You self-hosted the model. Nothing leaves your servers. Or so you thought. A poisoned open-source LLM can silently exfiltrate your most sensitive data through the very tools you gave it — and there's no source code to audit.
Thinking about AI and Search
I write about generative engine optimisation, LLM visibility, and how AI systems decide what to say — and whose content to surface. If you work in SEO and are trying to make sense of the shift to AI search, this is for you.
Latest Writing
Terminal programs like hledger store everything in plain text — making them lightweight, fully yours, and perfect for AI agents. Here's why the command line is making a comeback.
When you ask an LLM to rate its confidence, does the format matter? We tested four SOTA models with decimal (0.00–1.00) and integer (0–100) confidence scores across true, dubious, and nonsense labels. Decimal format produced more conservative estimates on ambiguous inputs and dramatically better cross-model agreement. Integer format caused surprising failures — GPT-5.2 alternated between 0 and 100 on obvious nonsense. The culprit? Tokenization. The 0. prefix appears to anchor models into calibrated probability-reasoning mode that integers simply don't activate.
5 minute paper