/self-audit

Verg measures itself.

Curation is Verg's biggest weakness. Every pattern is chosen by one person's taste. That's a single point of failure critics have rightly flagged. This page doesn't solve it. It surfaces it — honestly, with the same instruments we'd use on anyone else's claims.

Patterns analyzed

63

latest convergence.json

Mean novelty (counter)

n/a

Haiku skeptic verdict · 0=rehash · 1=novel

Contested

0

base curator & skeptic disagree

1 · Slurry test

Each pattern label is embedded (locally, via nomic-embed-text) and compared to a bank of 25 hand-written generic “AI trends right now” phrases. Cosine similarity ≥0.85 → the label is indistinguishable from what any LLM would produce given an empty prompt. Those patterns get flagged and de-ranked.

49
9
5
sharp49(77.8%)marginal9(14.3%)slurry5(7.9%)

The 3 patterns flagged as slurry are exactly the generic-sounding labels critics pointed to on launch. They're still visible on /emerging, just dimmed and sorted to the bottom. No hiding. No silent filter.

2 · Counter-curator

A second curation pass with a different taste. Where the base curator asks “what are sources converging on?”, the counter-curator (Claude Haiku) asks: “is this convergence genuinely novel, or a rebrand of something older?”Two independent tastes. Disagreement is the interesting signal.

63
novel0(0.0%)mixed0(0.0%)rehash0(0.0%)unclear63(100.0%)

The skeptic's bar is strict. Zero patterns rated fully “novel.” 0 rated as rebrands of older ideas. 0as meaningful recombinations with clear predecessors. This is honest, not flattering. The current pattern set carries a lot of substrate from prior decades — and that's worth saying out loud.

3 · Curator-space map

Every pattern plotted in base-curator × counter-curatorspace. X = CI score. Y = novelty rating. Click a point to see the pattern. The top-right quadrant is where both curators agree there's signal. The bottom-right is contested — base says signal, skeptic says rehash. If the cloud drifts toward the bottom-left over time, that's the slurry problem getting worse.

4 · Curator agreement

Cross-referencing the base curator's score with the counter-curator's verdict yields four buckets:

aligned+0(0.0%)aligned−0(0.0%)contested0(0.0%)neutral0(0.0%)
  • aligned+ both curators agree this is signal
  • aligned− both curators agree this is noise
  • contested the curators disagree — worth a human look
  • neutral neither strongly signal nor noise

7 · Outcome ledger

Every pipeline run snapshots each pattern's state. Over time this produces an empirical record — not a curator's opinion about what's signal, but what actually happened to each pattern: did its source count grow? Did its CI score decay? Did it survive to the next snapshot?

tracked

1166

growing

0

stable

0

decaying

0

died

0

Survival rate: 0.0% · Growth rate: 0.0% · Spanning 18 days from 2026-04-16 to 2026-05-04.

8 · What this tells us

  • The current pattern set carries heavy substrate from prior decades. Mean novelty is out of 1.0. That's not a failure — many of the most consequential shifts in AI are rediscoveries amplified by new compute and data. But it's honest to say so.
  • The slurry test catches exactly what critics pointed to. The patterns called out as generic-sounding really are. The test is working as a falsifiability gate — before publication, each label has to be distinguishable from what any LLM would say.
  • Contested patterns are the ones worth a human's attention.When the two curators disagree, neither automated score is trustworthy on its own. That's the honest moment — and exactly where human judgement earns its keep.
  • This page updates automatically.Every pipeline run recomputes slurry, counter-curator, and agreement. The metrics aren't cherry-picked snapshots. They're the real running measurement.

Commitments baked into this page:

  • The slurry count is shown even if we'd rather it were zero.
  • When counter-curator and base curator disagree, we don't silently drop either verdict.
  • Mean novelty is published regardless of whether it's flattering.
  • The counter-curator prompt, scoring rubric, and threshold are in public source.
Verg · self-audit · generated from the live convergence.json