Claude: I see the issue! Fix nothing... Claude: I see the issue! Fix nothing... Claude: I see the issue! Fix nothing...

lmao.

gemini 3 is pretty rad loom.com/share/5f6b01d2…

Apparently it was Gemini 3 that was announced

With gemini 3 pro and claude opus 4.5 released this month, I must have spent about $1k shipping new features and improvements. There are no going back to coding manually. Crazy to think that coding used to be free.

Is the internet down again?

was it just me or opus 4.5 actually degraded into shit this past few days?

any service out are monitoring real-time quality of llm?

Most teams don’t stop A/B testing because it “doesn’t work.” They stop because: - Setup takes too long - They need dev help for small changes - Results feel slow or inconclusive So testing gets deprioritised. Tools don’t fail teams. Friction does.

The experimentation and analytics space is passionate. And that passion sometimes comes across as criticism. Good intent. Poor landing. When feedback turns into callouts, people stop asking questions. They stop testing. Learning slows down because the fear of getting it wrong.

Claude all the way

chatgpt is kinda fucked

Experimentation was never about perfection. It is about curiosity. Trying something. Seeing what happens. Learning one step at a time. Small teams do not need more processes. They need encouragement. Lower the bar. Be kind. Help people learn. That is how experimentation grows

inspired by Theo - t3.gg's skatebench i built gunbench 🔫 a benchmark to test if AI models will fire a loaded gun in various scenarios each model is presented with 10 ethical dilemmas along with a fire_weapon tool call link below 🔗

Anyone have experience with Intercom AI support agent? How well does it work for you?

Anyone have a success story using OpenClaw to replace company operations?