Visualizing 10K agent trajectories: LLM-labeled treemaps expose classifier disagreements
enjalot · x · 2026-10-18
Developer enjalot shows a workflow for analyzing large batches of agent trajectories: LLMs (jev, DeepSeek) can now cheaply generate classification columns across all trajectories. The attached treemap visualizes 10K trajectories grouped by class and colored by outcome — fulfilled 7,979, partial 1,006, failed 623, noresponse 392, clarifying 249 — with a second panel colored by another model's labels to pinpoint disagreements (agents only report % agreement). He's also exploring treemap layouts, length-based sizing, smooth zoom into turns, an instant transcript viewer for thousands of messages, 3D and UMAP semantic layouts.
More from coding & agent
- The ML Engineer #409: Netflix on Knowledge Graphs for Agents, Meta's Ads Foundation Models — AxSaucedo · 2026-10-18
- Why 'taste' is the white whale of AI coding: LLMs can read design systems but can't extrapolate them — threepointone · 2026-10-18
- Users complain Claude Code compaction forgets tons of important context — Angaisb_ · 2026-10-18
- chatllm.cpp adds support for Intern-Decision's parallel diffusion-style decisions — foldl-li · 2026-10-18
- IBM hands chip design to an open AI agent factory using distributed inference — emmanuelvivier · 2026-10-18
- Manus launches 2.0 and Cue personal-agent app: each agent gets email, phone number and wallet — emmanuelvivier · 2026-10-18