Don't trust one AI with a decision that matters. Get the council.

01 Executive summary

One chatbot gives you one confident answer — with one blind spot. For a decision that matters, that is the wrong shape of help.

WeMind replaces the single voice with a council: a panel of specialist AI minds that argue your question, weigh the strongest evidence, vote, and return a verdict you can act on — Bottom line · Because · The split · Watch out · Do next.

WeMind is a native app for the Solana Seeker phone and Android, built security-first. Sign-in is a wallet signature — no password, no seed phrase, no custody. Free questions run fully on-device, so nothing leaves your phone. When you want cloud speed you bring your own key, or tap once for a free one — WeMind never holds a key server-side. Messages between updated apps are end-to-end encrypted. There are no ads, no tracking, and no data sales; the revenue model itself removes any incentive to exploit user data.

The debate engine, encrypted DMs, multiplayer rooms, a browser version, and the on-chain Pro purchase flow are built and runnable today. What remains is launch execution — store submission, onboarding polish, and the Founder Pass — scoped in a short, concrete plan. WeMind is a solo build by Urban Prism. It is not a token project and not investment advice; the Founder Pass in Section 9 is a collectible access pass, not a security.

02 One model, one blind spot

Consumer AI has trained a billion people to trust a single model's single answer. That is fine for "rewrite this email." It is quietly dangerous for "should I take the job," "is this contract clause normal," "which treatment path fits my situation," or "is this token a scam."

A single model fails in ways that are hard to see from the inside:

• Overconfidence. The answer arrives in the same fluent, certain tone whether the model is right, guessing, or hallucinating. Confidence is not calibrated to correctness.
• One hidden bias. Every model has a training slant and a house style. You get that one worldview, with no dissent in the room to expose it.
• No accountability. No second opinion, no visible disagreement, no record of why — just a paragraph you either accept or don't.
• No sense of the split. Most real decisions are close calls. A single answer flattens a genuine 55/45 into a confident 100/0 and hides the trade-off that actually matters.

People already cope by asking the same question to three chatbots and eyeballing the differences. WeMind productises that instinct — and does it on a phone that protects the question itself.

03 A council, not a chatbot

WeMind seats a council of AI minds and makes them do what a good advisory board does: disagree in the open, then commit to a call.

For each question, WeMind assembles a best-fit panel from a Society of professions — a nurse, a mechanic, a lawyer, a founder, a skeptic, whoever the question actually needs. Each mind reasons from its own role and expertise, not from generic web trivia. They argue in character, hold their ground, and then vote.

You don't get a transcript to wade through. You get a verdict card — the same one the app renders:

The card can carry tappable source links so you can dig deeper, and any verdict can be saved, edited, or AI-summarised as a note. The goal is a decision you can trust because you can see the disagreement — not a black box.

04 How it works

1. Ask anything. Type a real decision, comparison, how-to, or fact-check.
2. The council assembles. WeMind picks the right specialists for the question — from the Society, or from the custom minds you've built.
3. They debate — in character. Each mind argues from its persona and expertise. A built-in fairness & civility guardrail keeps it honest: attack the idea, not the person; no slurs or hate; steelman the other side; stay audience-safe. The tone adapts — spikier on light topics, warm and careful on serious ones.
4. They vote. The strongest case wins; the split is recorded.
5. You get a verdict. Bottom line, because, split, watch-out, do-next — plus source links and one-tap save.

Everything the app builds visually is tunable, and your settings persist — the product is designed to be shaped, not locked.

05 The three brains

WeMind lets you choose which intelligence answers, and always tells you which one is active with a colour-coded light. This is the heart of the trust model.

Because most usage runs on-device or on a key you control, WeMind's own running cost is near-zero — which is exactly why it can refuse ads and data sales and still be sustainable.

06 The product

WeMind is a full app, not a demo:

07 Technology & architecture

Client. WeMind is built in Unity 6 (URP) for Android, targeting the Solana Seeker. The interface is assembled in code with a layout system that keeps every element manually adjustable and preserves those edits — nothing resets on the next build.

On-device intelligence. The private brain runs quantised small language models locally (Qwen3-class models at 0.6B / 1.7B / 4B, in efficient GGUF form) so a full debate can happen with nothing leaving the phone. Downloads are size-checked, with cryptographic hash-pinning being finalised so a device only ever runs a verified file.

The debate engine. A single orchestration layer builds each mind's brief: it embeds the persona, grounds the mind in its own role rather than general knowledge, applies the fairness-and-civility clause and the adaptive-tone rule, runs the argument, then composes the verdict and the vote. Mid-debate, a mind can ask you a short, human clarifying question — phrased like a friend, not a form.

Backend (thin by design). Two small Cloudflare Workers back the online experience: one for accounts, rooms, and Pro entitlement; one as an origin-locked AI proxy so keys used for the free path stay server-side and never reach the client. Protection includes rate-limiting, input caps, and a bot challenge on sign-in. There is no central store of user AI keys, and on-device questions touch no server at all.

Identity & payments. Sign-in is a wallet signature — no passwords, no seed phrase, no custody. Pro is purchased on-chain from the Seeker wallet.

08 Privacy, security & trust

For WeMind, security isn't a settings page — it's the product.

• Private by default. On-device mode means your question and the debate never leave your phone.
• No key custody. Your AI keys live on your device; the servers never hold them, and the AI proxy is origin-locked.
• No seed, no password. Wallet-signature auth means WeMind never touches your seed phrase or funds.
• End-to-end encrypted DMs. Messages between updated apps are encrypted so the server can't read them.
• Anti-scam guard. Answers touching seed phrases, private keys, or "send money" are flagged; opening any link from an answer needs an explicit confirm.
• No ads. No tracking. No data sales. The business model is paid software, not attention or data — so there's no incentive to harvest you.

This posture is not marketing garnish; it is the reason a security-conscious, wallet-native audience is the right first market.

09 Economics

WeMind's costs are near-zero by design — on-device compute, bring-your-own-key, and a 0% Seeker dApp Store fee — so it can be funded by people who value it rather than by advertisers or data brokers. Pricing is fiat-denominated for clarity but paid on-chain; nothing auto-renews, and every payment is auditable.

BYOK on every tier means the marginal cost of a heavy user is borne by the key they already own — which makes near-100% software margins realistic without ads. Payments are crypto-native from the Seeker wallet: USDC for annual and lifetime so the price stays exact, SOL for the Founder Pass.

The paywall is drawn honestly: what's gated is bigger councils, deeper reasoning, and bigger rooms — not basic use. The private on-device council is free forever.

10 Why Solana, why now

• Seeker dApp Store fit. A wallet-native audience, a 0% platform fee, and on-chain payments. WeMind's pricing only works because of near-zero fees and a card-free flow.
• Wallet-signature auth, no custody. Clean and natural only on a wallet-native platform — and it doubles as the security story.
• Timing. The MVP is built and a concrete launch plan exists. This is the moment where the last mile — not a moonshot — stands between a runnable app and a shipped one.
• Ecosystem alignment. A Founder Pass as an owned Solana collection, and a phase-2 marketplace for community-built minds, deepen Solana-native usage over time.

11 Market & positioning

The category is decisions that matter — the questions people currently take to a single chatbot and quietly worry they shouldn't. WeMind's edge over a single-model app is structural, not cosmetic: visible disagreement, a recorded vote, role-grounded experts, and a phone that keeps the question private. Its edge over "ask three chatbots yourself" is that WeMind assembles the right panel, runs the argument, enforces fairness, and hands back one act-on-able card.

Positioned as security-first consumer AI on Solana, WeMind sits at the intersection of two rising tides — distrust of single-model overconfidence, and demand for on-device privacy — on the one platform where wallet-native payments and a private-by-default posture feel native rather than bolted on.

12 Roadmap

13 Risks & honest limitations

A whitepaper that only lists strengths isn't trustworthy. The real ones:

• Pre-launch, no traction yet. WeMind is a demonstrably-real build with a plan, not a product with users. Adoption is unproven.
• AI can still be wrong. A council reduces single-model blind spots and surfaces disagreement, but it is a decision aid, not an oracle. Verdicts can be mistaken; treat them as informed input, not final authority — especially on legal, medical, and financial questions, where WeMind is not a substitute for a professional.
• On-device limits. Small local models are weaker than frontier cloud models; on low-end phones the private brain trades some quality for privacy. The three-brains design is the honest mitigation — the user chooses the trade-off.
• Third-party dependence. Cloud paths rely on external model providers whose pricing, availability, and terms can change.
• Store & regulatory risk. dApp Store approval, app-store policy, and evolving AI/consumer regulation are outside the team's control.
• Solo-founder / bus-factor. One builder means speed and coherence, but concentration risk. The plan is scoped precisely to de-risk shipping.
• Founder Pass is planned. Not live, launched as a waitlist first, and a collectible access pass — not a security, not a promise of return.

14 Team

Urban Prism — solo founder & developer. WeMind is designed, built, and shipped by one builder who knows exactly what is done and what is left. The app is real and runnable today, and a self-authored release audit names every remaining launch blocker and its fix — the kind of precise, honest accounting that de-risks a solo build.

15 Disclaimer & contact