Early build · scores computed by the deterministic engine from real, sourced developments (30 days) · weights v1.0-prior
Updated 06:00 UTC
next refresh in
Today › Methodology
Methodology · 1.1.0-prior

How the Vikshy Score works

Every number is computed by a deterministic, published rules engine, an LLM may extract candidate facts, but a versioned formula calculates the score. Same evidence in, same score out, forever reproducible.

What it measures

The Vikshy Score is a daily 0–100 reading of an AI provider's current competitive strength and market pull. It is not a "best AI" verdict, not a valuation, and not an opinion. Scores are comparative within the covered cohort.

The formula

Each of seven factors gets a score anchored on where the provider stands today, then tilted by its 30-day momentum:

Factor = Standing + 0.35 × (Momentum − 50), clamped 0–100.
Momentum runs 0–100 (50 = steady). Neutral momentum means the factor sits at its standing. A surge lifts it; a slide lowers it, but bounded, so no single launch dominates.

The Vikshy Score is the weighted sum of the seven factors:

FactorWeight
Capability24%
User & developer acceptance18%
Adoption & distribution18%
Product & release execution14%
Value & availability10%
Ecosystem8%
Reliability & institutional access8%

Weights and the momentum tilt are v1.0-priors, theory-based starting values that pass a published expert-elicitation and robustness protocol before final publication.

Interpretation bands

80–100Frontier leader
65–79Strong contender
50–64Challenger
40–49Focused player
< 40Trailing

Evidence tiers

Every input is graded. Primary (official docs, filings, status pages) and independent measurement (arenas, benchmarks, usage data) can establish facts. Reputable reporting corroborates. Community signal (Reddit/HN/X) is capped and can never establish a fact, only colour reception.

The launch-proof rule

A loud launch earns only a bounded, decaying lift (about +1 point on day one). A provider climbs durably only when independent benchmarks, real user preference, and adoption data confirm the change. Marketing spend earns nothing on its own.

What does not count

Rumours and unverified leaks; paid promotion; marketing reach; unverified social posts; opinion pieces. Government and legal actions are measured for market-access and operational impact only, never as moral or political judgment.

Full specification, evidence log, corrections history and the weight-derivation protocol publish with the public beta. This is an early build (1.1.0-prior).