2025-05-12 03:17:08
Made some notes on how Cursor works under the hood based on their security documentation - it turns out an organization's list of subprocessors offers a loose form of "view source" for their infrastructure!
https://simonwillison.net/2025/May/11/cursor-security/
2025-05-09 23:34:42
Gemini 2.5 now applies the 75% cached token discount automatically - previously you had to manually configure it
Potentially big cost savings here for applications that run prompts against the same long context, or continue existing conversations https://simonwillison.net/2025/May/9/gemini-implicit-caching/
2025-05-08 23:23:31
The lack of love for magic links is interesting
(I'm not a fan personally, especially since I'm never confident if they'll work on my laptop if I click the link on my mobile phone)
2025-05-08 13:50:43
This bug was affecting my llm-gemini plugin too: I just shipped a release fixing that https://simonwillison.net/2025/May/8/llm-gemini-0191/
If you're calculating costs against Gemini you should know that you need to add candidatesTokenCount and thoughtsTokenCount together for the total billed amount
2025-05-08 06:40:43
RT AI Engineer
👀 Speaker drop!
"The only conference your engineers NEEED to attend"
the SOTA in AI Engineering from @saranormous @simonw @gdb @anthropicai @raizamrtn @ben_kus @ArtificialAnlys @jessemhan @thedanigrant @willccbb @JustinLin610 @shresbm @kwindla @ankrgyl (and a few more to be announced....)
2025-05-07 03:17:46
A feature I would love to see from every single hosted API vendor is some kind of special case where if you prompt "what model ID are you?" it replies with a definitely-not-hallucinated stable version identifier
(If model vendors are going to start switching date-based aliases to different models now it's the least they can do!)
