Posts

Security research and AI hype usually exist in separate worlds. Security practitioners are paid to be skeptical; AI vendors are incentivized to oversell. Project Glasswing — Cloudflare’s months-long experiment running security-focused LLMs against their own live infrastructure — is one of the most honest bridges between those worlds published to date. Written by Grant Bourzikas, Cloudflare’s Chief Security Officer, the post is a 9-minute read that cuts through the usual AI-in-security narratives and gets into what actually happened when state-of-the-art models got pointed at real production code. ...

Enterprise AI budgets operate on annual spreadsheet logic. Claude Code operates on token consumption logic. Uber just learned — painfully, publicly, and to the tune of an entire year’s AI budget — that these two systems are incompatible. Uber’s Chief Technology Officer Praveen Neppalli Naga confirmed to The Information this week that the company exhausted its full 2026 AI budget by April — four months into the calendar year — because of Claude Code’s runaway adoption among its engineering organization. ...

Anthropic held its second annual Code With Claude developer event on May 6, 2026, in San Francisco — and the message was unmistakable: the age of compound, self-improving AI agents has arrived. The event, which was also livestreamed with satellite stops planned for London and Tokyo, focused less on model announcements and more on the architectural shift underway in how developers build with Claude. The centrepiece? A major expansion to Claude Managed Agents — Anthropic’s cloud-hosted platform for running production-grade autonomous agents — which entered public beta in early April 2026. ...

A benchmark result from Augment Code published last Thursday is getting significant traction on Hacker News today — and it’s one of those findings that’s more interesting the more you think about it. The headline: Auggie, Augment Code’s coding agent, outperformed Claude Code on Terminal Bench 2.0 — with 33% lower token costs — while both ran on the same underlying model, Claude Opus 4.7. No model swaps. No fine-tuning. Just smarter retrieval. A 67.4% pass rate for Auggie versus 66.3% for Claude Code, at significantly lower cost. ...

The AI agent funding wave shows no signs of slowing down. Dust, the Paris- and San Francisco-based enterprise AI platform, announced today a $40M Series B round co-led by Abstract and Sequoia Capital, with participation from Snowflake Ventures and Datadog. The round takes total funding to more than $60M. Sequoia is doubling down here — they also led Dust’s $16M Series A in June 2024, a strong confidence signal in a market crowded with enterprise AI platforms. ...

There’s a quiet architectural shift happening inside the enterprise AI deployments that actually work in production — and it doesn’t involve picking the most capable LLM. It involves constraining LLMs with deterministic systems that know when to trust them and when to override them. Eric Siegel, CEO of Gooder AI and author of The AI Playbook, published a detailed analysis in Forbes today documenting this trend, with firsthand examples from Instacart, HP, Salesforce, and Twilio. The pattern they’ve converged on is what Siegel calls Hybrid AI: combining traditional rule-based, deterministic predictive AI with LLMs to compensate for what he calls the LLM’s “Achilles heel — a deadly reliability problem.” ...

OpenClaw, the open-source AI agent framework, has published a comprehensive five-point security plan in May 2026 — and refreshingly, they’re leading with honesty rather than marketing. The team explicitly refuses to promise “risk-free AI agents,” calling such guarantees sales tactics disconnected from reality. What they are promising is a layered defence in depth. Here’s what the plan covers. 1. File System Protection: fs-safe The first pillar addresses one of the most fundamental risks in any agentic system: an agent escaping its intended workspace. Through path traversal, symlink abuse, or absolute path injection, a compromised or misbehaving agent could potentially reach files well outside its working directory. ...

For the past two years, enterprise AI buying decisions have been framed as a model war: GPT-5.5 versus Claude versus Gemini. Whoever wins the benchmark gets the deal. That framing is rapidly becoming obsolete — and Anthropic knows it. New VentureBeat Pulse survey data, published this week, reveals that the real strategic fight in enterprise AI isn’t at the model layer at all. It’s at the agent orchestration and governance layer — the infrastructure that decides how agents plan, call tools, access data, run workflows, and prove to security teams that they didn’t do anything they weren’t supposed to. ...

Google’s biggest developer event of the year kicks off Tuesday, May 19, at 10 a.m. PT — and if the leaks and official schedule are any guide, it’s shaping up to be the most agentic-focused I/O in the company’s history. What’s Confirmed The Google I/O 2026 keynote will be livestreamed on YouTube from the Shoreline Amphitheatre in Mountain View. Google CEO Sundar Pichai will open proceedings, and the product slate leaking out of Mountain View in the days ahead of the event is substantial. ...

What happens when you give 40,000 employees access to AI agents and let them build whatever they want? Prosus found out. The global tech investment company published the results in May 2026: 60,000 agents deployed, across dozens of portfolio companies, without any central mandate from headquarters. The playbook they’ve written from that experience — distilled in a report called The Coming Age of AI Colleagues — is one of the most data-rich enterprise perspectives on agentic AI published so far this year. The headline finding is counterintuitive and important. ...