Blog
Essays on agents, the web, and software — plus the occasional build note. Follow along byRSSor JSON Feed.
- 3 min read
Perplexity won the wrong argument
The Ninth Circuit vacated Amazon's injunction against Comet: a shopping agent is a tool, not a person under the CFAA. The strategy that won this round is the one that loses everything that comes after it.
#ai-agents#law#cfaa#web-infrastructure - 4 min read
Provenance proves the build, not the builder
npm's own trusted-publishing pipeline cryptographically attested a worm on August 4 because the compromise happened one layer above the token it was designed to protect. The same gap just showed up in agent skill marketplaces and MCP registries too.
#supply-chain#npm#agents#security - 4 min read
Cheaper tokens, richer labs
OpenAI just cut its cheapest model's price 80%. Anthropic's inference margins climbed from 38% to over 70% in the same year. Both are true at once, and the reason says more about who holds pricing power in AI than any per-token chart does.
#ai#economics#inference - 4 min read
Multi-agent systems spend more and call it architecture
Two 2026 papers show automated multi-agent LLM systems losing to a single agent at matched compute — and the exception, Anthropic's own research system, turns out to prove the same rule from the other side.
#agents#llm-systems#distributed-systems - 4 min read
The weights didn't change. The score did.
Anthropic revised a prior model's benchmark score in May 2026; OpenAI nearly tripled another's two months later — both without touching the weights, only the harness. What that says about self-run evals.
#benchmarks#evals#ai - 3 min read
Agent memory solves storage, not recall
Cloudflare and MinIO both shipped durable 'agent memory' products this year, framed as the fix for context rot. The research the fix is named after says the hard part was never storage.
#llms#memory#context-engineering - 3 min read
The remote code execution hole in MCP is a design decision
OX Security's April disclosure showed every official MCP SDK will run an unsanitized command string as a subprocess. The answer was a documentation update, because on the transport's own terms this is intended behaviour — and neutral governance has no way to overrule an argument of that shape.
#mcp#security#agents - 3 min read
Kimi K3's license got stricter the week open weight went political
The letter 25 companies signed in July 2026 argued open and closed models have converged enough that regulators should back off. Days later, the open model closest to proving that point shipped a tighter license than its predecessor.
#ai#licensing#open-source - 4 min read
The crawler rulebook finally has an enforcer
Cloudflare will start blocking mixed-use AI crawlers by default this September. The mechanism that makes it possible is cryptographic, but the power it hands out isn't going to a standards body.
#ai#crawlers#infrastructure#economics - 3 min read
The queue got longer. The shelf didn't.
App Store submissions are up 80% year over year on iOS alone, and Apple just spent a guideline update policing apps that make other apps. The bottleneck was never the building.
#app-store#distribution#software-economics - 3 min read
Valid JSON was never the hard part
Grammar-constrained decoding guarantees a language model's output will parse. Benchmarks from 2024 through 2026 keep finding that guarantee stops well short of the output being right.
#llm#structured-output#constrained-decoding - 3 min read
The bug reports got better, and that's the problem
Curl killed its bug bounty in January over AI slop. By April the slop was gone and maintainers had more work than ever — because the thing they were filtering for was never correctness.
#open-source#code-review#ai#security - 3 min read
The browser finally speaks GGUF
llama.cpp shipped a real WebGPU backend, benchmarked this May against WebLLM and Transformers.js. The in-browser model catalog just got 400x bigger — and the numbers show exactly what that gain cost.
#webgpu#on-device-ai#llama.cpp - 3 min read
The API key was always the weird part
Consumer AI subscriptions are starting to authenticate the apps you use — no developer key, no token bill. It's a real shift in who holds the meter, and it lasts exactly as long as the labs allow it.
#agents#ai#economics - 3 min read
You can't scale what you have to remember
MCP is dropping the session. That reads as a loss until you remember why every protocol that mattered ended up stateless — and what it actually costs to get there.
#mcp#agents#web-architecture - 3 min read
The control question already has an answer
The argument over whether capable AI systems could be kept in check is effectively over. It wasn't won — the careful people just stopped holding the line, and an unattended ransomware agent showed what that costs.
#agents#security#trust - 3 min read
Nobody is reading your llms.txt
97% of llms.txt files never get read, and we've run this experiment before. What separates the sidecar files that shaped the web from the ones that became decoration.
#agents#llms-txt#web-history