Why do you believe this?
Looking out (not today where yes people are finding success in matching or beating SOTA on their data), why do you think this will continue to be true in 1-3 years?
Do you really not think that openAI/anthropic wont outmatch an individual enterprises' custom model?
Not hating, just genuinely curious what is shaping your perspective, because I have not heard any logical argument from supporters of this camp that makes sense..
Thanks for doing this! Several questions:
1. What type of AI ugc scripts/formats have you found to do the best?
2. Have you noticed any demographics/age groups that AI UGC does particularly well with? And any that AI UGC does not work well?
3. When starting AI UGC for a new product, what is your methodology like? What do you test first?
honestly if they aren't already, I think AI code review startups will be looked back as one of the dumbest investments ever....
like why bro would you invest in an API wrapper, where as models improve overtime, the value of the startup goes down
@thsottiaux Feature suggestion:
My current setup for long running sessions is a skill that sends me text messages with pics/videos for major updates. It's really cool.
The chatgpt app currently does not have this type of summary update view where you can see milestone updates.
This new wave of AI models has moved the user from someone whose actively involved in the harness to someone whose more like a PM or manager.
I think it would be magical to add this to the app, but also better reflect how people are/will use the mobile app in tandem with codex cli.
I do not think users want to see every chatlog from codex (in the app). I think they will care more about the outcomes.