Copy problem or flow problem? Settle it before the file opens.
One comment, taken apart
“Make it more intuitive, and the tone doesn’t feel like us — can we just simplify step 2?”
Design review · 14 min in · pasted into D:bo
Step 2 — one primary action, two supporting lines. Owner: design, today.
#142 — scope held to step 2; redesign deferred to Q4 research.
Every record also teaches it how you decide — on your own or with a team.
In the product
This is what your team actually said. This is the decision.
Paste the raw comments from your last review. Nineteen seconds later: criteria, an action, a record.
Resolved · 0:19record #142
Criteria
Is “intuitive” a copy problem or a flow problem? Decide before step 2.
Action
Onboarding step 2 — cut to one primary action. Owner: design, today.
Record
Scope held to step 2. Full redesign deferred to Q4 research.
reused 3× sinceMeeting notes · Figma threads · Client email · Slack — same output, whatever you paste.
No bot in your meetingWorks from pasted textRecords survive the sprint
Personalization
It doesn’t just remember. It learns how you decide.
Everyone starts on shared defaults. As you decide, your own standards take their place.
- At 20 decisions — roughly two to three weeks — half of what D:bo cites already comes from your own calls.
- The rest stays shared. Accessibility rules hold no matter who is deciding, so this never drops to zero.
Measured across 9 weeks of internal use by a team of three. External cohort testing starts September 2026.
Decisions recordedFrom shared defaults
- 0Day one — every reference is a shared default.
- 20Half of what it cites is now yours.
- 30It settles here. Shared principles stay shared.
Why notes don’t do this
A notetaker keeps the words. A team needs the decision.
Notetakers are good at what was said. None of them tell you what to change, what to ask, or why you chose it.
Feedback · with notes
Every word is kept, including “make it more intuitive” — which nobody can act on.
Feedback · with D:bo
Problems are split from preferences, and the criteria to check first get named.
Scope · with notes
Nothing says whether this is a redesign or one screen, so the team re-decides later.
Scope · with D:bo
What changes and what waits is written down before the file is opened.
Rationale · with notes
Three days later the reasoning is gone and the debate restarts.
Rationale · with D:bo
The reason survives as a record the next decision reads.
Week two
The second time you hear it, you already decided this.
When similar feedback arrives, D:bo brings back what you settled on last time. The discussion starts from the decision, not from zero.
The second time is where it pays off. You stop re-deciding the same thing, and the arguing time goes back into the work.
Decision records in play
“Intuitive” means flow, not copy — checked against onboarding twice before.#142 · reused 3×
Premium is spacing and restraint for this client, not more ornament.#131 · reused 2×
Tone complaints are a copy problem until proven otherwise. Palette stays.#118 · reused 4×
What it decides
Four judgments it makes before you open the file.
One comment or a forty-minute review — it reads both against the same criteria.
- 01
Is this a problem, or a preference?
It separates what has to change from what someone would simply prefer.
Only the problems become work. - 02
Whose call is this, really?
It names who has to weigh in, and what it costs if the call is wrong.
You stop guessing who to ask. - 03
What has to be settled first?
It puts the one open question in front of you before the work starts.
A criterion you can act on, in one sentence. - 04
Why did we choose this?
It keeps the reasoning, not just the outcome, and reads it back next time.
The next decision starts where this one ended.
Beta
Teach it how you decide.
This round is for individual designers. You teach D:bo how you decide, then check whether it decides the same way you would.
Who we are looking for
One designer at a time — product, UX, brand, or freelance. You do not need your team’s permission to join.
What you get
- An agent trained on your judgment, not a generic model
- A read on how consistent your own criteria actually are
- Free use through the beta, and your records when it ends
What we ask
- Around ten decisions over two weeks, on real work
- Tell us where the agent decided differently than you would
- A short interview at the end, if your schedule allows
Apply as an early tester →
Opens a Google Form, about ten minutes. Applications stay open, and we reach out individually.
Questions
Before you paste anything.
What exactly is D:bo?
A design decision memory. You paste the feedback your team actually left — in a review, a Figma thread, a client email — and D:bo returns the criteria to decide by, the action to take, and a record of why you chose it.
How is this different from a to-do app?
A to-do app records tasks. D:bo records the reasoning behind them: what had to be decided first, what was deliberately deferred, and which question was still open when the work started.
What can I paste in?
Unstructured text. Meeting notes, Figma comments, messenger threads, client requests — no formatting or cleanup needed. The original wording is kept next to the decision.
Can we test it before adopting it?
Take five comments from your last design review and run them through. That is usually enough to see whether the criteria D:bo names match the ones your team would have argued to.
How does it learn my judgment?
Every decision you record becomes a reference for the next one. Early on, D:bo leans on shared defaults. As records accumulate, those defaults give way to your own standards — around 20 decisions, or two to three weeks, is where half the references have made that switch.
What happens to our data?
You own what you put in. Records are stored separately per organization, and if you ask us to delete something, we also remove what it contributed to the learned persona.
Last thing
The first cohort is small. Tell us about your team.
No bot in your meetingWorks from pasted textRecords survive the sprint