Learning and development · 2029
The verticals where being wrong costs something specific.
Law, medicine, engineering, the public record. In these the cost of a wrong answer is not embarrassment, it is a specific and often nameable harm, and that changes what a study tool is allowed to be. Every claim has to carry the source it came from and the check it survived.
Nothing described on this page is built. It is a 2029 line on a roadmap, and it is deliberately the furthest out of the three.
- 2029planned, not promised
- 0lines written
- 1engine it will run on
Why it is not a second product
One engine. Very different rooms.
This is the surface that most needs the thing the engine already does: a verifier drawn from a different model family than the agent it checks, so a shared blind spot cannot pass as agreement. In higher education that is a quality decision. In medicine it is the whole product.
The regulated verticals are last for a reason.
Decided
Settled
Citation is not optional here
A claim that cannot be traced is not shown at all in these verticals.
Cross-family verification is mandatory
Spent on every path, and the cost of that is accepted.
It is not free
This is the tier that pays for the others, and the landing page says so.
Not decided
Still open
Which vertical is first
Four are named. Choosing one is a two-year commitment.
The regulatory posture
Undecided in every jurisdiction, and most likely to move the date.
Firms or people
These are different products wearing the same name.
What we will not do
We will not let a model in a regulated vertical answer without a source, and we will not describe an unverified answer as verified. If the check cannot be run, the honest output is that the check could not be run.
If this is your problem
Tell us before it is built, not after.
If you are in one of these verticals and can tell us what would make this unusable, that is the most valuable thing anybody could send us in the next two years.