A few days ago I pointed an adversarial agent at Photo Developer and told it to be harsh. Comb the App Store copy, the site, the docs — find where the app is writing checks the release history can't cash.
It came back with a fair one. Some of the language, it said, was a touch bold for a product first released earlier this year. The tone of a tool with a decade behind it, on a tool with a few months.
It wasn't wrong. And then, by coincidence, Rogue Amoeba published The Aggregation of Marginal Gains two days later — and handed me the frame for why the critique stings a little and misses a little at the same time.
The 1% that Brailsford found
Rogue Amoeba's post borrows from Dave Brailsford and British Cycling: don't chase the dramatic overhaul, chase a hundred small things each improved by one percent. Better dashboards. A single, calmer permissions window. Update prompts that don't interrupt you. A license field you can copy with one click. None of it is a headline. All of it compounds.
Their conclusion is the honest kind: twenty-four years of those small gains have aggregated into something that didn't win Olympic medals but did power the broadcasts of them. That's not a boast. It's arithmetic. Small deposits, a very long time, compound interest.
I read it and felt two things at once, which is usually the sign there's something worth writing down.
Method and mileage are not the same thing
Here's the distinction the two-day coincidence drew for me. Marginal gains is really two things wearing one name.
One is a method: ask "how do I make this one percent better," ship it, ask again. The other is mileage: the total those one-percents have added up to over however many years you've been at it.
The method is available on day one. The mileage is not. You can practice the discipline from your very first release and still have almost nothing accumulated, because you've only run the loop a few dozen times, not a few thousand.
That's exactly where Photo Developer sits. The method has been in place since the beginning — the whole app is a stack of small, boring, corrective steps. Weight the luma path by chroma so shadow noise stops turning into speckle. Pin a kernel's region so a sub-rect render doesn't go black at 100% zoom. Read the white-balance convention off the slider gradient instead of reverse-engineering it from the render, and stop flip-flopping the sign. None of these is a feature. Each is a one percent. I have a changelog full of them and a graveyard of the ones that didn't work.
What I don't have is the mileage. Rogue Amoeba has been compounding since 2002. I've been compounding since spring. The adversarial agent was right that some copy borrowed the confidence of the mileage before the mileage existed.
What a four-month-old app is allowed to say
The wrong fix is fake modesty. Shrinking every claim into "just a little app I made," hedging everything, apologizing for existing — that's not honesty, it's a different kind of pose. And it would be false: I am genuinely impressed by what the thing does, and I built it, so saying so plainly is more honest than performing humility about it.
Less, but better. For eighty percent of the work. Those aren't marketing softeners bolted on to seem humble. They're a precise statement of where the aggregation currently stands. A young RAW developer that tells you it handles eighty percent of your edits is not underselling — it's measuring. It's the difference between "this replaces Lightroom" and "this does the part of Lightroom you actually reach for, and does it the way I think it should be done." I'm not trying to dethrone anyone. I left Lightroom for my own reasons; the app is what came of having opinions I couldn't file as feature requests.
So: bold about the method, honest about the mileage. The confidence belongs to the process — the willingness to keep finding the next one percent — not to a maturity the calendar hasn't granted yet.
Is there such a thing as perfect?
Probably not, and it's the wrong target anyway. Perfect is a destination; one percent is a direction. The whole appeal of marginal gains is that it never asks you to arrive. It asks you to improve the thing in front of you and come back tomorrow — vedlikehold, maintenance, the unglamorous work that is most of the work.
The target I can actually aim at is smaller and more honest than perfect. Get the core to good. Then, on the images where I can see the difference, get it to better-than-good. Not everywhere at once. Where it shows.
Rogue Amoeba spent twenty-four years proving where that road goes. I'm a few months in, borrowing their map. I don't get to claim their distance. But the map is free, and the next one percent is always right there — so I'll take the method now and let the mileage arrive the only way it ever does.