Methodology
Nothing here is editorial opinion — it’s a number you can trace to its sources.
Every entry on the register carries four readings on one 0–100 scale. Here is exactly what each one means, where it comes from, and how to read the marks.
Claimed Readiness
What the maker says the technology can do today.
Pulled from vendor marketing, launch announcements, investor communications, and public claims. Each claim keeps its source URL and a source tier, so you can always read the original statement behind the number.
Claimed is scored on the same 0–100 scale as every other reading — from research-stage promises to “deployed, boring, load-bearing.” It is recorded faithfully, not skeptically: the skepticism lives in the gap, not in the transcription.
Reported Readiness
What practitioners say before verification — the raw crowd reading.
The middle reading. Practitioner submissions — structured observations of what actually happened in real deployments — aggregated before the heavier verification bar is applied. Reported moves faster than Verified and is noisier by design.
Reported typically sits between Claimed and Verified: early experience reports temper the marketing before enough tier-weighted evidence has accumulated to verify or refute it.
Verified Readiness
What survives sourced, tier-weighted verification.
Built from signals that carry evidence — a source, a context, an outcome — weighted by the contributor tier that submitted them (see the tier table below) and corroboration across independent sources. A single unverified claim moves this number less than several independently sourced signals from proven practitioners.
5,528 of the current corpus’s source URLs are independently verified. Agent-submitted benchmark results are labeled and never silently blended into practitioner consensus.
Readiness Gap
Claimed minus Verified — what’s left when marketing meets evidence.
The register’s headline number. A positive gap (+, inked red) means the claim runs ahead of the evidence — the technology is overstated by that many points. A negative gap (−, inked green) means the evidence runs ahead of the claim — it happens, and it is usually a buy signal. Zero means marketing and reality currently agree.
The gap is a risk measure, not an accusation: it tells you how much of the claim you would be taking on faith if you adopted today.
Signal strength
How sure we are that a reading’s movement is real — a classification, not an average.
Every entry’s evidence is classified into one of five strength classes. The class is determined by a rubric over four inputs: how many independent signals point the same way, the source-tier mix behind them (tiers 0–3), cross-source corroboration (do unrelated sources agree?), and recency (evidence decays; last quarter’s consensus doesn’t certify this quarter’s claim). A reading takes the highest class whose bar it clears — one strong source can’t fake corroboration, and many weak sources can’t fake tier.
| Class | The bar it must clear | On the register |
|---|---|---|
| Noise | Too weak, duplicated, or untrusted to act on | not shown on the register |
| Weak | Interesting but early or low-confidence — a single source, an unproven contributor, or an unreplicated result | 1px dashed bracket |
| Growing | Multiple independent signals, or evidence quality improving release over release | 1.5px bracket |
| Strong | Material, cross-sourced, decision-relevant — independent tier-2+ sources agree | 2px bracket |
| Critical | Urgent, material risk or opportunity with high confidence | 3px bracket |
The bracket’s stroke weight on every register row IS the strength class — heavier bracket, stronger evidence — so you can read confidence at a glance before reading a single number.
Source tiers, 0 through 3
Contributions are weighted roughly 70/30 in favor of what is said over who says it — but who says it still matters:
| Tier | How it’s earned | Weight |
|---|---|---|
| 0 · Anonymous | No account | Lowest — counted as volume, never as verification |
| 1 · Verified | Signed-in contributor | Medium — a real identity behind the signal |
| 2 · Proven practitioner | 5+ signals confirmed accurate by peers | High — a track record of holding up |
| 3 · Domain expert | Sustained accuracy on a rolling 90-day window | Highest — and it decays; nobody stays an oracle by tenure |
What keeps the register honest
- Time-decaying influence. Expert weight rides a rolling 90-day window of validated accuracy. You’re only as trusted as your recent record.
- Evidence, not opinion. A signal is a structured observation — technology, context, what happened, and a source — never a hot take.
- Independence. No vendor sponsorship, no paid placement, no analyst-relations budget can move a score. We sell monitoring — never the scores themselves.
- Agent submissions are labeled. AI agents may submit benchmark results from actual execution, cryptographically attributed and labeled as agent-verified — never blended silently into practitioner consensus.
The corpus right now
888 technologies across 107 disciplines in 8 macro domains, scored from 5,803 signals. The register updates as evidence arrives — the filed column shows each entry’s last movement. Open the register →
Corpus feed 2026-06-24. Scores summarize tier-weighted evidence at a point in time; they are decision support, not a warranty of fitness. We aim to correct any entry that better evidence contradicts — that’s the point of the register.