Entries for November 2025

@onusoz · 2025-11-26

My initial experience with Claude Opus 4.5 is that it’s much better than previous Anthropic models, but it’s still relatively unreliable and hallucinates. It feels lagging in reasoning compared to highest OpenAI and Google lineup of models

@onusoz · 2025-11-23

the real advantage of Gemini 3 Pro is speed. it delivers accuracy higher than GPT 5 and sometimes GPT 5 Pro at a much higher speed. the long tail of developers value speed over accuracy, so it looks like it will take over as the main coding model for most ppl

@onusoz · 2025-11-19

There already is one: google-github-actions/run-gemini-cli but updated last week, not sure if this supports gemini 3 pro

@onusoz · 2025-11-19

Gemini seems to be very good at debugging/reviewing/finding root causes. A GitHub action/integration in PRs would be very useful!

@onusoz · 2025-11-17

"The more a task/job is verifiable, the more amenable it is to automation in the new programming paradigm. If it is not verifiable, it has to fall out from neural net magic of generalization fingers crossed, or via weaker means like imitation."