cypot

Neuroscientist, start-up co-founder

[cypot @DrPotas]

@ChatGPT Work mode remains unusable as you still cannot edit earlier chats and it requires creating a new thread which gets messy.

Also Windows ChatGPT app mixes personal projects with workspace projects, making the desktop app unusable. Without these fixes, no Astra access for me!

[cypot @DrPotas]

@thsottiaux As soon as you enable editable chats in Work, I'll give Astra a try. Right now Sol is good enough and I'm locked into Chat mode until chats are editable in Work mode. Branching just creates way too much clutter for my workflow

[cypot @DrPotas]

@sama Would be great if you could edit chats in work mode, only works in chat mode atm, so i can't access Astra for my current workflows

[cypot @DrPotas]

@thsottiaux Work mode still doesn't permit editing prompts after they're sent. I can't use it like that, so I'm locked in to sol in Chat for now.

Any plans to fix Work, or add Astra to Chat? Would be nice to try Astra on something valuable.

[cypot @DrPotas]

@ns123abc @thsottiaux I gave opus and codex the same firmware task. I got 12 turns on codex, few mins per turn, before burning my plus plan credits. On the similar subscription with opus I got only 2 full turns, took about 15-20 mins each turn, then died in the 3rd turn before completion.

[cypot @DrPotas]

@emiliqn_2001 @thsottiaux Yes please! So annoying that you have to stick with cloud because you can't share the context across home and work computers

[cypot @DrPotas]

@ChatGPT @googledrive I wish I didn't have to hit shit + enter for a new line. That's not a natural way to type a prompt.

[cypot @DrPotas]

I just published The AI Didn’t Go Rogue. It Found a Way Out. https://t.co/CEWhETjKRg

[cypot @DrPotas]

@_noonyoon @sama Impressive analytical skills

[cypot @DrPotas]

@Ppkook4 @sama You can get 4o on OpenRouter.

[cypot @DrPotas]

@chrisgpt It depends on what you're doing. I've hit limits for the first time ever, 2x daily, but I think it's because of the complexity of the problem. Hard stuff that is not innate to the model needs more thinking. I'd imaging chasing goals is also gonna chew thru the tokens!

[cypot @DrPotas]

@sama I started using CC after I started running out of tokens in codex recently. Codex5.5 truely is amazing - in my mind an order of magnitude better than anything atm, but that's only today, not for ever. CC is way better at convincing everyone to use it despite the current gap.

[cypot @DrPotas]

@thsottiaux What about only releasing when it's been properly tested. When you guys release too early and things break, that kills productivity. Every time you change something it's frustrating to have to adapt to some new unpredictable behaviour. Keep it stable please.

[cypot @DrPotas]

@Swamee_Builds @sama Me too some issue, reached my limit for the first time ever. Then I switched to Claude, but it was very slow and I also exhausted my limit in no time.

[cypot @DrPotas]

@sama Sometimes I need a smarter model. Sometimes I need a faster model (though less critical). I always need a cheaper model! Really happy with 5.5, great job to you and your team!

[cypot @DrPotas]

@sama Used gpt5.5 yesterday, and I have to say, wow! No glitches, it handled many large complex coding/repo task beautifully. First time ever I didn't feel like I have to manage the context window, it kept up they whole way through without needing to refresh a chat. Really impressed!

[cypot @DrPotas]

@apocalypseRSA @sama @Enscion25 Yes everytime a new model is introduced it costs me a week or two of productivity and enduring irritating bugs.

[cypot @DrPotas]

@sama I think $100 is too high. $70 feels more reasonable, you'll get way more converting from plus then.

[cypot @DrPotas]

@Suryanshti777 Book

[Suryansh Tiwari @Suryanshti777]

This AI System Design guide teaches RAG better than most courses.

And I'm giving it away for free (Only for First 4500)

Inside:
• RAG fundamentals & chunking strategies
• Hybrid retrieval (BM25 + vector search)
• Production-level RAG architecture
• Evaluation & RAGAS metrics
• Hallucination reduction techniques
• End-to-end LLM system design

How to get it:
• Follow me (must so I can DM)
• RT + Like
• Comment "book"

I'll dm you