DecAI Fit Metric

DecAI Fit is a composite index (scaled from 0 to 6) representing a crypto project's alignment with the principles of Decentralized Artificial Intelligence (DeAI). The metric is strictly constrained by the quality of provided evidence and is designed to resist manipulation (such as AI-washing and onchain-washing).

Calculation Algorithm

Project profiles are populated with binary (0/1) flags, which are used to calculate the final score using fixed formulas.

1. Component Scores (0–3 points each)

A) AI Reality — depth of AI integration:

AI_0: Has Tier‑1 declaration of AI intent (roadmap/official update), but no evidence of a working AI component at the protocol level (1/0)
AI_1: Describes what exactly the AI does (1/0)
AI_2: Has working demos/examples (1/0)
AI_3: Has metrics/benchmarks/quality assessments (1/0)
AI_4: Has technical description/architecture/model versions (1/0)

$a = \min(3, AI_0 + AI_1 + AI_2 + AI_3 + AI_4)$

B) Onchain Necessity — dependence on blockchain:

OC_1: Has contracts/addresses/network (1/0)
OC_2: Onchain = not just a token, but has useful logic (access/accounting/rules/DAO) (1/0)
OC_3: Onchain is critical (without blockchain, the product breaks) (1/0)

$o = \min(3, OC_1 + OC_2 + OC_3)$

C) Token Necessity — token utility:

TK_0: Token does not exist (1/0)
TK_1: Token is used to pay for services (1/0)
TK_2: Token is needed for incentives (rewards/staking/slashing) (1/0)
TK_3: Token is needed for governance (voting) (1/0)
TK_4: Token is replaceable (can be replaced by USDC/subscription without losing meaning) (1/0)

If TK_0 = 1, then $t = 2$. Otherwise $t = \text{clamp}_{[0,3]}(TK_1 + TK_2 + TK_3 - TK_4)$.

D) Evidence (by Tier): Presence of Tier‑1 evidence for key blocks:

E_AI_T1 (1/0)
E_OC_T1 (1/0)
E_TK_T1 (1/0)
E_only_T3: Sources are exclusively Tier‑3 (1/0)

$e = \min(3, E_AI_T1 + E_OC_T1 + E_TK_T1)$

2. Base Index

Preliminary sum of components: $S = a + o + t + e \in {0, \dots, 12}$

The Base Score ($B$) is mapped to a 1-6 scale:

0–2 → 1
3–4 → 2
5–6 → 3
7–8 → 4
9–10 → 5
11–12 → 6

3. Governance Caps

To ensure mathematical aggregation doesn't obscure fundamental gaps (e.g., strong token mechanics but zero verifiable AI), strict caps are sequentially applied to the Base Score. Each rule can only lower the final score.

R0 (AI Reality Zero-Forcing): If $a = 0$, the final DecAI Fit = 0. A project with no verified AI capabilities ($a = 0$) does not meet the threshold for Decentralized AI classification as of the review date.
R1 (Informational Noise Cap): If E_only_T3 = 1, the score is capped at a maximum of 2.
R2 (Missing Core Evidence Cap): If E_AI_T1 = 0 or E_OC_T1 = 0, the score is capped at 3.
R3 (Token Verifiability Cap): If a token exists (TK_0 = 0) and E_TK_T1 = 0, the score is capped at 4.
R4 (AI Reality Cap): The final score cannot significantly outpace the foundational AI score: if $a = 1$, cap = 3; if $a = 2$, cap = 5.

4. Obsolescence Penalty

If a project card hasn't been updated for over 550 days (last_verified_days > 550), the final score (if greater than 0) is reduced by 1 point.

Operational Definitions

To ensure objective evaluations, strict rules are applied when assigning binary flags:

Every flag (1) requires justification: A source link plus a short fact/quote. Without justification, the flag is set to 0.
AI_0 (Declaration of Intent): Set to 1 only if there is a Tier-1 source (roadmap, official docs) explicitly stating AI as a plan or intent, but lacking Tier-1 evidence of a working AI component at the protocol level (meaning AI_1..AI_4 remain 0).
AI_3 (Metrics/Quality): Set to 1 only if there are quantitative quality indicators, it's clear what was measured, the context is specified (version, date), there is a Tier-1 artifact (or reliable Tier-2 confirmation), and the metric relates to the project's actual product rather than a third-party API.
OC_3 ("Replace-the-Chain" Test): Set to 1 if hypothetically replacing the blockchain with a standard server/database breaks a core property of the product (e.g., trustless settlement, censorship resistance, verifiable reward distribution). If "onchain" only means payment or token issuance, it is set to 0.

Anti-Manipulation (Anti AI-Washing)

Source Priority: Tier-1 (code, contracts, official docs) > Tier-2 (aggregators, analytics) > Tier-3 (marketing, influencers, rumors).
Conservatism: In case of data shortage or conflicting sources, flags default to 0 until re-verified.
Ignoring Marketing: PR, claims, and "partnerships" without verifiable artifacts do not raise flags to 1.