for the context builder, you might want to test out NotebookLM; this really shines in the case of pre-structuring and organizing from lots of different but related sources of information. You could still make a follow-up interactive round with another LLM ... but NotebookLM is really a fantastic tool to ingest, structure of, and derive from multiple information sources. use the chat to create overviews, deep dives, etc.; include more sources from other websites, etc.; also save useful output to notes and complement information or merge successful takes; then use the notes to even increase the focus. It's worth a try

For a more practical approach I recommend the video below. It shows step by step how to create a masterplan and then further instructions to produce code. It's for Claude 3.5 but lot's of parts are the same with ChatGPT. kzbin.info/www/bejne/nH7EY62HftCEra8si=9_ePznrZkLL_Xxc1

Thank you for making this video. Subscribing to your channel, will keep an eye out for more.

This video is eye-opening and mind-blowing! More gems like this please!

Thank you for sharing your workflow! Congratulations for having the courage to go on youtube, in front of a camera and share what you have learnt with the world.

You can just wait O1 official, it will be agentic framework, don't waste your time to do dirty work yourself

why u record mmono

👍

It saved my ass from being kicked out of university. I had only 5 hours left before the assignement deadline. I hadn't been to any classes and had zero chances of doing the assignment. But, magically, o1 did it for me! And it passed! Now I got a job, so I am safe. (University is just to get a diploma to sort people, it is not about learning. The real learning happens by actually doing stuff at work.)

This is very useful! Thanks!

O1 is the first model where I actually saw some creativity. I was working on a problem and asked it to pick the best direction but I forgot that I've said it. So I went on to try to solve the issue and the o1 took a different path! I was about to curse it, but I realized it actually proposed a much more elegant solution so I went with it and it was damn good. Not a bad model. But working iteratively... not good. Huge HUGE outputs, and then goes back to previous replies, not very good for that. STill, I use it as it's much smarter than 4o

Yes I have notice that o1 models work diffrently than standard LMM - thanks for tips:)

Awesome thanks I use gemini lt but the pro main one but now seriously confused and need to know if I should use flash instead, is it actually better then pro, you didn't really mention the main gemini one at all specifically? Thanks

I subbed after 5 minutes. Could tell it would be good and it was.

How do you export all your context into documents?

Your videos and content are awesome, have seen/read all of it in X. Please do more of these, thanks very much for sharing your knowledge.

Sullys vids are great

Thanks for sharing! Just wanted to alert you that the first 18 seconds or so is only on the left audio channel. I didn’t hear anything since I was only using my right earbud.

Great video! Really insightful workflow, thanks for sharing. Excited to see what the future holds for these models, specially, transitioning into agentic models!

Thanks for sharing your knowledge!

Why nothing about Lama3.

Do you use any public Q&A datasets for LLM evaluation?

Great one . .

This channel is going to go places.

Awesome video! Love the practical insights on model costs. More on o1 series usage would be great!