scaffolding, and feedback structure that lets an AI agent actually accomplish a task reliably — focuses on two well-trodden domains: coding and math theorem proving. But what do you do when your task is neither? A recent essay by a practitioner going by “datavorous” proposes a concrete framework for exactly that question, and it’s worth walking through before you start building your next harness. ...

If you’ve been running AG2 (formerly AutoGen) and haven’t touched it since before this month, there’s a good chance your upgrade path just changed shape entirely. AG2 hit its v1.0.0 milestone days ago, and the framework has now shipped v1.0.2 — a rapid pair of follow-up releases that flesh out the new protocol-driven architecture. Here’s what actually changed, confirmed directly against the official GitHub release notes. The Big Picture: What “v1.0” Actually Means Per the official AG2 v1.0.0 release notes: “AG2 v1 is here! The protocol-driven framework we started building as autogen.beta in v0.12 is now the mainline framework, and the classic framework has moved to its own repository.” ...