🛠️ Career Strategy

The Competence Quadrants

Exceptional talent doesn’t arise by pure dumb luck. People develop it by choosing the right work, taking intelligent risks, and accumulating a track record of impact that speaks for itself. 

By knowing what patterns to look for, asking the right questions, and filtering for behaviors over backgrounds, you can easily recognize it. 

And AI can help you. Just plug the below prompt into your AI tool of choice (e.g. ChatGPT, Claude, Gemini) along with an uploaded PDF of your (or a candidate’s) resume:

If you want the prompt, please sign up for the newsletter :) 

🤖 AI Tactic

Solving a Million-Step Task with AI

In a recent post, I wrote about how today’s AI models are terrible at following complex, interlocking rules. They’re dazzling with words, useless with logic. Fluency without fidelity.

In another post, I shared a talk I gave called “The Source of Strategy”, which partly focused on atomization of processes.

So it was very flattering to read a new research paper that solved a million step problem (the Tower of Hanoi puzzle) via AI, using substantially the approach I described.

It’s called Solving a Million-Step LLM Task With Zero Errors. That’s not a typo. One million steps, zero hallucinations. I often face hallucinations after 10+ steps!

How is this result even possible?!

They decomposed the work into atomic steps. Each mini-task was handled by a small agent whose only job was to do that single thing exactly right.

Then came the magic: voting. The researchers created a panel of voting agents to decide the next move. The AI looked at all the votes, then determined what to do.

You and I are used to seeing LLMs fall apart when the task gets long, layered, or rule-heavy. And they still will, if you treat them like a solo genius. But this paper makes the case that with the right system design, these “vibe-based” models can actually start acting like rule-following accountants. Not because they got smarter, but because we gave them deterministic micro-tasks.

Which makes me think: the next wave of useful AI won’t come from smarter models, but from better orchestration and structure.

If you try this approach (or remix it in interesting ways), reply back and let me know.

💼 Job Board

The $300,000 Whale

Here are the 3 most interesting remote job openings I’ve seen this week:

If you want the jobs, please sign up for the newsletter :) 

That’s it for this week.

-Rahul from The DesAI Digest

P.S. Reply back to this email with a business challenge you’re facing! I’d love to help.

P.P.S. if you liked this, forward it to a friend. And if you hated it, forward it to an enemy.