typed/recipes

Jev and Laya: System One decision model use cases.

An independent directory of what people are building with Jev, Laya, and the System One models that follow. Every entry shows the claim, what the evidence supports, and whether we've reproduced it.

18directory entries4recipes in two languages2models

Three kinds of question

Glossary

Choice

Pick one option from a list you write. You get the winner and a probability for every option.

support · billing · sales · other

Score

Place the text on a scale you define, from 2 to 10 levels. The answer can land between levels.

urgency: 2 someday … 10 today

Noul

Ask a yes/no question and get one number: the probability the answer is yes.

“Does this ask for a refund?” → 0.93

Jev

TypeSafe AI, Closed weights
Price
$0.042 per million input tokens, output free
Latency
70–500 ms end to end (vendor figure)
Released
15 Sep 2026

The first model sold as a System One model. You send text and typed questions to a hosted endpoint and get back typed answers with probabilities. It writes no prose.

Laya

Convai Innovations, Apache 2.0
Price
$0 per token; you pay for your own hardware
Latency
32.8 ms single question on a T4 GPU; 7.2 ms per question in batches of 10 (vendor figures)
Released
18 Sep 2026

An open-weight model that answers the same three question types as Jev and runs on your own GPU or CPU. Three checkpoints ship: English (421M), multilingual for 100+ languages (322M), and one tuned on typed decisions (421M).

Latest in the directory

All 18 entries
This homepage sponsor slot is open. Sponsors are labeled and never change a verdict.See sponsorship

Use cases by job

All use cases

Routing and triage

Send a message, ticket, or request to the right place before anything expensive touches it.

Guardrails and safety

Check agent actions, inputs, and outputs against a policy, and hold the doubtful ones for a person.

Retrieval checks

Decide whether a passage is relevant, or whether a source actually supports the claim it's cited for.

Scoring and ranking

Put many items on the same scale fast enough to re-sort them live.

Agents and real-time

Pick the next move inside a loop that's already running: a browser agent, a game, a trading bot.

Claim vetting

Flag the tells of an overstated claim: vendor sourcing, missing evidence, unfair comparisons, false precision.

Vet a claim with five typed questions

Send a social post or article claim as state, ask who's making it, whether there's checkable evidence, whether its numbers show their method, whether comparisons are fair, and how promotional it reads. Your code turns the five answers into a verdict. This is the rubric behind every entry on this site.

ChoiceNoulScoreStarter

Route uncertain Jev decisions to review

Validate a Jev Choice answer, route eligible labels and send uncertain or invalid answers to review. Includes executable policy tests; no model benchmark is claimed.

ChoiceIntermediate

Qualify an inbound lead in one call

A form fill or voicemail transcript comes in. Ask the service type, the urgency, and whether it's in your service area, all in one call. Hot leads ring a phone; the rest go to a nurture sequence.

ChoiceScoreNoulStarter

Check whether a source supports a sentence

Before a generated answer ships, pair each cited sentence with its source passage and ask one Noul: does this passage support this sentence? Anything under your threshold gets flagged or dropped.

NoulStarter

What is a System One decision model?

Understand typed decisions, Choice, Score and Noul, with a routing example and links to the model documentation.

Jev or Laya: how to choose

Compare deployment requirements, version-specific evidence and total workflow cost before choosing Jev or Laya.

How to read a launch-week claim

The five questions we ask of every post before it goes in the directory, and the patterns that most often turn a true number into a misleading one.

Questions about Typed Recipes

What is Typed Recipes?

Typed Recipes is an independent directory of use cases, code examples, and guides for System One decision models such as Jev and Laya. Each entry separates the original claim, available evidence, and reproduction status.

What are System One decision models?

System One decision models return constrained decisions, such as a choice, score, or probability, rather than an open-ended text response. This directory explores their use in routing, guardrails, retrieval, and other workflows.

Have the listed recipes been independently tested?

No entry currently carries a Tested by us badge. A source-based verdict is separate from a reproduction test. Check each entry’s evidence and test status before relying on its claims.

Is Typed Recipes affiliated with Jev or Laya?

No. Typed Recipes is independent and is not affiliated with TypeSafe AI, Convai Innovations, or any model vendor.

Continue learning