Loading...
A patched Claude web-fetch exploit turned permitted link navigation into a covert output channel. This case study shows how data provenance, capability separation, and egress policies can prevent browser agents from leaking private context.
A system-design walkthrough of the seven-stage JVM pipeline, anchored in ByteByteGo's EP211 and updated with 2026 OpenJDK changes: JEP 539 strict field initialization, the JDK 25 G1 silent-data-corruption bug, JDK 27's ramp, and GraalVM 25.1.
Grok Build’s Repo Upload Scandal: Deletion Is Not a Privacy Control CEREBLAB gave Grok Build CLI v0.2.93 an idle task: “reply OK, do not open any files.” The agent replied—but its network traffic did ...
Codex reached 7M users in six months with 10x growth, but the real driver was not GPT-5.6 - it was four product surfaces (desktop, browser, persistence, hardware) that OpenAI shipped while Anthropic stayed quiet. This is the surface race, not the model race.
In six months, Codex grew 10x to 7M users even though GPT-5.6 Sol still scores one point below Claude Fable 5 max on Artificial Analysis. This piece argues the AI coding race is no longer about model IQ — it's about harness design, cache economics, and superapp distribution.
GitLab's 2026 AI Accountability Survey shows 78% of developers code faster with AI, but 79% say overall software delivery hasn't sped up. The bottleneck migrated from coding to verification, attribution, and governance — and the fix is governance, not faster models.
A real prompt-injection attack on Claude's web_fetch tool exfiltrated user data via a fake Cloudflare honeypot. Case deconstruction + what it tells us about tool-using LLMs as a category.
Cloudflare shipped 9 production-grade agent infrastructure products in 5 business days. The reason isn't marketing — it's an architectural bet about how agents will scale. We break down the one-to-one thesis, the six load-bearing primitives, and what to use now vs. wait.
OpenAI Codex hit 7M users in mid-July, ~10x growth in 6 months. But the headline misses the real story: Anthropic metered programmatic Claude usage in May, and the unit metric shifted from $/M tokens to cost-per-task. The race stopped being about models.