I write about using Claude, Cursor, and Copilot to ship real production software.
No demos. No hype. Just what actually works.
Follow if you’re a dev building systems.
In the far future, AI will be so ubiquitous that humanity will develop 'dumb devices' that are more emotionally reassuring.
This will be colloquially referred to as 'Artificial Stupidity'
So I can't add 2SV to my @google email account because my org won't allow it.
I can't add it to my org because it's not set up on my Google email account.
Is this a joke?
Can someone make a scanner client where I can open the software and save a document locally in like 2-3 clicks?
I'm so over @HP 's client. Genuinely dreadful, enshittified pap.
I've been working a lot in the power industry lately and the tech side of things is REALLY unreliable.
Almost anything related to publishing data seems on a best efforts basis; some data providers fail every day.
Forget your Five Nines, we're talking <75% availability.
Making me wonder how we actually have really reliable power grids at all 🤔
New claude technique unlocked:
typing ! in front of a command from inside claude will run whatever follows as a bash command.
No more quitting claude to reauthenticate with AWS 🙌
I took the @karpathy autoresearch loop and pointed it at markets.
25 AI agents debate macro, rates, commodities, sectors, and single stocks daily. Every recommendation scored against real outcomes. Worst agent by rolling Sharpe gets its prompt rewritten by the system. Keep or revert. Same loop, prompts are the weights, Sharpe is the loss function.
Trained the agents on 18 months of market data. 378 iterations. 54 prompt modifications, 16 survived.
The system learned which agents to trust using Darwinian weights — geopolitical, commodities, and the @BillAckman quality compounder rose to the top.
The agents even figured out their own portfolio manager was the weakest link before we did!
Deployed the trained agents. +22% in 173 days. Best pick: AVGO at $152, held for +128%.
The final prompts are evolutionary products — shaped by market feedback, not human intuition. Now running live with my own capital.
https://t.co/3KSAEKfoph
Part hedge fund, part research experiment :)
@MorganColeBooks Bash safety checks. They're making it impossible to use Claude without having to answer 'Yes' 20-30 times per prompt and there seems no good way to manage them
Claude spontaneously adding a hyper-rigorous check on top of settings has meant that virtually each prompt now requires 3-5 manual permissions. Massive degradation in quality.
My AWS bill is forecast to be 30% higher. I click on the 'Generate explanation button on the console'. I get
'Your bill is forecast to be 30% higher this month'
We're still early on AI
🤣