@thsottiaux A next step could be an option that enables the agent to work and think in the background without you telling it to do so, reminding you of something, or making suggestions, etc. Something a real assistant or friend would do.
@thsottiaux Fix issues that have been present for weeks now. Half of the projects are missing from the overview and are only accessible through search. Every few hours, Codex is unable to submit messages, but only in some chats. The same applies to reloading them. Some are no longer reachabl
@DanielMiessler If there is any budget left at the end of the year, they buy random stuff to consume it, because a year later it may be less etc. With a mindset like that, you don't want or create any change.
@DanielMiessler The problem with these big companies is that they attract a lot of people who just want a leadership position but lack the skills. After a while, you have a lot of people who don't feel any responsibility for great success, nor are they able to create one.
@kimmonismus They are not. You can't compare pay-per-use with a subscription, so API pricing has to be a lot more expensive to lead people into a subscription. If a daily use of a gym costs $10, that doesn't mean a monthly contract is worth $300. That's not calculated this way.
@thsottiaux Merge them together within one chat, one project overview, one chat history. Why tear it from another? The user could differentiate via the model picker.
@argofowl It is expensive and slow. It unnecessarily inflates every task. It needs hours for something you could do in 15 minutes with 4.8. For the moment, I'm not impressed, which is sad.
@jumperz They got more expensive, but I don't think that 5.6 got better. I used it now for 24 hours and have switched all tasks back to 4.8, the hard ones to Fable. I have two Max plans; all in all, the Claude plan is the better deal.
@thsottiaux I would recommend switching back to 5.5 and using Fable for difficult problems. In theory, it's more expensive, but in reality, it doesn't take two hours for something that could be done in 10 minutes.
@bridgemindai 5.6 is expensive, slow, and s.... I switched my tasks back to 4.8. It fixed everything that 5.6 ruined in a 2-hour run, within 15 minutes. Fable for the hard stuff.
@OpenAI My first impression is that this model is extremely slow, needs hours for simple tasks, and does a dozen unnecessary things while working. That leads to a large amount of token consumption. I'm not able to detect where it is smarter than 5.5. Fable is a lot better in my opinion.
@wasiancandy Sie essen abends, abgesehen von den Kommentierenden, die alle krank sind und ihr Gewicht von Luft und Liebe halten, bis sie dann Ozempic nehmen und abnehmen, was klar zeigt, dass es doch das Essen war. In der Wissenschaft nennt sich das diet underreporting.
@xburak Vacuuming only works if you know how to build one, not! This is one side of an AI agent: do something you can't. If you are best at coding, you're possibly not the one who's going to be good at orchestrating something. Lead someone.