Show the Rest

Last August I published eleven UX principles, one essay each, over nine days. I have been going back through the record I put out while we were building BeMo, checking each claim against what building it actually taught. This edition is about two of those essays, published seven days apart, that disagree with each other. I did not notice at the time. Building the product is what surfaced the contradiction, and this is the edition where the miss was a promise I borrowed from 1888.

The promise I borrowed

Principle 2 was called "You Press the Button, We Do the Rest." The title is George Eastman's line, coined to sell the first Kodak: photography without the darkroom, chemistry hidden behind a shutter button. I applied it to nonprofit software with full confidence. The manifesto's one-line version read: "Our job is to hide the complexity and deliver the outcome. You describe the result. You get the result." The essay went further: "All you did was press the button. The system took it from there. That's not just smart software, it's respect for your time." And it closed by making the borrowed promise explicit: "like George Eastman, we believe technology should disappear behind the scenes, so your vision can take center stage."

Invisible automation as respect for the user's time. That was the pitch.

The promise I published a week later

Principle 9 was called "Agency Belongs to the User." Its subtitle was four words: "AI suggests. Humans decide." The manifesto's version: "AI can recommend. You decide. BeMo is a partner, not a pilot." The essay argued that nonprofit leaders "don't need tech that takes the wheel. They need tech that rides shotgun, offering directions, but never grabbing the steering wheel." And it contained the sentence that principle 2 never reckoned with: "friction isn't always bad. Sometimes, it's the space where insight happens, where decisions are made, nuance is added, and missions are protected."

Read them side by side and the contradiction is plain. One principle says the system should disappear. The other says it must stay where you can see it. One promises to do the rest. The other says the rest is where your judgment lives. I published both inside seven days and saw no tension, because in August 2025 nothing was built yet and a promise costs nothing until someone has to trust it.

Why Eastman's promise does not transfer

Eastman could hide the darkroom because a photograph verifies itself. You look at it: either it is your daughter on the porch or it is not. The output carries its own proof, so nobody ever needed to trust the darkroom.

Knowledge work does not verify itself. A shortlist of ten funders looks equally plausible whether it is grounded in your program's reality or confidently wrong. A drafted appeal reads fine whether or not it matches what your organization actually said last year. The output does not carry its own proof; the proof is in the work behind it. Hide that work and trust has nothing to stand on. "We do the rest" is a fine promise when the rest is chemistry. When the rest is judgment, it is exactly the promise an overstretched development director has learned not to believe.

There is a second failure under the first. Some of "the rest" is the job. Deciding which five funders fit your mission, which story leads the annual report, whether the board sees the risk paragraph: that is not overhead to automate away. By my own principle 9, it is the space where the mission is protected. Automating the mechanics respects someone's time. Automating the decision replaces the person, and dressing that up as convenience is the move our own positioning rules now prohibit by name.

What building taught

Every month of building made the system more capable, and the design question was never the one principle 2 assumed. It was never "can it do the rest." It increasingly could. The question that decided every design was "how does it show the rest": what did it read, what is it proposing, what evidence sits under the proposal, and where does the human decision land. The complexity we hide is mechanical, the Boolean syntax, the formatting, the re-keying. What the work rests on stays visible, and the system says so when it does not know something, because a stated gap is something a person can act on and a confident blank is not.

What changed in BeMo

The contradiction resolved in favor of principle 9, and not quietly. This spring our internal positioning canon, the written rules every piece of copy gets checked against, ratified the agency principle by quoting it back: AI can recommend, you decide. The shipped behavior rule is the same sentence wearing engineering clothes: the AI proposes, the user decides, nothing happens without consent. And a companion rule says the win belongs to the user, not the software.

For the standing background work that principle 2's promise most tempts, the rule we hold ourselves to for everything we build next is receipts or silence: either the system can show what it did and the evidence behind it, or it surfaces nothing. "Ran this morning, nothing new" is an honest report. Unverifiable helpfulness is not. The same rule reaches further than I expected: even encouragement has to cite its evidence, because praise without receipts is noise from a machine.

The scorecard, both ways. From principle 2, the outcome-first instinct held: nobody logs in wanting to build a workflow, and hiding mechanical complexity is still right. What did not survive was disappearance as a trust posture, and the button itself was the wrong metaphor: real work is not one press followed by "the rest," it is a sequence of decisions with mechanics between them. Principle 9 held entirely. It is the best-aged claim in the archive, and it earned the receipts the rest of the manifesto still owes.

In one sentence, the delta: I promised "we do the rest," and building for people whose work runs on trust taught us the honest promise is "we show the rest, and you decide."

Eastman hid the darkroom because the photograph was its own proof. We show ours because a shortlist is not.