One idea per issue
A deep dive covers a single concept properly instead of five badly. If it needs two issues, it gets two issues.
About
Machine learning,
Machine learning,
explained like you're smart.
Not like you already know it. AI, But Simple is a weekly deep dive that takes one idea from the field and walks it end to end, with the math kept honest and the jargon left out.
Why it exists
The field moves faster than anyone can read.
Roughly 400 papers land on arXiv every weekday. Almost none of them are written for the person who has to ship something on Monday.
Most explanations pick a side. Either they hand you an equation with no intuition, or an analogy with no substance, and you end up trusting neither.
We write the version in between: the idea in plain language, the mechanism drawn out, and the math shown once you can see why it has to be that way. Six minutes, once a week, no hype.
Ten thousand researchers, engineers, and data scientists read it, from labs and teams at MIT, OpenAI, NVIDIA, Google, Stanford, Amazon, Meta, and Microsoft.
How we write
Four rules, held to.
01
02
Intuition before notation
You should understand the shape of a thing before you meet its symbols. The math shows up once it explains something.
03
Sources, always
Every claim traces back to the paper, the repo, or the posting it came from. No secondhand summaries of summaries.
04
No hype
Nothing is revolutionary until it holds up. A result gets described at the size it actually is, benchmarks and caveats included.
10,000+
Readers
200+
Issues published
7
Live trackers
The process
How an issue gets made.
The same four steps every week, which is the only reason it arrives every week.
No. 1
Watch the whole field
PRO pulls papers, repos, model releases, job postings, and interview questions from hundreds of sources every day, and ranks what actually moved.
No. 2
Pick the one that matters
Out of everything that surfaced, one idea gets chosen: the one that changes how you'd build something, not the one with the loudest launch.
No. 3
Take it apart
Read the primary sources, rebuild the intuition from scratch, draw the mechanism, then check the math against the paper.
No. 4
Cut it to six minutes
Everything that doesn't earn its place comes out. What's left is the version worth reading.