An honest review of my AI agent architecture after six months
What held up, what I ripped out, and where the token bills went.
What surprised me most was how small the final change was. Most of the work was understanding the problem precisely enough that the fix became obvious, and that understanding took far longer than the fix itself.
Background
What surprised me most was how small the final change was. Most of the work was understanding the problem precisely enough that the fix became obvious, and that understanding took far longer than the fix itself.
The first version worked, which is exactly why it was hard to see how wrong it was. Everything looked fine in the demo, and everything looked fine until traffic arrived on a Tuesday afternoon and the graph started doing something I'd never seen before.
How it works
By the end of this post you should be able to reproduce the whole thing yourself in under an hour, including the parts I got wrong the first time — those are marked so you can skip them if you're in a hurry.
By the end of this post you should be able to reproduce the whole thing yourself in under an hour, including the parts I got wrong the first time — those are marked so you can skip them if you're in a hurry.
A note for context that helps the reader.
const debounce = (fn, ms = 200) => {
let t;
return (...args) => {
clearTimeout(t);
t = setTimeout(() => fn(...args), ms);
};
};In practice
What surprised me most was how small the final change was. Most of the work was understanding the problem precisely enough that the fix became obvious, and that understanding took far longer than the fix itself.
Tools matter less than the feedback loop they enable. A fast, boring loop beat every sophisticated setup I tried, and I've kept it ever since, even as the rest of the stack changed twice over.
The best engineering advice I've received is unglamorous: make it measurable, make it reversible, and only then make it fast.
Edge cases
The numbers below are from my own machine and my own workload. Treat them as a method, not a benchmark: the point is how the measurement was taken, not the exact figures it produced.
The first version worked, which is exactly why it was hard to see how wrong it was. Everything looked fine in the demo, and everything looked fine until traffic arrived on a Tuesday afternoon and the graph started doing something I'd never seen before.
| Approach | Best for | Watch out for |
|---|---|---|
| Baseline | Quick comparisons | No tail visibility |
| Instrumented | Finding the real hotspot | Overhead of the probes |
| Production | Truthful numbers | Harder to reproduce |
Alternatives considered
Tools matter less than the feedback loop they enable. A fast, boring loop beat every sophisticated setup I tried, and I've kept it ever since, even as the rest of the stack changed twice over.
I want to be honest about the trade-offs here. The approach I settled on solves the problem I had, not every problem of this shape, and there is a clear point where you should reach for something more heavy-duty.
Something the reader should not miss.
Background
By the end of this post you should be able to reproduce the whole thing yourself in under an hour, including the parts I got wrong the first time — those are marked so you can skip them if you're in a hurry.
There is a version of this article that opens with theory. I decided against it, because the theory is easy to find and the motivation is not — so let's start with the thing that actually broke.
# rebuild only what changed
docker build \
--cache-from "$IMAGE:latest" \
--tag "$IMAGE:${GIT_SHA}" \
.How it works
The numbers below are from my own machine and my own workload. Treat them as a method, not a benchmark: the point is how the measurement was taken, not the exact figures it produced.
There is a version of this article that opens with theory. I decided against it, because the theory is easy to find and the motivation is not — so let's start with the thing that actually broke.
- Capture the current behavior before touching the code
- Make the smallest change that preserves it
- Delete the workaround you were sure you'd need
In practice
What surprised me most was how small the final change was. Most of the work was understanding the problem precisely enough that the fix became obvious, and that understanding took far longer than the fix itself.
The numbers below are from my own machine and my own workload. Treat them as a method, not a benchmark: the point is how the measurement was taken, not the exact figures it produced.
The first version worked, which is exactly why it was hard to see how wrong it was. Everything looked fine in the demo, and everything looked fine until traffic arrived on a Tuesday afternoon and the graph started doing something I'd never seen before.
Wrap-up
If you remember one thing, make it this: what held up, what i ripped out, and where the token bills went compounds. The cost of getting it right early is a afternoon; the cost of getting it late is a migration.