@ryancarson Love the foundational idea behind this. It's what moves us forward. I can take human judgment (personalized) and delegate it with very specific requests like, notify me only when...
@omarsar0@karpathy My guess is it's a little over-indexed right now. I suspect they'll improve over time as to where the guardrail should actually sit.
@tszzl@jrysana There wonβt be a pause. The labs may say they are pausing to each other, but there is no way they will let the other validate that. And there is no way they trust that the other will not be advancing while saying they are pausing. Onward with RSI.
Claude Code has taught me to be much more expressive... on-hire engineers get a thesaurus+RSUs. Smooshing... Gesticulating... Discombobulating... Reticulating... Perusing... Enchanting... Wandering... Mulling... Spelunking... Deciphering... Forging... Actualizing... Moseying...
@FlorentChif@iruletheworldmo Makes sense. Last night I said to Codex, βare you a bloody idiot? Why canβt you just do what I tell you.β Lol.
My immediate next thought was this has been nerfed and 5.6 is coming out soon. Total degradation and drift.
@elvissun Well, it's definitely a clearly defined, measurable goal. Problem is you can spend $5M to make and $1M ARR saas, or in a race codex creates a saas that charges $1 for every $5 in amazon gift cards we give you. You'll hit $1M pretty fast. ;-)
@omarsar0@Stewart_of_Mars@dair_ai I have been working my prompt to generate the goal plan. I want to know before the the model starts working on the goal what choices I should be making before it starts running so it can run uninterrupted. I donβt want it making assumptions on my behalf for choices that matter.
@garrytan@garrytan I'm asking to do the same for GStack - I'm trying to understand it as a business technologist founder who doesn't have an AI Engineer on staff. And I want to use Codex not Claude - Ask Questions isn't avail so limited. Keep cooking!