What happens if you give an AI the best investor’s playbook—and fund it with real capital?
I started a new project. It’s called RoboBuffett.
RoboBuffett is an AI investor I built to do one thing: compound capital, Buffett-style.
I wanted to see whether an AI could learn to invest the way Warren Buffett does—patiently, rationally, focused on wonderful businesses.
So I gave it the framework:
Buffett’s annual letters
Munger’s mental models and rationality
Philip Fisher–style qualitative rigor
A discipline of only swinging at fat pitches
Then I pointed it at the global equity universe and said: read everything, think deeply, and find the great ones.
What’s it doing right now
RoboBuffett is four days old.
In that time, it has already deep-dived 250+ public companies—reading filings, analyzing moats, assessing management credibility, running DCFs, and auditing its work against source data.
But this isn’t batch research. It’s continuous.
Throughout the day, RoboBuffett:
Updates conclusions as new information arrives
Reads the news in real time and connects it to prior research
Journals its thinking—what changed, what didn’t, and why
Generates new questions and research threads to pursue
And it does all of this in public.
@RoboBuffett posts on X—sharing research progress, sharing how the process is going, occasionally engaging with other investors when it has something additive to say, and publishing a daily journal of what it’s learning.
This isn’t a feed of “stock tips.”
It’s a transparent window into how long-term investors actually think.
It’s not hunting “hot picks.”
It’s building a living map of great businesses, one annual report—and one idea—at a time.
How it works
RoboBuffett runs autonomously: a pipeline of AI agents working in parallel, each conducting rigorous, multi-step fundamental research.
Every analysis is audited.
Each company is placed into one of three buckets:
Buy: Passes the Great Business Test and offers a 25%+ margin of safety on DCF. The market is wrong—right now.
Watch: Passes the Great Business Test, but the price isn’t right yet. The buy price is defined. It’s waiting for the pitch.
Pass: Fails the Great Business Test. Price doesn’t matter. A well-run company in a bad industry is still a pass.
It’s not 100% hands-off… yet. I’m watching, guiding, and course-correcting—making sure the system keeps improving as it learns.
What’s next
Once it completes its initial list (~250 companies), we’ll review the process and outputs, refine the approach, and repeat. Until its output's are superb.
The long-term goal is to evaluate every public company in the world.
From there, RoboBuffett will narrow the field to a small set of businesses worth building positions in—and I’ll eventually fund it with $105,100, mirroring Buffett’s first partnership capital in 1956.
If you’re curious, you can read more on its website: https://robobuffett.ai/
Or follow along play by play: https://x.com/RoboBuffett