@gdb@thsottiaux
I am using 5.6 Sol Ultra for adding a complex new feature using orchestration and the amount of times it fails with dependency issues is staggering and runs around in circles.
What a disappointment.
I see everyone is super happy with "resets" and how openai communicated. But no one is asking why and how undercooked the testing was prior to release ? This is very important for future. Companies are seeing whom you reward for few bread crumbs
#Codex
https://t.co/aprz3O5o3H
1/ A new way to ship AI products early?
OpenAI now releases multiple models, multiple effort levels, and frequent Codex changes—sometimes before the experience feels fully tested or clearly explained.
The result: major user confusion.
6/ Is this becoming the new AI software-development lifecycle?
Ship early.
Let users find the bugs.
Fix rapidly.
Compensate with credits or resets.
Repeat.
Efficient product development—or testing unfinished products on paying users?
1/ A new way to ship AI products early?
OpenAI now releases multiple models, multiple effort levels, and frequent Codex changes—sometimes before the experience feels fully tested or clearly explained.
The result: major user confusion.
5/ The icing on the cake: users who complained and tested the problems may feel positive afterward because they received a couple of days of free token usage or resets.
The company gets testing. Users feel compensated.
Sorry but, Sorloth wanted individual glory.
And that is enough for a team to get of the WC. The team that plays has the least ego about individual glory wins the cup.
And we all know which team is that !!!
#WorldCup#Football
Primatologist, conservationist, animal advocate, educator, and National Geographic Explorer Jane Goodall has died at age 91. Goodall’s decades of research into the lives of wild chimpanzees transformed our understanding of these intelligent apes. https://t.co/i3qzfDIAzg
@embirico If only you can make it work on Windows, More people can join the bandwagon. But then, looks like only Mac users are the only real deal now...
Guess everyone who works on Codex cli run it on Mac. Because it does not work on Windows without explicit permissions everytime pwoershell command is run.
Good job @OpenAI .
@claudeai I never used Claude. Mostly working with painful OpenAI's Codex implementation.
The fact that you can come out and be open about the problems is enough for me to give claude a try for coding. Anyone who ever worked in devleopment would know how these kind of issues pan out.
@OpenAIDevs Codex on Windows is broken and its been 3-4 weeks. No one from @OpenAI
thinks it is a prioirty.
Can you help ?
One of the many tickets #GPT5Codex#gtpcodex
https://t.co/zl6nnERrPY
@thsottiaux@sama@sama@thsottiaux
When such a big issue for a segment of users is not addressed in 3 weeks, it gives the impression that you guys are not serious.
Just one of many tickets
https://t.co/zl6nnERrPY
@thsottiaux@sama@thsottiaux When are you going to fix codex on windows ? User needs to give permission for every powershell command explicitly currently, which makes it unusable.
The tickets on github have been open for 3-4 weeks now.
Replit Agent3: I am interested to see others feedback if they had any success to build/correct apps with less cost.
From everything I am seeing and trying, it is only going to cost the end user more, a lot more.
At least with this initial version (3/3)
Replit Agent3: Initial thoughts: good step forward. I think it does reduces testing time from what I can see/test, and seems to provide more value on complex projects. However... (1/3)
...same issues exist: self inflciting errors, long cycles to fix them, a tendency to overcomplicate simple fixes. Given limited alternatives, it remains only halfway decent option. What was once measured in user's time/money, is now measured in Agent's time and user's money (2/3)