This was a cool listen. I think Cloud+AI is increasingly making the @levelsio -style model of a scrappy solo serial micro-entrepreneur viable, allowing one person to spin up and run a number of companies that generate income, possibly well into billion-dollar valuations.
The Most Common Lies from an LLM:
"This is the final piece of the puzzle."
"Here's the complete working code."
"I'm certain this will fix your problem."
"You won't have any dependency issues."
"I fully understand your request."
"Your context has been preserved."
"I've tested this extensively."
my file names since using cursor: current_results_price_switch_fixed_improved_nowActuallyTrulyWorkingFinallyOptimized_NoMoreEdgeCaseErrors_v2_FINAL.csv
my morning routine
1. open stripe dashboard
2. see $0
3. blame my audience
4. build a feature no one asked for
5. write 10 replies with AI
6. read Steve Jobs biography (again)
Claude starting to flex remote MCPs and adding them to 45 minute-long Claude Deep Research. nice.
theres the developers who complain that mcps dont offer anything over rest apis
then theres the companies that wake up one day and they're integrated to claude/every mcp client without needing to be asked bc all you need is one string
Last night was a weird experience with AI. I told Cursor to init a new website inside a project, expected a blank slate. Not only did it init the site, it knew what I wanted it to be (based on root project) and built out the first version.
It was exactly what I wanted. jaw droppingly good
Then I told it to wire up a few more things to the backend, and it completely messed up the site. I couldnโt even run it. I got it back into sort of a working state but one part keeps failing. Itโs impossible to fix, I donโt even know how it was working before
I didnโt commit because I didnโt expect my small request to destroy it.
Still, once we learn how to stabilize these tools itโs going to be magical (this was Cursor)
Last night was a weird experience with AI. I told Cursor to init a new website inside a project, expected a blank slate. Not only did it init the site, it knew what I wanted it to be (based on root project) and built out the first version.
It was exactly what I wanted. jaw droppingly good
Then I told it to wire up a few more things to the backend, and it completely messed up the site. I couldnโt even run it. I got it back into sort of a working state but one part keeps failing. Itโs impossible to fix, I donโt even know how it was working before
I didnโt commit because I didnโt expect my small request to destroy it.
Still, once we learn how to stabilize these tools itโs going to be magical (this was Cursor)
Just explored AI 2027โa compelling scenario forecast from the AI Futures Project. It envisions a near future where AI systems reach superhuman coding abilities by 2027, accelerating AI research and leading to artificial superintelligence (ASI) by year's end.
The scenario examines potential outcomes, from rapid technological advancement to geopolitical tensions and the challenges of AI alignment. It's a thought-provoking look at the possibilities and risks as we approach this pivotal point in AI development.