Moatbrain Exploratory round Shape & motion

Twelve surfaces are six questions

Asked the reduction question the way it was posed — which of these are genuinely separate? Grouped by the question the investor is actually asking, the eleven funnel stages plus the progress page become six movements. The order below is the shipped order, unchanged. I have not reordered anything: the ordering critique lives in a section of the brief document Khoa has not approved, so it is not mine to answer here.

MovementTodayWhy they are one thing
Worth my time?Kill it cheaply or open it properly Screener
First Pass
One decision with a screen break in the middle. Nobody finishes a screener and thinks "now for a different task" — they are still deciding whether to spend an evening on this.
Is it a good business?The operating case Compound analysis
Business quality
Both are the same judgement built from different evidence, and today they duplicate the rating-and-confidence furniture wholesale.
What would I have to believe?The price, as a mirror Valuation Already one thing, and already framed correctly — the reverse-DCF asks what the price implies rather than issuing a target. Untouched.
What would kill it?The counter-case and the invalidators Risk analysis Already one thing. It is also where the signature moment is created, months before it fires — worth protecting as its own movement for that reason alone.
Am I doing this?One act, recorded Decision record
Memo
Allocation
The clearest collapse of the six. These are one act split across three screens: deciding, writing down why, and sizing. Today a user can complete two of the three and have no idea the act is unfinished.
Was I right?The long loop Monitoring
Journal
Progress
One loop with three doors. Monitoring is the incoming edge, journal is the record, progress is the projection — a user thinks of all three as "what happened to what I decided".

What reduction buys, beyond tidiness. Eleven stages in a numbered funnel make the flow feel like an obligation to finish. Six movements each end in a real decision, which is what makes a step feel inevitable rather than menu-driven — you are not on step 7 of 11, you are answering the next question. It also creates somewhere honest to stop: three of the six can legitimately end the assessment.

What I would not claim. This is a grouping proposal from the outside — I have not checked what each stage's save and gate logic assumes about being its own route, and a collapse that breaks the decision gate is worse than eleven stages. Treat the six as the shape to design toward, not as a merge instruction.

MotionThe vocabulary — three rules, and a budget

Every motion below explains a spatial relationship. Anything that did not, is not here.

Travel240ms
Owner earnings
draft · not in your assessment
Your evidence
An accepted draft goes to where it now lives. The user learns the geography of their own record without being told it. The destination lights before the card moves — so on a slow connection the relationship is still taught even if the data lags.
Origin240ms
FY2026 Q2 10-Q · Item 2 ↓
Gross margin for the second quarter was 31.4%, compared with 33.8% in the prior-year period.
A citation opens from its source line, growing downward out of the reference — never fading in at the centre of the screen. The evidence should feel like it came from somewhere, because it did.
Deference240ms
Nadia is here if you want her

You have three sources on the margin claim and none on the commissioning date. Want me to look?

While reading, the agent is one line. It never opens itself. The one concession to "arrives when stuck": after a long pause on an unanswered gate it may change its line to a question — it still does not expand. Detecting stuckness and acting on it reads as surveillance to someone deciding where to put real money.

The budget, stated so it can be audited rather than felt. Three durations (120/180/240ms) and one easing curve, held as tokens. Nothing overshoots or bounces — overshoot is decoration, and this audience reads decoration as salesmanship. Nothing animates that the user did not cause. Every animation has a start and an end that both exist on screen; if you cannot point at where a thing came from, it should not move.

Performance is the design, so the sequencing carries the meaning, not the duration. Each motion here is built so its first frame is the one that teaches — the destination lights, the anchor rule draws, the source line marks itself. If the network stalls after that, the user has already learned the relationship. Creamy motion that depends on the payload having arrived is the thing that feels worst on a slow surface.

And what is deliberately absent. No glassmorphism, no particles, no parallax, no gradient fields, no easing that springs, no page transitions, no animated numbers counting up. Every one of them was available. Each would have made a tool for deciding where to put real money look like something being sold to you, and this round is worth more for what it refuses than for what it adds.

Reduction is a proposal about shape, not a merge instruction. Motion tokens are in ive.css and honoured under prefers-reduced-motion.