# The Human Learning Rate
*The job is not to learn a thing. It's to become the kind of person who can learn anything, fast.*
By [Nye's Digital Lab](https://paragraph.com/@nyewarburton.eth) · 2026-10-18
learning, rate, ai, creativity, process
---
[Share](https://paragraph.com/@nyewarburton.eth/jq0VQW8QXB5GMkZrcU8m)
_Nye's Digital Lab is a weekly scribble on creativity in an age of rapid change._
* * *
**Essay # 101**
For twenty years, the jobs that I worked were in **a loop.**
As an animator...
Pose the character. Scrub the timeline. Fix the pose. Scrub again. A few hundred times a day, for a shot that lasts four seconds on screen.
The best animators I worked with were not the ones with the best first pose. They were the ones who **scrubbed** the most. They found out they were **wrong faster** than everyone else, so they were **right sooner.**
I didn't have a word for that until years later, when I started training AI models and found it sitting at the top of the settings file.
* * *
A **Variational Autoencoder, Source:** [**https://www.baeldung.com/**](https://www.baeldung.com/)
### The Dial
When you train an AI, specifically a neural network, there is one number you set before almost anything else. It's called the **learning rate**. It decides how big a step the model takes every time it finds out it was wrong.
Turn it too high and the model overshoots. It lurches past the answer, overcorrects, and never settles down. Think of the friend who changes their whole personality after one bad date.
Turn it too low and it crawls. A million tiny steps, still nowhere. Think of the friend who has been "about to start" that _hilarious podcast_ since freshman year.
Here's my thought.
> **People have a learning rate too!**
Your learning rate is how fast you turn **experience** into **ability.** Not how smart you are. Not what you know. How quickly the distance closes between _I don't know_ and _I shipped it._
Two students take the same class, make the same mistakes, get the same notes back. One of them is a different person a month later. The other is the same person with more notes.
We measure seat time, credentials and tuition. Those are not results. **Nobody writes the learning rate on the transcript.**
* * *
_Small consumer drones used in Ukraine. Source:_ [**https://aerospaceamerica.aiaa.org**](https://aerospaceamerica.aiaa.org)
### Rate Beats Depth
If you want to see what a high learning rate looks like at full scale, look at Ukraine.
Analysts watching the drone war track something called the adaptation cycle: how long it takes for one side to field a new weapon and the other side to counter it. For most of military history, that cycle ran in _years._
> Ukraine is doing it inside a long weekend!
Its cheap interceptors were knocking down drones that cost many times more. No vendor sold them that. It came from absorbing an enormous number of failures and fixing every one of them, in public, quickly, in front of people whose lives depended on it.
> The weapon was never the drone.
>
> The weapon was the learning system behind it.
My guess, and _I'll call it a guess,_ is that an organization that learns three times faster than its rivals doesn't do three times better. **It wins outright,** because the gap compounds and the slower side never gets a single round where it's ahead.
* * *
### The Bottleneck Isn't the Machine
The machines got fast. Every tool I open answers before I've finished asking.
> So why doesn't anyone feel smarter?
Because the machine's loop closed but ours, as humans, didn't. We still learn on the old schedule. Try the thing. Wait for the critique. Wait for the grade. Wait for the quarter to end.
A student who gets feedback in an hour and a student who gets it in two weeks are not learning at the same rate, even in the same class. The second one is running with the dial turned nearly to zero.
Here's the good news. An AI's learning rate is set by somebody else.
* * *
**Yours isn't!**
* * *
My electronics knowledge tops out around **"I think the smoke is bad."** There are thousands of engineers with a thirty-year head start in this space, and I'm not going to out-study them at fifty. That's a race I lose by showing up.
So I'm not running it. I'm looking for the spot nobody is standing on.
There are thousands of PhDs working on how robots balance. There are maybe a few hundred people on earth working on how a robot should _move so that a human being doesn't hate it._
> Which line do you think I'm getting in?
The skill underneath this isn't technical. It's being able to say, in one clear sentence, **what you actually want.** I learned this from bossing around AI agents. A vague instruction gets you confident garbage. A precise one gets you something useful. _Then I noticed I was doing the same thing to myself._
Most people aren't slow. They're aimed at nothing in particular, at high speed.
* * *
_Generative Improvisation # 45,_ [_Leonardo.ai_](http://Leonardo.ai) _/ Flow State_
### Nobody Learns Fast Alone
The uncomfortable math of trying new things is that most attempts **fail.** In fact, I often expect to fail. A community that runs a hundred experiments gets ninety-nine flops and a few things that work, and the few pay for everything many times over.
The failures aren't waste. They're the price of the search. They're paid by individuals, and the payoff lands on everybody.
So here's where I'd start.
Pick one thing you want to be able to do. Try it today, **badly.** Embrace how bad you are at it. Show it to someone who will tell you the truth. Fix it tomorrow. Learn and get better at it. And when you can actually do it, teach the next person.
You don't control how fast the machines learn.
**You control the length of your loop.** Jump in.
Make it Happen.
* * *
Thanks for reading. I do this _every week._ If you vibe to the ideas I express, consider **subscribing** or **sharing** with friends.
[Subscribe](https://paragraph.com/@nyewarburton.eth/subscribe)
* * *
Colophon
--------
This essay is the synthesis of a collection of verbal improvisations that were recorded and then refined with generative models. Additional revisions were done with human writing.
**Nye Warburton** is an artist and educator from Savannah, GA.
For more information: [nyewarburton.com](http://nyewarburton.com)
* * *
Links
-----
_Related essays in the Lab._
* [The Sprint Battlefield](https://paragraph.com/@nyewarburton.eth/the-sprint-battlefield) — Ukraine doesn't have better weapons. It's that they learn faster.
* [Should Robots Look Like Us](https://paragraph.com/@nyewarburton.eth/should-robots-look-like-us) — maybe we should leave the uncanny valley where it is
[Share](https://paragraph.com/@nyewarburton.eth/jq0VQW8QXB5GMkZrcU8m)
---
*Originally published on [Nye's Digital Lab](https://paragraph.com/@nyewarburton.eth/the-human-learning-rate)*