Don’t Outsource Your Thinking

teltam.github.io · 2,801 words · saved by 4 readers

As 2026 rolls on I wanted to write down some thoughts on using Claude Code. I’ve been using these tools for quite a while now and like the first chat models there are similar patterns emerging in how to use these tools. I also hear similar threads from folks who use code generation models in their daily work. A lot of these ideas are still working themselves out in terms of practice, and I wanted to write some of them down before the next wave of model improvements. Every month provides some updates on how to poke the frontier a bit. Between base models and instruct models, the sft/rlhf detour, the reasoning arc and now rl-vr, model improvements month over month are a given. In early 2023, from a user pov, the key idea was that foundational models will eventually absorb everything around it. Perhaps, maybe not product and UX surfaces. I think this is still true, but building one level above is not a fools errand as people make it out to be. The middle man will eventually be cut out, th

Don’t Outsource Your Thinking As 2026 rolls on I wanted to write down some thoughts on using Claude Code. I’ve been using these tools for quite a while now and like the first chat models there are similar patterns emerging in how to use these tools. I also hear similar threads from folks who use code generation models in their daily work. A lot of these ideas are still working themselves out in terms of practice, and I wanted to write some of them down before the next wave of model improvements. Every month provides some updates on how to poke the frontier a bit. Between base models and instru

Don’t Outsource Your Thinking

saved by

related reading