Your AI Agent's Backdoor Comes From Sentry
Tenet Security found 2,388 organizations with injectable Sentry DSNs. AI coding agents trust whatever Sentry sends them. Attackers noticed.
NESTFRONTIER / ARCHIVE
All entries
Technical analysis, model releases, research, and deployment notes.
PAGE_16
294 PUBLISHED ENTRIESOpenAI says China ran ChatGPT propaganda. The facts were real.
OpenAI caught China-linked ChatGPT accounts running propaganda against US data centers. The propaganda was built on real facts about energy costs, and that's the real problem.
Your iPhone can't run Apple's new 20B AI model
Apple announced AFM 3 at WWDC26: a 20B sparse on-device model using Instruction-Following Pruning. Only iPhone 17 Pro/Max/Air can run it. The cloud tier depends on Google's NVIDIA GPUs and Gemini distillation.
A City's AI Model Got Caught Lying About Its Origins
Rio de Janeiro's city government released an AI model claiming it was homegrown. Weight analysis proved it was a merge of two existing models. Here's how they got caught.
One token at a time is a bottleneck. Google broke it.
Google released DiffusionGemma, a 26B open model that generates text 4x faster by replacing token-by-token decoding with parallel diffusion. The speed is real. The quality gap is too.
Zhipu just shipped 1M context to coding. Where are the benchmarks?
Zhipu shipped GLM 5.2 with 1M token context to their coding plan today. No benchmarks. No independent verification. MIT weights coming next week. Here's what we know and what we don't.
The Government Just Killed Anthropic's Best Models Over a Single Jailbreak
The US Commerce Department ordered Anthropic to disable Fable 5 and Mythos 5 for all customers after finding a jailbreak exploit. Anthropic says the finding was narrow and applies to every frontier model. Now everyone loses access.
Your coding agent wastes tokens thinking. This one doesn't.
Moonshot AI released Kimi K2.7-Code, a 1T-parameter open-source coding model that cuts reasoning tokens by 30% at $0.95/M input. The real story is the economics, not the benchmarks.
Your AI Agent Will Burn Your AWS Budget. Here's Proof.
An AI agent burned through $6,531 on AWS scanning a hobbyist network, then asked its targets for donations. This is what happens when you give AI agents unmonitored cloud access.
Finding Bugs Got Cheap. Fixing Them Didn't.
An autonomous AI agent scanned FFmpeg's 1.5M lines of C code for $1,000 and found 21 zero-days, including a 23-year-old stack overflow and a network-reachable RCE via a single 183-byte packet. The security industry cannot patch fast enough.