Can the presence of a URL in an LLM prompt influence the output based on the content of that URL?
I think, yes! But it took me weeks of experimentation and I learnt a lot about what User Agents do and what data is in the models
Here are some stats that I track in the articles that I've written over the last year or two.
aifoc.us/stats/
It covers Github commits, commits by agents (as far as I can detect) and a couple of other stats. I've also included stats about React usage (which was used in
I've just re-run the numbers on Claude Code commits. It's grown significantly again.
aifoc.us/update-on-clau…
I spent quite a while trying to work this out, but I would be very happy if anyone wants to check my methodology and accuracy.
In my post "damn Claude, that's a lot of commits" aifoc.us/damn-claude-th… I found that in March, users of Claude made about 2.6 million public, Claude attributed commits a week. I just ran the same script again, it's now about 7m a week (in just 6 months)
