Tony Hirst

Every so often I tinker with Claude.ai to refactor code. It typically makes huge claims about how many lines of code it has refactored out, but diffs rarely back that up.

Last year, I “retired”, taking the early option to leave via a mutual resignation scheme. The break was pretty much a clean one, except for a visiting position to see out getting JupyterLite and browser executed R code deployed to support a student in a secure environment (one of His Majesty’s pleasure facilities).

I admit to using genAI for coding but I’m still reluctant to engage with in pretty much any other context except a bit of OCR (particularly with not-quite-Roman alphabets such as Old and Middle English texts), but eben then with a lot of care (genOCR doesnlt like minority positions, preferring the statistically most likely next token; for a digital humanities scholar that could potentially lead to an interesting way of probing biases, but as I don’t want to contribute to the genAI world, even...