AI Development general thread.

Update; have just gotten one for the wife so she can make worksheets for the kids, and I can abuse it when I run out of mine.
Serious question, I'm wondering how someone not doing development would absolutely show the limits.

Disclaimer, that's my main focus so haven't really done other stuff with it.
 
Serious question, I'm wondering how someone not doing development would absolutely show the limits.

Disclaimer, that's my main focus so haven't really done other stuff with it.
I tried with gemini and it just wasn't able to work.
You can just tell codex to generate X amount of images in Y style and it will just happily go about its business and do what you want.
 
Update; have just gotten one for the wife so she can make worksheets for the kids, and I can abuse it when I run out of mine.
If you switch codex between accounts is there any penalty to your usage at all or any issues continuing where you left off?
 
If you switch codex between accounts is there any penalty to your usage at all or any issues continuing where you left off?
Doesn't seem to care. Like I can continue using the same projects and everything.

Found a nice app that switches the desktop app.

And for opencode:
 
Anyone used opencode? I had a issue a while back with codex and ollama ( running on Linux ). The interaction between harness and model seemed off.
 
Anyone feel like openai codex limits have been adjusted and limits have been reduced the last couple days specifically? I've just been experimenting this week with it and the first few days the limits were reasonable. Last couple days seem unusually limiting...
 
Anyone feel like openai codex limits have been adjusted and limits have been reduced the last couple days specifically? I've just been experimenting this week with it and the first few days the limits were reasonable. Last couple days seem unusually limiting...
There is the issue again of the whole model at capacity. Many started complaining early morning our time yesterday and lots of complaints around accounts and usage etc. Thursday seems to be their main release cadence and something seems to have crept in again.
 
local or frontier models? Hows it perform vs the codex harness?
No local models. I mostly use GPT 5.5 and 5.4, and use Chinese open weight models via open code zen for simple things.
The big thing is that it seems to be much lighter on tokens than using codex for some reason. Don't notice much of an intelligence difference though.
Big thing for the desktop app as I said earlier is it has very good access to an integrated browser, so you can tell it to fix something and it will loop and check. Which is very good for UI work.
 
Anyone feel like openai codex limits have been adjusted and limits have been reduced the last couple days specifically? I've just been experimenting this week with it and the first few days the limits were reasonable. Last couple days seem unusually limiting...
Yes, they either silenty adjusted usage limits or there is a bug. Everyone is explaining on twitter I've burnt through 58% of my weekly on my pro 20x account in 1 1/2 days. So definitely something broken.
 
Update; have just gotten one for the wife so she can make worksheets for the kids, and I can abuse it when I run out of mine.
You should register for a team account then you pay for 1 and can use another for free. Basically R399 for both.
There are promo codes floating on twitter for SA but yeah, I rotate accounts.
 
Top
Sign up to the MyBroadband newsletter
X