How to Refuse Suggestions

When refusal becomes the main job

When you polish a product like Fecit alongside AI, an interesting pattern emerges. Your main job is no longer “accepting suggestions” but “filtering them out.”

AI proposes a lot. New features, screen layouts, copy, even the big-picture direction of “what if the app became more like this?” If you take them all, the app converges on some average you’ve seen a hundred times before.

So the main labor of building a product becomes refusal. And the grounds for refusing can’t be invented on the spot. You need philosophy, user observation, and your own principles stocked up in advance to refuse.

What I actually refused today

“Let’s make overdue events stand out in red so ‘past deadline’ is more visible.” — AI’s initial suggestion. Standard to-do app behavior. I refused.

A schedule is a schedule. It doesn’t need a pass/fail judgment attached.

Fecit’s calendar is a record that “something happened at that time,” not a checklist that gets graded as done or not done. A frame where a red badge pops up once a deadline passes, telling you “you failed,” didn’t match the calendar’s philosophy in Fecit. The grounds for refusal became a declaration of what the product is.

“Let’s make a ‘Today’ tab that gathers only today’s items.” — AI proposed this, noting that the “decide what to focus on now” surface felt weak. I refused.

I’m not making a Today tab.

The reason is simple. In Fecit, “what to do right now” is handled by a single pinned Focus slot at the top. If I add another place to stack multiple things as “today’s to-dos,” it blurs the backbone of Fecit, which is to pick just one.

“Let’s ship a lighter version first and upgrade later if it works.” — I’ve refused this many times.

There is no half-built middle version. You either do it or you don’t.

I’ve watched too many “temporary” things become permanent. The moment you say “I’ll do it properly later” is often actually the final decision.

“Let’s slightly adjust the speed of this interaction.” — I refused, but AI quietly left it applied anyway (…). I noticed later and rolled it back.

The grounds for refusal become the design

If I bundle these four refusals in retrospect, each one revealed a sliver of Fecit’s direction.

• A schedule is not something to be evaluated.
• The mechanism that reduces “what should I do now?” anxiety should be singular.
• Most “temporary” builds stop being temporary.
• Things that share a screen should move as one mass.

These sentences don’t come out if you try to write them sitting at a desk. They surface naturally as you answer “why doesn’t this fit Fecit?” every time AI pitches something. Refusal builds the product’s language.

What AI is good at

So paradoxically, AI is most useful as a thing that throws out a high volume of stuff to disagree with. Options I wouldn’t even consider on my own all spill out in dialogue, and as I refuse them one by one, what this product is not becomes clearer.

Accepted suggestions become features. Refused suggestions become philosophy.