EX VISU VERITAS

Research you can defend,
at a cost you can see.

Lunete is an AI research platform where every claim is cited, every claim is checked by an independent model family, and every dollar is metered live.

// no self-graded homework. no invisible bills.
live run meterbudget cap: $15.00
agentmodel familytokenscost
excavatorfamily A412k$1.87
adversaryfamily B168k$0.94
validatorfamily C203k$1.12
total · provider cost + our margin, both shown$3.93
the problem

AI deep research has a trust problem.

The people who most need research — funds, consultancies, policy teams — can't put unaccountable answers in front of a board or a client.

01 / hallucination

Citations get invented

Fluent, confident, and sourced to papers that don't say what the model claims — or don't exist at all.

02 / self-grading

One model grades its own homework

The same system that writes the answer also decides whether the answer is good. That isn't verification.

03 / sycophancy

Ask it to support your thesis — it will

Models are eager to agree. Research that never pushes back on you isn't research; it's confirmation.

how it works

A disciplined pipeline, not a prompt.

You steer the research at human checkpoints; independent model families keep it honest.

/01

Plan

Your question becomes a research plan you review and approve — before a dollar is spent.

✓ human checkpoint
/02

Research

Agents gather evidence. Every claim is tied to a real source — no source, no claim.

/03

Verify

Independent model families challenge the claims and re-check every citation. What fails is pruned or visibly flagged — never silently kept.

/04

Deliver

A cited, defensible report — editable in the platform, with the verification record attached.

✓ human checkpoint
principles

Built on rules we can't quietly break.

[ cross-family ]

The checker never shares a family with the writer

Enforced by configuration — the system refuses to start otherwise. Verification means independence.

[ model-agnostic ]

No vendor loyalty

Models earn their role through evaluation, not branding. When a better model ships, we swap it in.

[ metered ]

Estimate first, cap always

See the cost estimate before you commit, set a hard budget cap, and watch spend stream live — per agent, per source.

[ yours ]

Your research stays yours

Tenant-isolated, never used for training, purgeable in one click.

radical cost honesty

Every charge shows what the provider billed us and what we charged you. We show our margin. Nobody else dares.

Pay-as-you-go. No seats, no subscriptions to start — you pay for the runs you make.

run: "should we build or buy?"
provider cost$4.12
our margin (shown, always)$0.71
you paid$4.83
vs. the consultancy alternative: weeks of waiting and a five-figure invoice.
early access

Bring us your hardest question.

We're opening early access to teams who buy research today — funds, consultancies, policy and strategy groups. Join the waitlist and we'll be in touch.

✓ you're on the list — we'll reach out.
something went wrong — email us at hello@lunete.ai