Does 0% type-error mean it never returns the wrong answer?

No, and that is the whole point of the video. The 0% figure is guaranteed by construction because the output shape is fixed by the schema, not an accuracy measurement, and a typed response can still be factually wrong.

Did Reindent benchmark the speed and price claims itself?

No, it says so. The launch figures are treated as vendor claims throughout, and the description frames the speed and price numbers as Diogo Almeida's rather than anything the channel measured. That honesty limits how far you can take the video.

If a typed answer can still be wrong, what is the practical alternative?

The description does not propose one. It separates two guarantees, a schema guarantee and a correct decision, and stops there, so there is no recommended validation loop, threshold strategy or fallback pattern to copy from this one.

Is this a review with original testing, or commentary over launch footage?

Commentary over launch footage. The entry labels it an explainer, and the channel states it has not benchmarked the speed claims, so nothing here is original testing. Read it as a framing argument rather than evidence.

Who made the video, and whose figures does it treat as vendor claims?

Reindent published it, treating Diogo Almeida's speed and price figures as vendor claims throughout rather than measured results.

What does a typed output fail to guarantee?

That the answer is factually correct. The entry says the video separates a guaranteed output shape from a correct decision.

What does the channel admit it has not done?

Benchmarked the speed claims itself. The entry labels the clip commentary over launch footage rather than original testing.

CommunityJev Media & Talks

Type-safe is not the same as right

Reindent separates a guaranteed output shape from a correct decision, and treats the launch speed figures as vendor claims.

See highlights

Open youtube.com

Format: Video
Published by: Community
Author: Reindent
Last verified: Sep 2026

Tagscommunity video explainer

Works forJev for Classification

Used inJev for Software & DevTools

Highlights

The description says Diogo Almeida's speed and price figures are presented as vendor claims.
It says a typed output does not guarantee the answer is factually correct.
It says TypeSafe's 0% type-error figure is guaranteed by construction, not an accuracy measurement.

Watch out

Commentary over launch footage. The channel says it has not benchmarked the speed claims itself.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

YouTube: Reindent on why a guaranteed output type is not the same thing as a correct answer.

Hacker News: Simon Willison's overview of Noul, Choice, and Score, and where a decision model fits.

benwerd53 pointsJev introduces a new shape of LLMRead the thread on Hacker News

Hacker News: swyx posts the Latent Space interview with Diogo Almeida.

swyx4 pointsJev: System One Models for Prod, Not God – With Diogo Almeida, CEO, TypeSafe AIRead the thread on Hacker News

X: Diogo Almeida in five minutes

More Jev resources like this

IBM Developer: probabilities, not words

Max Jesch, on IBM Developer, walks Jev as a System One model that returns probabilities for classification instead of generating text.