AI Product Frontends
Building the UI on top of a model API — the practical round these companies actually run.
9 topics
Streaming Responses in the UI
Getting tokens from a model API onto the screen as they arrive, and the event model you are actually consuming.
core25 minChat UI Architecture
The state model behind a conversation surface — messages, streaming status, branching, and where the transcript actually lives.
core25 minCancellation & Abort
Letting the user stop a generation, and making sure the request actually stops rather than just disappearing from the screen.
core20 minRendering Model Output Safely
Treating generated text as untrusted input — markdown, HTML, links, and code blocks that a model was talked into producing.
core25 minTool Calls & Agent State in the UI
Showing what an agent is doing — tool calls, results, approvals and multi-step traces — without turning the screen into a log file.
core25 minLatency & Perceived Speed for LLM UIs
Working with seconds-long responses — where the time actually goes, and the interface techniques that make the wait acceptable.
core20 minErrors, Retries & Degraded Modes
The failure modes specific to model APIs — rate limits, refusals, truncation, mid-stream drops — and what the UI should do about each.
core25 minStreaming with RSC & Server Actions
Where the model call belongs in an App Router app, and how tokens get from a server component or action to the client.
deep25 minCost, Telemetry & Feedback Capture
Measuring what a model feature costs and whether it works — token accounting, quality signals, and the privacy line.
deep20 min