The code still works. Nobody's looking.

February 2nd, 2026. I shipped openSAM, a full Rust rewrite of the original $SAM framework, down from roughly 31,000 lines to about 5,000. Ninety-nine percent lighter, by a rough measure that’s actually true and not a marketing number. No new token. No hype cycle timed around the release. Just better code, doing the same job the original did, running on a fraction of the hardware.

Nothing happened.

what I expected versus what I got

I expected, at minimum, a bump. Not a repeat of the original run, I wasn’t naive about that, but something. A signal that the people who’d been part of the SAM story would come back and look at what the cleaned-up version could do. Instead the chart stayed flat. The GitHub stars trickled in at the same slow rate any solid open source repo gets when nobody’s currently talking about it. The code was, by every technical measure I could check, better than it had ever been. The room had already moved on to whatever was next.

That’s the specific lesson people skip when they tell the SAM story as a comeback arc. Better engineering did not buy back a single unit of attention. The attention economy doesn’t refund you for quality. It doesn’t check your commit history before deciding whether to look. It looks or it doesn’t, and once a room has emptied out, a cleaner rewrite of the same idea is not, by itself, a reason for anyone to walk back in.

the code was never the thing being priced

This connects to the thing I keep coming back to about the original run: the market priced attention, not code, on the way up. OpenSAM is the same fact proven from the other direction. Good code on dead attention does nothing. If code were the asset the market was actually valuing, a 99% lighter, cleaner, more capable rewrite should have moved something. It moved nothing, because the thing that had been priced the first time around was never really the 31,000 lines. It was the story people were telling each other in real time. Once the room stops telling that story, the code can be flawless and it changes exactly zero of the outcome.

I don’t think that means the code didn’t matter. It means code and attention are two separate assets that occasionally sit on top of each other and get confused for the same thing. You can improve one without touching the other at all.

manufactured luck, tested against the counterfactual

Here’s the part that actually matters more than the disappointment of a flat chart. The original SAM run could always be explained away as a fluke. Right timing, right platform, a market that happened to be paying attention to exactly that kind of story in September 2025. openSAM was the test of that explanation. Same builder, same underlying idea, meaningfully better execution, done without a token drop or a hype push riding shotgun. If the first run had been a gift, a pure accident of timing, this one should have failed in a completely different way: nobody notices, nothing works, the technical side falls over under real use.

That’s not what happened. The technical side held. The rewrite worked exactly as designed, on cheaper hardware, with less code to maintain, for months. What didn’t come back was the attention, and that’s a different failure than the code failing. Manufactured luck isn’t the claim that you can guarantee the room shows up twice. It’s the claim that the parts you actually control, the execution, the discipline, the willingness to rebuild from nothing, are repeatable outputs and not a one-time gift. OpenSAM is the proof of that half. Doing it twice, cleaner, without the hype, and having it hold technically even while attention stayed away, is the receipt that the skill was never the lucky part.

where it actually landed

I wound the whole project down in May, on my own terms, and kept the holder snapshots as the only artifact that any of it happened. Not because the code failed. It didn’t. Because I’d tested the thing I actually wanted to know and gotten the answer. The system that builds the tool is mine and repeatable. The system that decides whether a room is paying attention on a given month is not, and never was, and pretending otherwise is how people end up rebuilding the same dead thing four more times waiting for a room that already left.

The code still works. Nobody’s looking. Those are both true at once, and neither one is an accident.

This is what I write about every week. Subscribe if it landed.