Loading
Each one gets its own route and its own shape — whatever the subject actually needs. Some are finished; several are registered in full and written a module at a time, and the cards say which.
Designing physical things and getting them made.
The judgment that separates senior from mid.
Senior Frontend Trainer
Judgment under constraints, not syntax recall. Retrieval practice, spaced review, and transfer tasks pointed at your real codebase.
Spaced practice · nothing to read start to finish
Accessibility to citation grade
The version that settles an argument rather than starting one: criteria by number, what a scanner provably cannot catch, and the unsolved problem of streaming text meeting a screen reader.
Complete · all 12 writtenabout 175 min
Front-end performance under streaming load
INP first, because that is the metric a token stream actually fails while scoring fine on everything else. Measurement, React under high-frequency updates, and payload discipline as a guest in someone else’s bundle.
Complete · all 12 writtenabout 174 min
Planning like a senior engineer
Framing, design docs and non-goals. One invented enterprise SSO project, planned end to end the way a senior engineer would.
Complete · all 12 writtenabout 168 min
Prototyping in hours, not sprints
Velocity treated as a tooling problem, because that is mostly what it is. A starter kit holding decisions already made, a mock agent backend, and timed runs against unfamiliar products.
Complete · all 11 writtenabout 157 min
Agent graphs and interrupts
Reading a stateful agent graph the way the person rendering it has to. Nodes, state, checkpoints and interrupts, and the UI surface each one demands.
Complete · all 16 writtenabout 233 min
Streaming interfaces
Everything past request, wait, render: transports that survive a corporate proxy, partial answers, cancellation that reaches the server, and a run that outlives the tab.
Complete · all 16 writtenabout 224 min
Making an agent’s work legible
Trace rendering as an information-design problem. The span model first, then reasoning, citations, confidence and cost — disclosed at the altitude the reader is actually at.
Complete · all 15 writtenabout 213 min
Grafting onto someone else’s app
Tenancy rather than ownership. Delivering a component into a host you cannot rewrite — the cascade, Shadow DOM, micro-frontends, CSP, and inheriting somebody else’s brand instead of fighting it.
Complete · all 16 writtenabout 241 min
The agentic UX canon
HAX, PAIR, Apple’s HIG and NN/g read accurately enough to disagree with — starting from the fact that most of the canon was written for systems that predict, not systems that act.
Complete · all 11 writtenabout 152 min
Knowing whether the agent UI works
Correction rate, acceptance, abandonment, escalation — defined precisely enough for someone else to compute, and read honestly, since every one of them has a flattering second reading.
Complete · all 11 writtenabout 161 min
Enough security to raise the right flag
The rendered surface is attack surface. Injection reaching the page, PII in a trace view, residency, and the telemetry choice that lands inside an audit — enough to notice and escalate, not to own.
Complete · all 11 writtenabout 157 min
How the money works — for the things you build yourself, and for the companies that buy software.
UX diligence as an instrument
Building the scorecard before anybody asks for one. UX maturity, front-end debt, agentic readiness and graftability, scored red/amber/green with a remediation estimate in weeks.
Complete · all 12 writtenabout 176 min
Speaking value creation
Translating a design decision for people who think in EBITDA and hold periods — and refusing to overclaim, which is the fastest way a design function loses the room for good.
Complete · all 12 writtenabout 166 min