Using Opus 4.6 on claude code in terminal on an old iMac and in some ways it still feels like more than enough for so much.
The eyepatch doesn’t make it hard because it blocks 50% of your vision, it’s because it kills the depth perception..
I wanted some kind of voice talk in T3 code to orchestrate/talk through threads remotely like I used to do with codex.
Realized I can just give codex the T3 mcp, and now I can orchestrate all my T3 code (including Claude) via codex voice and it’s magical. Ty @theo & team.
This has also figuring stuff out in t3 code settings even easier.
It was annoying going back and forth between the in app thread I was asking for help in and the settings panel. Now I just ask via voice what the T3 thread is saying/send messages to it and get updates via voice.
The even crazier next step has been using @keleftheriou’s Iris to have codex voice on my Apple Watch to do this, in case I’m also doing things on my phone and can’t/don’t want codex voice running on it.
Absolutely insane times for software we are living in.
Does anyone know/have a good answer to this?
Isn’t a large part of a dot’s work done in codex threads on cloud or local machine, and don’t those draw on usage?
Genuinely curious to understand, as I have many new codex threads of my dot doing things to maintain pages etc (which is nice but I didn’t explicitly ask for).
Fully included in the Pro plan, doesn't draw on your weekly usage.
Readers added context they thought people might want to know
As of 22 October 2026, OpenAI’s dots are unavailable to ChatGPT Pro users in the EU. OpenAI excludes the European Economic Area (which includes all EU countries), the UK and Switzerland from Pro access.
[Source: OpenAI documentation](learn.chatgpt.com/docs/dots)
Hey Claude Duet figure out all the sites i shouldn't visit because of this php thing and block them automatically on all my devices so i don't go to them by accident thanks
The fact that I am noticing things like this and they bother me means I am probably spending too much time online.
I am really trying to use 6.1 Sol as a daily driver for a lot of things when using OpenAI models, but it is constantly missing things/getting things wrong, to the point that I am wondering if I should just go back to 5.6/selective Astra usage.
So I was looking into implementing this and found that with a big tool call mid-task and/or compaction itself, you probably always end up going over 100K. Not a big deal, but obviously would prefer not going over.
So was looking to set it at 90K, but found that /autocompact only accepts values >100K. And while you can override in json, then it applies to all models...
Any chance you'd consider tweaking so /autocompact can happen at values below 100K @lydiahallie?
This is also really good for non-code work!
Thanks for sharing @trq212
I first tried out T3 Code in March, and it wasn’t quite for me. It was clearly built for devs, and lacked things I needed.
Have tried it out again this week (nightly), and even as a simple knowledge worker am loving it!
Can’t recommend it enough. Nice work @theo and team.
There’s a favorites section in the general ChatGPT Space, and for the life of me I cannot figure out how to favorite a specific page/doc.
Sometimes I think we are being a bit too mean to Google about Gemini, but I tried to use it today to find something in a @Theo video, and now I don't think we're being mean enough.
Just found out you can’t edit Spaces from the ChatGPT mobile app… which kills an incredible amount of this feature’s usefulness.
I am trying my best to figure out and integrate these new features, but there are papercuts everywhere..
Interesting, but like many things with dots/threads dots starts, I wish there was a bit more info as to what's going on under the hood.
When not running locally, does it have local file context like agents.md? Does dot copy that over to cloud when local isn't available?
