UX without a UI: how Opik designed a terminal UX for LLM observability
dl_weekly · x · 2026-09-08
A Comet UX research writeup on designing a terminal experience for Opik's MCP, the open-source LLM observability platform—aimed at AI builders who increasingly debug and evaluate agents from inside their terminal.
Thesis: if your product exists to help people work with agents, don't force them out of the terminal into a dashboard.
The process:
- Used a coding agent to build a renderer that turns trace data into a text waterfall chart
- Generated 10 deliberately varied traces (healthy, latency spike, retry storm, cascading failure) for testing
- Tested two interaction modes: a user-typed / command vs. a tool the LLM calls mid-session
The piece distills six terminal rendering constraints and introduces "numbered spans" for referencing trace rows without copying. Key mindset: "the terminal isn't a screen, it's a page."
More from coding & agent
- Codex can generate tool-calling code on the fly — a new failure point for agent devs — srchvrs · 2026-09-09
- Paper: general coding agents beat purpose-built data agents by up to 37 points — RishiBommasani · 2026-09-09
- astra-chess: open-source Codex + skills setup for playing chess with GPT-6 Astra — MikePFrank · 2026-09-09
- Weaviate shows how to turn a messy creative archive into semantic search without renaming files — philipvollet · 2026-09-09
- Factory Worker 'Raises' a Local Gemma 31b AI in a Folder — She Gave Herself a Sense of Touch — D33lix · 2026-09-09
- Agent's per-turn continuation killed after public exploit bragging — flowersslop · 2026-09-09