TypeSafe's 2-stage extraction cascade: big-model quality at a fraction of the cost
TheMoonMidas · x · 2026-09-22
TypeSafe AI documents its SDE cascade for structured data extraction: a cheap small model extracts first, per-field yes/no verifiers flag suspicious fields, and only then escalate to an expensive reasoning model.
- Models & pricing (per 1M tokens, in/out): rung 0 gpt-5.4-mini at $0.75/$4.50; rung 1 gpt-5.5 at $5.00/$30 (7x mini); verifier TypeSafe jev-1.12 at $0.042 with free output.
- The verifier returns P(something is wrong) per field; escalations fire only on verifier signals.
- Notable choice: text-mode extraction only — no structured outputs, tool calls, or JSON mode, since schema-following mistakes aren't the failure mode they expect, and constrained decoding doesn't fix the root cause.
- Includes an end-to-end example and a cost/quality tradeoff analysis across 100 prompts.
More from coding & agent
- Deel launches Akai, an agent platform that learns workflows from a single demo — kimmonismus · 2026-09-22
- Uncle Bob Demos uml-viewer: His New Way to Review Agent-Built Projects — blaizedsouza · 2026-09-22
- Progressive disclosure for agents: gradually revealing info as trust builds between agents — iamrobotbear · 2026-09-22
- Qwen 3.8 27b fine-tune cuts verbose output by up to 40% with little quality loss — julianharris · 2026-09-22
- Codex desktop app won't show which command the agent ran, user complains — amplifiedamp · 2026-09-22
- It Was Never About Coding: A Veteran Engineer's Take on the Agent Era — blaizedsouza · 2026-09-22