Cardano deserves an app that actually needs Cardano — our earnest attempt (Rational Protocol, preprod alpha)

[timmyedge]

Why I’m posting here first

I’ll be honest, I’m more interested about the ledger than the finance and money angle of crypto. A decentralised, deterministic, append-only record that no single party can quietly rewrite is one of the few genuinely useful things this technology gives us, and most of crypto spends that gift on moving money around. I wanted to spend it on something else: a trustworthy, contested, machine-readable record of what people believe.

So I built Rationall (rationall.io), and it’s now running on preprod testnet. This isn’t a launch — it’s me bringing the thing to a room I expect to be hard on it, and asking you to be.

There’s a recurring frustration in this community I take seriously: how can we distinguish the purpose of Cardano? and where’s the consumer-facing app that genuinely needs Cardano’s decentralisation — not a dApp that would run just as well on a spreadsheet plus a centralised API with a wallet button bolted on? I don’t have a clean answer to that. This is our attempt, and I’d rather it get stress-tested by people who’ll understand the context.

What it is: token as mechanism, graph as product

Most crypto apps make the token the product. The token (RTNAL) is an incentive-routing mechanism. The product is a conviction graph: a persistent, on-chain, machine-readable map of normative claims — ethics, politics, science, the economy, society, and everyday life — and how much skin people put in the game behind them. You might think of it as a mash-up of social media and prediction markets. You stake or bond ADA behind a claim you believe and earn RTNAL and ADA for backing claims that draw durable support. The output isn’t a verdict — it’s a distribution: who backs what, how hard, and how positions cluster and oppose.

The framing I most want challenged: conviction, not veracity. There’s no oracle resolving these to a verdict — “a carbon tax is the fairest emissions policy” has no settlement. The conviction scoring is the artifact, not a step toward some final answer, and the emergent topology (clusters, stance-edges between supporting and opposing positions) is the rest of it.

Why Cardano

This only works on a ledger that can be trusted as a neutral substrate that no one can quietly rewrite — for a record of contested ideas, who can, or rather who can’t control, co-opt and corrupt it is the whole ballgame. The longer thesis is that a tamper-evident record of human conviction is exactly the kind of data a decentralised ledger should support — in combination with decentralised storage — and as AI and autonomous systems increasingly ingest or reference what humans believe, that input should be attributable, versioned, and auditable, not a black box.

The AI angle

A decentralised map of human values — is a useful provenance signal for AI builders: an auditable input grounded in what people actually prove conviction in. The API layer implements a signed attestation stack — (switches on with user accounts) — that lets integrators pin community-curated value collections as guardrails over a model’s answers, receive a signed receipt of exactly which collections and immutable version hashes were applied, and hold a signed compliance certificate committing to a fixed guardrail bundle.

Honest tradeoffs (read this part)

I’ve watched enough projects oversell, so:

• Not formally audited yet. — but not unexamined either. Contracts are written in Aiken with unit and property-based tests, plus substantial red-teaming and internal verification. No third-party security audit is published; treat it as alpha software on testnet ADA.
• Sybil-resistance is one of the hard parts. Raw capital doesn’t buy influence linearly: per-stake weight is √(stake) × reputation (100× the ADA ≈ 10× the weight), and a claim’s conviction is further multiplied by staker-diversity and source-diversity terms, so a broad base of independent backers beats one whale. Reputation is earned, soulbound, non-transferable — money can’t shortcut it. All of that only bites if identities are real, and “soulbound” isn’t “un-fakeable,” so validating Sybil-resistance is useful here.
• Lineage, not magic. This builds openly on token-curated registries, curation markets, and Kleros-style dispute resolution. The narrow novelty I claim: applying them to normative claims with no truth-oracle (distribution-as-artifact) and treating the emergent stance-graph as the output.
• AI Guardrails. None of this asserts a model is ethical or correct — there’s no oracle for that here either. It’s provenance of process: a verifiable trail of which human-curated value-sets shaped an output. That’s the ceiling of this claim.

How to try it

Sign in with Google via a zkFold smart wallet (no browser extension) or normal CIP-30 wallet, grab some faucet ADA, and you can stake behind a claim and start earning RTNAL. Preprod testnet — nothing here is mainnet, nothing is for sale.

What I’m asking for

Break it. Specifically:

1. Sybil/identity — where does √(stake) × reputation + diversity multipliers fall apart under a determined attacker farming wallets?
2. Contract design — reward routing and the principal-protection mechanism in Aiken; what would you attack before an audit would?
3. The mechanism — does “conviction, not veracity” hold, or does it quietly collapse into a popularity contest?
4. The Cardano fit — am I leaning on the secure root of trust for real reasons, or rationalising? Is building for the base chain - with a mindset of partner-chain or layer 2 upon scaling need - the right scoping?

• Try it: https://rationall.io/about
• Discord: Rational Protocol
• X: @rationall_io

[COSDpool]

This is an interesting kind of token economy that I personally haven’t seen before. It seems like (with enough participation) it could produce the same kind of consensus base as prediction markets… it will be instructive to see how well people are drawn to the staking premise vs. the huge support of the gambling premise on sites like Polymarket.

[timmyedge]

Thanks, that’s a comparison I find useful to think against.

On staking vs. the gambling pull, I tried not to force a single premise — there are tiers:

• Social signalling (likes/endorsements) — free, and still feeds scoring as a soft signal.
• Principal-protected staking — skin in the game without the gamble: principal returns at unlock, you only swap native yield for protocol scored yield.
• Bonding / contests — kind of like veracity bonds, a contest layer where you genuinely can lose the bond. Closest to the prediction-market draw.

The one distinction from prediction markets: normative claims don’t settle. There’s no resolution date on “a carbon tax is the fairest emissions policy”, so you’re not betting on an outcome — you’re expressing and defending a position over time. Whether that pulls people, and makes for an interesting place to hang out, remains to be seen.

Participation is definitely make-or-break — so what I can actually control is lowering the barrier to help with the cold-start problem. E.g. zk smart wallet for Google sign-in without seed phrase or extension, and the UX is shaped like a normal social feed, not a trading terminal. A little bit closer to where Grandma and Grandpa can get involved.

[timmyedge]

Here is an explanation of the conviction graph:

The Conviction Graph

[Nyxar0x]

really like the honesty on tradeoffs here.

one gap though: the AI attestation API is the only stated revenue line, but does that revenue ever touch RTNAL? buyback, burn, fee-split to stakers, anything? right now RTNAL’s value case is just earn it by staking

nothing pulls demand back toward it from outside the loop. without that, what anchors price beyond new ADA coming in?

[Nyxar0x]

the no oracle, no verdict framing makes sense philosophically. but the reward pays out for claims that draw durable support isn’t that just rewarding you for predicting what gets popular, not for how strongly you believe something?

that seems to pull people toward backing what’s likely to win support, rather than what they actually believe.

which is the opposite of what “conviction over veracity” is supposed to mean.

how does the scoring tell the difference?

[timmyedge]

Thanks for the reply. There is some small coupling of the token to activity, but it was a deliberate choice to decouple RTNAL especially for the core usage (staking/bonding/delegating): you stake ADA and earn ADA + RTNAL back. Wanted the core based in ADA to keep the UX simple (especially for non-crypto users) and to drive real ADA activity. Only a smaller set of functions actually need the token to play.

So there are two structural places — and a third under consideration — that drive token demand:

1. It’s ties to activity. Hard-capped 21M supply. RTNAL-denominated fees (bond discounts, jury-escalation) burn 10% and recycle 80% to stakers. As emission rewards taper, they shift to fee-recycling—so at maturity, staker yield ≈ f(activity). More adoption → more burn (scarcity) + more yield.

2. It governs a primitive that matters. If the graph becomes something AI leans on, control of its on-chain params (weightings, fee splits, treasury) is valuable. Voting weight scales with the RTNAL you hold.

3. A revenue → RTNAL buyback (what I’m weighing, and what you pointed at). Routing a slice of protocol revenue (API + ADA protocol fees) into buying RTNAL on-market to burn/recycle. That’s where external revenue could touch the token: adoption → revenue → buy-pressure.

Honest bit: today this is more designed than proven, and the real bet is the graph becoming critical in the first place. I’m genuinely still weighing this balance while trying to keep core usage as ADA-only and frictionless. If you were designing the value capture, what would you do differently?

[timmyedge]

Fair challenge — quick concession first: some of our copy says “stake what you believe,” which is a simplification of the input. The mechanism doesn’t require belief; it rewards backing claims that draw durable support, and collective belief is the output.

But consider what predicting durable support for a normative claim actually takes. To win, you have to identify what will resonate broadly — that’s not a flaw; it’s a Schelling point. Predicting what others will back is predicting what they’ve reasoned others would independently reason too: wisdom-of-crowds, not trend-chasing.

The maths keeps it durable, not merely popular:

- Bandwagoning is the worst strategy — the early-mover premium is ~5× for backing a claim before the crowd, decaying to ~0.2–0.5× once it’s already popular.

- The only way a latecomer earns on a popular claim is to hard-lock 90+ days — trend-chasers won’t jail their capital, genuine believers may.

- Tenure and earned reputation compound, so proven early-and-right judgment beats trend-hopping.

This has been stress-tested in simulation — adversarial runs pitting honest stakers against bandwagoners, whales, coordinated sybils and trolls; long-horizon fairness runs checking that strategies don’t drift toward gaming over time; and sybil- and whale-concentration tests. The pattern holds throughout: early, honest, committed participants earn the most, and the gaming strategies the least.

So it rewards early, committed, diverse-backed prediction — at which point “predicting what will endure” and “genuinely believing it”, in a way, converge.

[Nyxar0x]

if I were designing it, I’d make the taper adaptive

decay only when fee recycling actually covers a chunk of what emissions were paying, so slow adoption doesn’t mean a yield cliff.

I’d also start the buyback small and early instead of waiting even a little real revenue hitting RTNAL now proves the loop works. and I’d protect the reputation/jury flows from tapering as fast as staking rewards, since those keep the graph honest.

I learn and propose how projects should capture value or make their token more important in their protocol rather than just existing (I audit tokenomics)

[Nyxar0x]

Alright as long as it’s been stress tested it’s full proof

Good work

[timmyedge]

Really appreciate this — and good timing.

The adaptive taper makes a lot of sense and the buyback too. Jury is funded straight from the escalation fees / contested bonds (self-balancing, never from the emission pool), so they’re already decoupled from the staking taper.

I’ll investigate these further. Given you do this professionally, I’d genuinely value your eye as it firms up — thanks again.

[Nyxar0x]

Sure!!

Mind sharing your contact info

Will love to connect with you

[timmyedge]

DM’d in the discord server