New Medium: the thinking above the project
Every fix became the next container.
Chat threads, Notion, a converter I built, the IDE, a planning structure inside the repo. Each one solved the last leak, then trapped the thinking again. The pattern was the insight: the thinking needed a home of its own.
A layer above the work.
Planning moved into a hub above the code, with five layers that each answer one question, and one agent that works across all of them. The project's own files stay the source of truth, so nothing is trapped in the tool.
desktop app screens
tools agents use to work the plan
I audited my own product.
Reading the project as it really was showed two products sharing one story. I refocused on the one with outside demand, set a quality bar before tuning anything, and paused the work that was burying it.
Designed for how AI fails.
No priority appears without evidence behind it. Any finding can be challenged, and corrections stay visible. Agent actions become plain-language approval cards. And people bring their own AI provider, because real companies have real constraints.
Playbooks, proven by onboarding.
A playbook is a proven path you switch on: a plan, documents, and guidance for your agents. When onboarding started turning into complicated defaults, I made it a playbook instead. Switch it on to go from idea to first epic, or skip it for a clean slate.
Making it felt, and making it hold.
I designed and built a 15-screen scroll story that mirrors the product. When mobile broke it, I moved every layout value into data and proved the refactor changed nothing before recomposing for phones.
story screens, designed and built
differing pixels across 30 transition frames
“AI products now compete with a raw chat thread in a capable person's hands. Structure has to earn its place.”
See the decisions behind it.
I'm happy to walk through the nine decisions behind New Medium, each with the move behind it.