Name: Glass Box
Brand: DEPLOYR
Price: 349 USD
Availability: PreOrder

Picture this

Who it is for

A hiring firm screens CVs with AI. Glass Box gives each candidate a TEE-signed receipt of the model and prompt, so the process can be audited.

Two ways to own it

What you get

Base$349

$87 deposit to reserve

TEE inference integration, receipt contract, audit viewer, API

You choose: chain
You choose: name and logo
You choose: accent color
You choose: model choice

Customfrom $2,500

Built around you

Reshape it entirely: new features, your flows, your brand end to end, integrations you need. We scope it with you.

Everything in Base
Custom features and design
Priority build

Why it is new

What exists today, and what this does differently

What is new

Infra and standards exist; a consumer-facing receipt app is still uncommon.

Closest things out there

OpenGradient offers a TEE attestation checker, Meru (hackathon) issues per-inference receipts, and an IETF draft defines Attested Inference Receipts; no shipping consumer receipt app found.

Checked with web searches in October 2026. A search is not proof that nothing similar exists, and the market moves fast, so we recheck before every build.

What it unlocks

The human in it

Fairness you can prove to the people affected.

The Regatta Board

Races it can enter

buildathon · Ritual · rolling

Ritual builder programs

unknown

Official page ↗

Onchain AI; eligibility unknown

Their rule on existing code: unknown

What you own

The whole app, not a starter kit.

The complete Glass Box app: a firm runs AI screening in a sealed enclave, every decision prints a signed receipt the person can check, appeals re-run with proof and a Monday audit catches swapped models, with a paywall and receipt agent built in.

Base price · $349

The user app, complete

Every buyer gets the same finished app. Basic branding and copy changes come through a quick form.

Three step onboardingHome

Shown once: every decision gets a receipt, the person can check it, you audit every Monday.

Receipt status and next actionHome

Receipts this week, cost per screen, audit streak and one next step.

Receipt verifierReceipt

Model, rubric, receipt and attestation hashes checked; CV text never published.

Three real casesReceipt

A changed rubric, a fair no explained, and a quietly swapped model caught.

Appeals inbox and re-runsAppeals

Roll back a rubric and re-run affected screens.

Drift alertsAppeals

Alerts the day a model or rubric changes. Gated behind Team.

Weekly audit sampleAudit

Sample receipts, check every hash, sign the tally. Gated behind Team.

Regulator exportAudit

One verifiable file of every receipt. Gated behind Team.

Model choice and plansMe

Large or small open model with cost per screen; Starter, Team and Compliance through USDC checkout.

Open the full app demo

Add-on

Owner console

Your admin side: revenue, paywall editor, users, rules and the agent log.

Revenue and paywall editorRevenue

MRR, incoming USDC, payment methods, tiers and gates.

Receipt rulesRules

Receipt chain, rubric visibility, per criterion scores and retention.

Receipt agent logAgent

The clock drift and unreadable hash failures and fixes, plus costs.

Roadmap build requestsRoadmap

v2 and v3 scopes with add-on prices.

Top-up

Growth and referrals

Referral and growth mechanics, switched on when you are ready to grow.

Outcome share cardAppeals

Candidates share that they could check the decision.

Tool partner loop and referralsGrowth

Sign a hiring tool partner, receipt links in rejections and referrals.

The money path, built in

How it earns for you

Candidates check receipts free; firms pay monthly in USDC for volume, audits and compliance.

StarterFree100 receipts a month, candidate viewer, appeals inbox

Team$49 a monthUnlimited receipts, weekly audit sample, drift alerts

Compliance$200 a monthRegulator exports, retention controls, audit sign off

About $0.001 to $0.009 of inference and $0.0004 of Base gas per screen.

Prices are yours to set. Illustrative, not a promise of income.

Growth loops, built in

How it spreads

A receipt link in every rejection email
Firms listing receipted AI screening on job posts
AI tool partners adding receipts by default
A plain guide to what receipts do and do not prove

Roadmap · build the next phase with us

Where it goes next

v1 included

Receipts end to end

Open model in a TEE with attestation
Receipt contract on Base
Plain language candidate viewer
Appeals inbox and re-runs
Weekly audit and paywall

Included

Lending and insurance

Credit and claims rubrics
Adverse action letters with receipts
Regulator portal
Bias sampling by group

Add-on from $800

Receipts for any AI tool

SDK for AI tool builders
Receipts on Ritual
Public verifier page
Receipt badges for job posts

Add-on from $1,400

For investors and grant reviewers

The pitch, ready

Make every AI decision about a person come with a receipt that person can check.

Problem: People rejected by AI cannot see which model or rubric judged them, and firms cannot prove fairness.

Solution: AI jobs run in a sealed enclave and sign the model, rubric and answer onchain, with appeals and audits.

Why now: Firms screen with AI at scale while audit duties for automated decisions are arriving.

Market: Hiring, lending and insurance teams with audit needs, from free to $200 a month.

The ask: Ten hiring teams and one applicant tracking partner this quarter.

Funding is never guaranteed. Check each program's rules.

For hackathons and demo days

Your 2 minute demo

Home: Lerato rejected in 11 seconds.
Receipt: verify; rubric v7 hid a gap year penalty.
Owner Agent: clock drift and plain language fixes.
Appeals: roll back to v6, re-run, interview offered.
Audit: Monday sample gated behind Team.
Owner Revenue: change the Team price.
Owner Growth: sign a tool partner.
Owner Roadmap: lending and insurance.

Many hackathons require code written at the event. See the Regatta Boardfor each race's rule.

Make it entirely yours

Custom rubrics, new enclave hosts, regulator portals and branded viewers, scoped on a call from $1,000.

Custom from $2,500

BLUEPRINT

Why it matters and how it really works

The need

Firms already use AI to screen CVs, loan applications and claims, and the people affected get a one line rejection with no way to know which model or which rules judged them. Even a well meaning firm cannot prove afterwards that the model it described is the one that actually ran.

What exists and what is new

The building blocks exist. OpenGradient offers a TEE attestation checker for inference, Meru (a hackathon project) issues per-inference receipts, and an IETF draft defines Attested Inference Receipts. Research verdict: rare, medium novelty. No shipping consumer-facing receipt app was found. What this app adds is the person-facing layer: a readable receipt for the candidate, the rubric version on every decision, a one tap appeal with proof, and audit tooling for the firm. It packages existing attestation infrastructure; it does not invent it.

The data it runs on

Per decision: a fingerprint of the model weights, a fingerprint of the rubric or prompt, a fingerprint of the input (the CV), the output score, and the enclave's attestation quote. Fingerprints and attestation references go onchain; the CV text and the full output stay encrypted with the firm for a retention window the firm sets.

1.The mission

Make every AI decision about a person come with a receipt that person can check.

2.Why people can't resist it

Which model judged me? A candidate opens the receipt and sees the model, the rubric version and the score per criterion, in plain words.

Appeal with proof When a rule looks wrong, the appeal carries the receipt. The firm can re-run on another rubric and issue a new receipt.

Monday audit Hiring leads spot check ten receipts a week and keep a clean audit streak they can show regulators and clients.

3.How it works for real

Enclave inference: The firm's screening model runs inside a TEE (for example an Intel TDX or NVIDIA confidential computing GPU host, or Ritual's TEE execution). The enclave produces an attestation quote binding model, prompt, input and output hashes.

Models: Open weight models with published hashes: a large tier (around 70B parameters) for nuanced screens and a small tier (around 8B) for high volume, chosen by the firm and pinned per rubric.

Receipt contract: A small contract on Base stores receipt records: attestation reference, model hash, rubric hash, input hash, score, block time. Optionally mirrored on Ritual for native onchain AI verification.

Attestations: Rubric versions are published as EAS attestations by the firm, so a rubric change is visible and dated.

Verifier and viewer: A web viewer checks the enclave quote against the vendor's root certificates and the onchain record, then renders a plain language page for the candidate.

Indexer and alerts: An indexer watches receipts for model fingerprint drift or rubric changes and alerts the firm, which is how a silent model swap gets caught.

API: A REST API for the firm's applicant tracking system: submit a job, get a receipt id, link it in the rejection email.

4.Cost per use and pricing

Per screen, roughly: model inference about $0.001 (8B) to $0.009 (70B) for a 3,000 token CV and rubric at open model hosting rates, a confidential computing premium on top, and about $0.0004 of Base gas per receipt record. Example: a firm screening 2,000 CVs a month on the large tier pays about $20 to $40 in compute and under $1 in gas. Prices are estimates and change often.

Starter $99 a month

Up to 1,000 receipts, the candidate viewer and one rubric.

Team $399 a month

Up to 10,000 receipts, rubric versioning, appeals inbox and weekly audit reports.

Regulated Custom

Dedicated enclave hosts, longer retention, exports for regulators and auditors.

5.Growth loop

Every rejection email carries a receipt link, so each candidate sees the brand. Candidates who were treated fairly share that they could check; firms list 'receipted AI screening' on job posts to attract applicants, which pulls in other firms.

6.Risks and guardrails

Proof is not fairness: A receipt proves what ran, not that the rubric is fair. The app shows the rubric version and encourages human review; it does not certify fairness.

Privacy: CVs are personal data. Only hashes go onchain; text is encrypted, retained for a set window, and deleted on schedule. Data protection laws such as POPIA and GDPR apply.

Enclave trust: TEEs depend on hardware vendors and have had vulnerabilities. Receipts name the attestation type so verifiers can weigh it.

Gaming: Candidates could probe rubrics to game them. Firms can choose fingerprint only mode, with rubric text released during appeals.

Legal exposure: Receipts may be used in disputes. Firms need legal review before switching it on, and the app is not legal advice.

7.Build order

Run one open model in a TEE host and produce an attestation quote per job
Deploy the receipt contract on Base and write records from the job runner
Build the candidate viewer with plain language explanations
Add rubric versioning with EAS attestations
Add the appeals inbox and re-run flow
Add drift alerts and the weekly audit sample

8.Where this is going

Now
Receipts for CV screening at a handful of firms that want to prove they use AI carefully.
Next
Loans, insurance claims and tenancy checks, anywhere an AI decision affects a person.
Later
Regulators and courts accept attested receipts as standard evidence of how an automated decision was made.
The world it builds
No one is judged by a machine they cannot question. Every automated decision about a person comes with proof they can hold.

Costs and tools are estimates at the time of writing. General information only, not legal or financial advice.

OWNER PLAYBOOK

Own it, then chart your voyage

Nine routes from one vessel. Tap a port to open its chart.

Why own Glass Box

Glass Box gives every AI answer a signed receipt so the people affected can check what model and prompt was used. Owning it means the clients who need audit trails pay you, and your onchain record shows you build trust infrastructure. On someone else's tool, your clients' audit history lives on their servers.

What’s stopping you?

“I can't code.”

You do not need to. Glass Box ships ready to deploy, and every path here is written for non coders. When you want to change something deeper, you can hire help or ask DEPLOYR support.

“Is it really mine?”

Yes. It is deployed from your wallet, the contract address is public, and the brand, users, and revenue are yours. DEPLOYR does not hold your keys or take a cut of your app's revenue.

“Does a receipt make the AI fair?”

No. It proves what ran, not that it was right. Say this plainly to clients. It makes the process auditable, which is a step towards fairness.

“What about privacy of the data?”

Store hashes and receipts onchain, never raw CVs or personal data. Check data protection rules where your clients operate.

“What if the TEE or provider changes?”

Receipts already issued stay valid. Keep your provider options documented so you can switch if needed.