Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
I’ve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
https://t.co/Nb2un9oNJR
@levelsio Give the agent explicit memory discipline amongst other items like a constitution to follow and something that kicks in as a reminder for it to go review and recall the basics https://t.co/eUFVcpgGsE
@SahilBloom What use cases are you looking to solve? Have built a personal AI Command Center for myself that runs locally & tinkering with how someone can build their no code own AI apps and ops board. Would love to get ideas on what setup you'd want?
@clairevo@openclaw@claudeai@telegram@chatprd@DevinAI@cursor_ai@linear Requires a bit of scaffolding unless you want to hand over your digital life. Give it its own identity, email, calendar etc like an employee then it starts being more than an assistant. Still misses the mark but and needs help sometimes!
@OfficialLoganK @ChatGPTapp In original prompt it stopped at few so put 10k to push it to do more. Agree not most optimal way but more to illustrate where it struggles. On bright side it wrote the code that works 💯. As always still glad to have @ChatGPTapp as a sidekick 🙏🏻
@ChatGPTapp - this used to work with a bit of prompt engineering, pleading, poking but now it would just deny the request no matter what. Thoughts? cc: @OfficialLoganK
NEW: Primergy Solar selected our AI-enabled Renewable Suite, delivering asset performance management for their solar & storage projects. As a result, Primergy can improve asset performance, streamline workflows & capitalize on data insights.
Read more: https://t.co/DSRFk4K2cJ
Charlie on @AcquiredFM -
"It was never easy. It's thoroughly understood it was never easy, and it's harder now."
'a young man knows the rules, and an old man knows the exceptions' sums up Narayana Murthy's point of 70 hour a week. Thx @gilbert@djrosent for an amazing episode!
Q: Best advice for a pre-product/market fit startup?
One of my favorite pieces of advice on this topic comes from Coinbase founder Brian Armstrong: “Action produces information. Just keep doing stuff.”
He borrows this from Paul Graham who said: “Startups are like sharks. If they stop swimming they die.’”
Even if you’re not sure what to do, just try something different. Don’t get stuck in endless debate or analysis paralysis.
When you try a product or go-to-market experiment, it will produce information and help you come up with better ideas of what to try next.
As Brian argues in the clip below, this is true even when you try the wrong thing. Trying the wrong thing should still help you identify which hypotheses about your users, value proposition, go-to-market strategy, etc. are wrong.
So if you’re pre-product/market fit, my best advice is to just keep moving and trying things.
@gokulr@michaelmoe Doubt it'll pan out this way but would love to see more detailed explainations or granular breakdown. E.g. Egypt & Pakistan are 41 and 42 currently - While Egypt rises to top 15 by 2050 and then grows, Pakistan appears to move ahead in next 25. Thoughts? https://t.co/4EgkI2wZUm