The prime signal newsletter is up & running ! Don't miss a beat if you are a founder, a leader or a student of your craft !
Subscribe 👇and let me know your thoughts writing.aref.vc
You're both half right..
Tobi describes the input side. Large language models (LLMs) lower the cost of starting, so someone who spawns many parallel threads gets more of them off the ground. and that's good for ADHD because the ignition barrier is ineffective anymore.
You
Tobi and I had an argument:
- He claims LLMs advantage people with ADHD, who start many things, LLMs work, people hop between them
- I claim they advantage super-disciplined folks, LLMs are tremendous entropy-generators, controlling which is the name of the game
Who is right?
In chemical kinetics, a reaction runs no faster than its slowest step. Speeding up every other step does nothing.
Factories got electric motors around 1890. Productivity did not jump until roughly 1920. Firms had to redesign the whole factory around the motor, not just swap
Economists, is it possible for everyone, I do mean everyone, in the knowledge work economy to produce 2-3x their current output using AI, but the aggregate GDP growth rate remains the basically same?
There are two types of growth:
Demand capture
Demand creation
Let’s unpack this.
Demand capture is when users search for something they already want, e.g. “video editor.”
To capture that demand, you need to appear in the search results.
Then there’s demand creation. You do