Operating across initiatives.
At a six-person pre-seed company, nothing happens one at a time. This is how I sequenced three competing priorities, fixing the product's foundation, building the feature roadmap, and supporting the story we told investors and publishers, without letting any of them stall the others.
1.0What was in flight at once, August to December 2026: the component library, thirteen scheduled roadmap projects, and the pitch push. Launched features are named; unreleased ones are not. Earlier work is listed below.
The constraint
Three priorities. One small team.
As Game Caviar's only product designer, I wasn't just responsible for craft. I was responsible for judgment about where design time went, in a company where every hour had visible opportunity cost. Three initiatives needed attention at once, all business-critical, none optional:
- 01
Fixing the product's foundation. A disconnected component system I identified early on and raised with the CEO in May 2025 as a priority to address immediately, before it became harder to untangle. 0 to 1 Design System →
- 02
Building the feature roadmap. Research-driven features to bring publishers into our pilot program. Product Strategy →
- 03
Supporting the story we told investors and publishers. High-concept work that helped show the long-term value of the platform.
All of it had to move through a small engineering team: 3 people when I joined, 4 since.
Operating principles
Five rules for sequencing the work.
Never let foundational work compete head-on with feature work.
When I found the disconnected component system, I made the case that it needed attention immediately. But we couldn't pause the roadmap to fix it, so instead of asking for dedicated time, I sequenced the work into the features: every new feature became a chance to add a properly connected, documented component. The design system advanced as the roadmap did, and never had to win a separate priority fight. See the design system →
Let real signals reprioritize the roadmap, not internal assumption.
The Full Screen Profile had been identified through research and backlogged, until a publisher told us it was a requirement before they'd commit to our pilot. That conversation, in May 2025, moved it to the top of the queue. It shipped that August, and the finished feature moved that publisher into contract review the same day. See the research →
De-risk big bets by shipping a smaller real version first.
We had a long-term vision for an internal agentic AI tool that could generate real excitement with publishers and investors. Rather than wait on a full build, I scoped Advanced Search on existing APIs: it solved a real, specific search limitation now, and laid groundwork for the larger vision. It came in at 40 engineering points, against 93+ for just the next phase, and shipped in March 2026.
Use high-concept, low-engineering-cost work to carry business conversations.
I designed the Dashboard and Financials concepts for pilot and investor conversations, produced the visual artifacts, and built a file that cycled through white-label treatments so each one could be adjusted and dropped into a pitch deck. Concepting started July 15 and ran between feature work. The concepts needed no engineering time, but they still had to be credible in front of a room: they supported more than a dozen pitches to publishers, developers, and service providers around Gamescom and a UK trip, including one to Microsoft/Xbox that led to two additional leadership meetings, without pulling a 3-person engineering team off work already in flight.
Make engineering bandwidth a design decision, not just an engineering constraint.
With 3 engineers when I joined, 4 since, and a Django-to-React migration running underneath everything, every design decision had to account for build cost up front. That's part of why the design system was built incrementally, why Advanced Search leaned on existing APIs instead of new infrastructure, and why the investor concepts stayed conceptual until they'd proven their value.
What ran in parallel
None of it was sequential.
To make the overlap concrete, here is what was in flight between March 2025 and today:
- Diagnosing and rebuilding a disconnected component library: raised with the CEO in May 2025, tracked in Linear since January 2026, and still ongoing, including a Figma-to-staging pipeline (in progress) and accessibility standards that didn't previously exist.
- Running client interviews that served as both usability testing and product discovery.
- Designing and shipping the Full Screen Profile, from re-prioritization in May 2025 to delivery that August, including new components that fed directly into the design system.
- Concepting, prototyping, and shipping Advanced Search, a phased step toward a larger AI vision, in March 2026.
- Designing through a 2026 roadmap of thirteen scheduled projects between late July and December, tracked in Linear.
- Producing the pitch concepts, visual artifacts, and white-label file for more than a dozen publisher, developer, and service-provider pitches from July 15, 2026.
They were concurrent, competing for the same small pool of engineering time, and each one's priority shifted in response to real signals, a client requirement, a resourcing constraint, a pitch opportunity, rather than a fixed roadmap set months in advance.
The throughline
It's a lot like chess.
The common thread across all three initiatives isn't a specific tool or technique. It's a consistent way of making tradeoffs under real constraint: you develop your pieces while making threats, not instead of making them.
Build foundational work into feature work rather than against it. Let real client and market signals override internal assumptions. Ship a smaller real version of an ambitious idea rather than waiting to build the whole thing. And treat engineering bandwidth as a design constraint to design around, not someone else's problem to manage separately.