Santiago Palomino · NotesEngineering notes from my projects, plus research, security and AI economics.https://santipvz.dev/B300s are starting to look like cloud instances, not mythical hardwarehttps://santipvz.dev/notes/b300s-are-getting-easier-to-rent/https://santipvz.dev/notes/b300s-are-getting-easier-to-rent/B300 access is spreading across GPU clouds, with public spot and preemptible prices. That is a real shift, even if it doesn't mean unlimited capacity on demand.Wed, 04 Nov 2026 00:00:00 GMTBenchmaxxing is a real incentive, but a high score isn't proof of cheatinghttps://santipvz.dev/notes/benchmaxxing-evidence/https://santipvz.dev/notes/benchmaxxing-evidence/There is solid evidence that training aimed at a benchmark task can move scores and rankings. That is different from proving a particular lab secretly trained on the test set.Wed, 04 Nov 2026 00:00:00 GMTOctober was a rough month to be responsible for securityhttps://santipvz.dev/notes/black-october-cybersecurity/https://santipvz.dev/notes/black-october-cybersecurity/PHP, LiteLLM, certificate infrastructure and Binance all had a bad October. The incidents were different, but the same old weak links kept showing up.Wed, 04 Nov 2026 00:00:00 GMTWhat if an LLM could read a neural network's weights?https://santipvz.dev/notes/weight-space-programming/https://santipvz.dev/notes/weight-space-programming/A set of experiments asks an LLM to inspect, edit and even write neural-network weights directly. The results are wild, and still very early.Wed, 04 Nov 2026 00:00:00 GMTConnecting a team's autonomous stack to a simulatorhttps://santipvz.dev/notes/simulator-bridges/https://santipvz.dev/notes/simulator-bridges/My work connecting Auria to CarMaker and EUFS Sim 2, with NATS at the integration boundary and Rerun to inspect a run.Mon, 02 Nov 2026 00:00:00 GMTRead hyperscaler capex like a model metrichttps://santipvz.dev/notes/hyperscaler-capex/https://santipvz.dev/notes/hyperscaler-capex/Big Tech will spend well over $700 billion on infrastructure this year. The headline number depends on accounting choices, so I read it the way I'd read a sensitivity figure, operating point first.Fri, 30 Oct 2026 00:00:00 GMTThree numbers for the AI market, and the fine print on eachhttps://santipvz.dev/notes/ai-market-revenue/https://santipvz.dev/notes/ai-market-revenue/Venture money, lab revenue and the price of intelligence all hit records this year. Each one has an accounting footnote that changes how you should read it.Tue, 27 Oct 2026 00:00:00 GMTWhen a prompt becomes a shellhttps://santipvz.dev/notes/prompt-injection-shell/https://santipvz.dev/notes/prompt-injection-shell/Two Semantic Kernel bugs this year turned prompt injection into code execution, and web pages are now hiding instructions for AI agents. The lesson is old: the model is not a security boundary.Tue, 20 Oct 2026 00:00:00 GMTAI found twelve OpenSSL bugs the same month curl gave up on AI bug reportshttps://santipvz.dev/notes/ai-bug-hunting/https://santipvz.dev/notes/ai-bug-hunting/In January 2026 an AI system found every vulnerability in an OpenSSL security release, and curl closed its bug bounty because of AI-generated noise. Both stories are true, and they are the same story.Tue, 13 Oct 2026 00:00:00 GMTTokens keep getting cheaper and AI bills keep going uphttps://santipvz.dev/notes/cheap-tokens-bigger-bills/https://santipvz.dev/notes/cheap-tokens-bigger-bills/A curiosity of AI economics. The price of a fixed level of intelligence falls about 13× a year, yet teams spend more every quarter. Both are true.Tue, 06 Oct 2026 00:00:00 GMTLooking inside the model is a breakthrough, with an asteriskhttps://santipvz.dev/notes/looking-inside-models/https://santipvz.dev/notes/looking-inside-models/MIT Technology Review put mechanistic interpretability on its 2026 list of breakthrough technologies. What it can do today, and why researchers still argue about how far it goes.Tue, 29 Sep 2026 00:00:00 GMTCoding agents you can run on your own machinehttps://santipvz.dev/notes/local-coding-agents/https://santipvz.dev/notes/local-coding-agents/A short field guide to open-source coding agents that work with local models, what each is good at, and the one limitation they all share.Tue, 22 Sep 2026 00:00:00 GMT