isamu_2000

17 ms·

1. 1.▲
   by isamu_2000 1mo ago

   luna is the first model that has outdone gpt-5-mini on the pareto frontier for some of my high value, cost sensitive ai product workflows. it's both cheaper (by about 60% in real world use) and higher quality based on my test harnesses
2. 2.▲
   by isamu_2000 3mo ago

   We’re using GPT-5-mini in an enterprise data-processing workflow, and we too see that GPT-5.4 nano performs materially worse for our requirements, roughly 30% worse as measured through our test suite.
3. 3.▲
   by isamu_2000 4mo ago

   totally agree. and hearing this one-sided diatribe spoken with so much conviction makes my eyes roll to the back of my head, he just "knew" everything was all ai generated.
4. 4.▲
   by isamu_2000 5mo ago

   i suggest having codex review claude's plans. i consistently find it catches 2nd order effects of planned changes much better which can lead to a total change of direction in some cases.
5. 5.▲
   by isamu_2000 2y ago

   https://github.com/dorny/paths-filter ? looking into it. Definitely going through and pinning all my 3rd party actions to specific commits e.g. party/package@<commitsha>.