How one designer keeps up with three products
I'm the only designer at Peek, and in 2025 I was on three products at once. I can't be in every review, ticket and release. So I built a design system any engineer or AI session can build from, three AI skills that run my process from kickoff to release, and I started shipping front-end changes as my own pull requests.
Product
Process and systems
Role
Sole product designer
Team
With PMs and engineering
When
2025 – now
Platform
Figma, Claude Code, GitHub
6
handoffs run through the same audit
7
missing screens caught on one handoff
1st
front-end redesign merged from my own PR
The Issue
One designer, three products. Here's what held up and what didn't.
→ 01
What worked
Figma files engineers trust
What was missing
Design gets skipped when a sprint is under pressure
→ 02
What worked
A design system in Figma
What was missing
Screens drift: 3 different community filters, 370+ status chips
→ 03
What worked
QA by me before release
What was missing
Only as thorough as my week allows
Feature
Design System 2026
Write the system for whoever builds next, people or AI.
A system only in Figma didn't stop drift.
What I built. The rules, import keys for about 80 tokens and every component, layout patterns, and a guide for briefing an AI design session.
The trade-off. Every change is logged with its before state, which slows edits down but makes any change reversible.
The pivot
Result
[Confirm] An adoption signal, such as components used in code.
Audit counts, then the rules that answer them.
Feature
Three skills: kickoff, handoff, QA
Let the skills draft; keep the decisions human.
My process depended on my memory: kickoff notes, handoff checks, QA lists.
What I built. /ux-project runs a project from brief to direction, /handoff audits a file before it's Dev Ready, and /qa-release tests the build against what was promised.
The trade-off. Nothing goes into Jira without my go-ahead, and product calls stay with the PM and me.
The pivot
Result
On one feature, 7 missing screens, a targeting contradiction and a terminology mismatch were caught before build.
Before
After
One handoff before and after /handoff.
Feature
Shipping my own front end
Design around the AI constraint, then ship it as code.
Our AI-search community pages were built for AI to read, but real renters were arriving too.
What I did. Mocked up the redesign, ran the app against real communities, and raised it as my own pull request, then revised it after engineering review.
The constraint. Moving the local-area section above the listings took citations from about 43% to about 98%, so I kept it there and designed around it.
The pivot
Result
Merged on Oct 28, 2026 after engineering review. Engineers now review real code, not a picture.
The community page before and after.
In their words
“
What I learned
I'd measure time saved from the first handoff. I know the skills catch things, but I don't have a before and after for how long a handoff or a QA round takes.
Also shipped
Design System 2026 handoff kit · /ux-project, /handoff and /qa-release · AEO community page pull request
