Product Designer, Builder.
Chicago
Joined October 2009
- Tweets8.7K
- Following1.7K
- Followers1.8K
- Likes14.3K
Claude’s GitHub achievement of pair programmer extraordinaire 4 times, and Pull Shark twice.
And you think you’re under-recognized for your work.
> Surprisingly the CLI has grown 2X in usage since
Bad news @thsottiaux - that's not us using it. It's us telling Claude to use it while we wait and hope for a new model.
So 40,000 people are on the Pro 500 chatGPT plan?
There were reports of usage draining faster than anticipated for some users over the weekend. We've identified approximately 40,000 accounts that had an inconsistency on which usage limits where the actual usage that was applied did not correspond to their actual selected plan. We are reaching out individually to all these users and have corrected the issue. If this affected you, we'll add credits to your account. Sorry we got this wrong.
Do you think the people at Slack have ever sent each other a video? And if they have, have they ever watched it? The video player is so janky. It plays to the end then sets itself back a few seconds, so if you go to play it again it starts from 0:57/1:00
I have been a huge supporter and defender of Open AI and Codex. But it's got to the point where the gap between 6.1 Sol and Opus 5.5 is so vast.
The GPT 6 model family is unusable for anything even slightly vaguely defined. Astra and Sol both create their own conditions and then burn billions of tokens unless babysat.
This has been true since Astra launched. But trying to prompt it in any way like you can Opus is unusable.
A few weeks ago I started having Opus prompt and babysit Codex for the stuff it is good at, via the CLI.
Today I've decided it's literally the only way I can use it.
@thsottiaux you should be embarrassed. I trust you are working to fix it.
I'll give you full access to every single thing I've ever done.
I'm rooting for you. But this is incredibly frustrating.
Feedback ID: 01a13279-5c3b-77f3-9f5e-b11916c30046
Hot take:
Coding is a means to an end.
Building products is not about coding. It's about understanding the problem, and making a solution.
Coding doesn't matter. Results matter.
Being at capacity at 1am Central Time. This means OpenAI are either:
1. Rush-training a new model to compete with Opus 🤞
2. They are really hair-on-fire fucked for capacity
3. Both
This is such a simple but insightful question.
I've never hit a limit in Claude I wasn't shooting at or expecting.
OpenAI models consistently surprise me. A task I think might take an hour runs overnight and burns a whole week of usage.
It creates rules for itself, then is trapped by them.
I can't trust it. So I have to get Claude to babysit.
There are weeks where you feel like you get a days worth of work done.
There are weeks where you feel like you get a months worth of work done.
This feels like a week where I got a year's worth of work done.
