I think one of the crucial things that we as humans need to come to grips with in the age of AI and going forward is that knowledge is cheap, knowledge is inexpensive, but value is found in decisions. This is why, even though we have Claude code easily able to write code for us (knowledge work), you don't see a lot of startups just coming out of nowhere because of the decision fatigue.
How humans can mitigate against decision fatigue by being informed and not being overwhelmed is what defines success going forward.
This is how I’ve been using Fable 5.
1. I asked Fable 5 to plan. Planning is in 2 phases:
- Phase 1: When planning, I tell the problem that I'm trying to solve and ask it to reverse prompt me, asking for clarifying questions, and don't assume anything.
- Phase 2: I use the reverse prompt that it gives me and use it against Fable 5 again. We go through planning with Fable 5 until all the things have been resolved. I ask it to create a document that can be handed over to the less intelligent agents like Opus 4.8 and Sonnet 5.
2. I use this document with Opus 4.8 to start the implementation, but in the prompt I make sure to say to delegate tasks with Haiku and Sonnet 5.
This has been great with my token usage!
Now, enjoy ClawIDE with your Opus 3.7 for ultimate productivity! Multi task and context switch better, and access your sessions even if you’re in the bathroom!! 💩🚽
https://t.co/yM74yFjBEj
We fixed a bug where rate limits on Claude subscriptions weren't properly adjusted for long context requests in Opus 4.7.
We've reset 5-hour and weekly rate limits. Enjoy Opus 4.7!
I kept ending up with a mess of terminal tabs whenever I ran Claude Code on more than one branch at a time.
So I built a little tool to keep track of them. It's called ClawIDE: a self-hosted web UI for parallel Claude Code sessions.
Here are all it's feature: https://t.co/KSEELfHuM8
Here is how to get it: https://t.co/lubVhG1zOQ
Would love your feedback!
With AI, I feel like there is no excuse to be unprofessional anymore. You have to be professional at work, and businesses have to be professional with their customers. You can’t use the excuse “I didn’t know” anymore.
Dumb question but I see everyone raving about @karpathy’s autoresearch agent. Do I need a bloody H100 GPU just to use this? I don’t have $25,000 just lying around!!!
I keep seeing people say that the Citrini paper 2028 Global Intelligence Crisis paper is wrong because they don’t consider government intervention. Has government intention ever worked properly for us?
Government intervention is a pipe dream against AI. They will always respond too late, with the right intentions, but wrong execution to favor the highest payers of their election expenses and ignore the masses.
93% of U.S. jobs are exposed to AI automation.
That’s $4.5 trillion in salaries.
The biggest risk isn’t factory work.
It’s financial managers, developers, analysts.
This shift is faster than most people think.
Full breakdown here: https://t.co/rik5AmPzw5
@Cognizant@blocks
Policy question:
If a model provider says ‘we need controlled testing before deployment,’ and government responds by excluding them…
Are we selecting for safety or against it?
Incentives matter.