5 Comments

Sir! Wow. This the best source of information that I can use. Sorry the words fail. Your analysis and conclusions are prophetic. 👍🤪👍👍👍 Out

Expand full comment

While o1 may not be great at most of things as in the post (I did not explore much anything else with o1 yet), but I would vouch o1 for its coding capability. It is outstanding... To add perspective, a hybrid approach for coding works 10x better than just one platform. e.g. I use claude > perplexity > o1 > then again claude to perfect my code. It works great... perplexity is in loop for research and it does a good job.

Expand full comment

Interesting. I am just starting to use o1 for coding.

I like the idea of mixing and matching tools. Will have to try. You can also run your query in o1, get a pass of code, then put it in VSCode/continue to use other tools, etc.

Expand full comment

As it stands right now, with o1 limitations, what would be the best stable choice of llm for system 2 programming?

Expand full comment

o1 is literally the only game in town for system 2 thinking. It's your go-to for coding and math and complex queries.

Simpler queries, general knowledge and help with writing are best done on other models, like Claude 3.5 Sonnet.

Expand full comment