Skip to content
JevDirectory.org
PatternsStructured-data-extraction cascade

SDE cascade

A two-stage extraction cascade: a mini model extracts, a Noul battery verifies, and a confidence gate escalates to a reasoning model.

Category
Patterns
Also known as
Structured-data-extraction cascade
Related terms
4
Directory entries
6
Docs
docs.typesafe.ai
Added
2026-09-24

Definition

The pipeline runs gpt-5.4-mini for extraction, verifies each field with one battery of Noul questions in a single call, and escalates to gpt-5.5 when the gate reads 0.7 or below. On an internal 100-prompt sweep it sat up-and-left of every single model on the cost/quality chart.

Its cautionary example is a jsonschema-valid but semantically wrong record that passed validation, showing that verification questions catch errors a schema cannot.

Tagspatternsextractioncost

From the directory

A two-stage extraction cascade: a mini model extracts, a Noul battery verifies each field in one request, and a 0.7 gate escalates to a reasoning model, sitting on the cost/quality frontier.
Cookbooks & DemosDocs#official#cookbook#extraction
Official
Architectural patterns for System One systems: speculative fan-out, confidence-gated routing, composite scoring, and intent routing.
Practices & PatternsDocs#official#patterns#architecture
Official
450GitHub stars
Text-only classifier for 261 federal tax forms that sends each PDF page's text to Jev, which returns a probability over forms and page kinds behind a confidence gate; no model is trained.
Repos & SDKs#community#typescript#classification
Extract dates by asking seven Choice questions about a date's shape and parts, then resolving them in code: five of six examples auto-accepted and missing dates flagged at 0.46 confidence.
Cookbooks & DemosDocs#official#cookbook#extraction
Official
Turn wine tasting notes into 67 numeric columns for a CatBoost regressor: five rounds of reading its own errors cut held-out RMSE from 3.09 to 1.77 across 800 unseen reviews.
Cookbooks & DemosDocs#official#cookbook#research
Official
Regex finds candidate emails, phone numbers, and amounts; Jev picks the requested span so code copies a verbatim value that cannot be invented, with no free-text extraction required.
Cookbooks & DemosDocs#official#cookbook#extraction
Official

From the community

Posts from builders shipping with Jev right now.

Follow @typesafeai

Jev repositories worth a look, in Japanese

やぁ!兄弟たち! Jevに関するGitHubの実用性と発展性がありそうなリポジトリをまとめたよ! やはり、高速判断を要するComputerUseや完全自動トレードなんかに対しての活用が多い印象だね! Jevは公式のウェイトリストも1日ほどで承認されるけど、待たなくてもVercel AI GatewayからModel: Show more

Reply

Company invoices, classified for the books

Real-time Clippy that wakes up when you struggle

I built real-time Clippy with Jev. It quietly watches how you use the product and only wakes up when it thinks you’re struggling. Hesitating? Confused? Stuck? Clippy knows. Even its reactions are controlled by Jev. 👀

Diogo Almeida
Diogo Almeida
TypeSafe AI
@CompleteSkeptic

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x

Reply

One-click invoice finder

A visual reference finder

I built a visual reference finder with Jev One single prompt → 100 images from Cosmos, NASA, and The Met my new rabbit hole for creative work 🌻  Show more

Diogo Almeida
Diogo Almeida
TypeSafe AI
@CompleteSkeptic

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x

Reply

Post scoring with SuperX

Jev + SuperX = virality solved ✅ Every post gets 61 questions in ~1s for $0.0004 🤯 > fitted on 9,481 real posts from 207 creators > picks the viral post 2 in 3 times > never rewards reply bait So: write, score, rewrite, stop when it peaks. Free, no signup. try it below ↓

Rob Hallam
Rob Hallam
@robj3d3

I spent the last 8 hours building a viral post classifier with Jev. It's now better at spotting viral posts than I am. And it's as good as Fable 5.1, but 100x faster.

Image
Reply