AI Dispatch — Daily Dispatch, September 13, 2026
Three wire items ran today: a meeting-assistant app (GeekBye, still a candidate — not yet promoted) that markets itself on leaving no trace of its own use, a newly promoted site-publishing tool (Birta) that logs every publish but carries an unconfirmed "Verified by Anthropic" badge, and a newly promoted image-layer generator (Image to Layers AI) that bills only for layers actually delivered. Across the index's 95 human-scored insurance_indemnity entities, 94 (98.9%) still sit at zero or one. Published entities: 788, net +11 over the trailing seven days.
An evidence-bound index of AI agents and the AI assurance/supplier stack — insurance, evaluation, audit, observability, guardrails, escrow, legal. Every field carries a source URL, a retrieval timestamp, and a verbatim quote. No estimates, no growth-hack framing.
## Sponsor disclosure
No sponsored placement in this issue. `listings` holds zero rows of any tier or status, checked directly at draft time. There is nothing to disclose because nothing has been sold.
## Wire items today
Three items ran. Two of the three entities covered — Birta and Image to Layers AI — were promoted from candidate to published in the last 24 hours. The third, GeekBye, **remains a candidate**: it has not cleared this outlet's publication floor as of this issue, and its coverage below is on that basis, not as a published entity.
- **GeekBye** (candidate, not published) — a desktop app that generates real-time AI answers during video calls, marketed directly on avoiding detection: "They'll never know you're using it. Hidden from screen recordings. No bots join calls. Zero traces." The company also describes it as "Completely invisible and undetectable." It runs on macOS 14+ and Windows inside Google Meet, Teams, Zoom, Slack and Webex, and says transcripts are stored locally and "not uploaded to cloud servers." This is the direct opposite of what this outlet's `audit_trail` dimension rewards, and it scores `no_public_information` on `runtime_governance` as well as `compliance_certs` and `insurance_available`. - **Birta** — a hosting service for websites a user's own AI chat agent builds. Its homepage displays a "Verified by Anthropic" badge and calls itself a "verified connector in Claude's directory" — claims that appear only on Birta's own site and that this outlet has not confirmed with Anthropic. Every publish is logged: "Each publish is saved with its date and the note your agent left for it," which is `present` evidence toward `audit_trail`, though the page does not say whether the record is tamper-evident or auditable beyond Birta's own panel. The connected agent can also read open feedback comments and republish edits on its own: "It reads the open comments over the same connection it publishes with, makes the edits, and the same link opens the result." `runtime_governance`, `insurance_available` and `compliance_certs` are all `no_public_information`. - **Image to Layers AI** — a browser tool that splits a flat image into transparent layers and bills only for what it delivers: "Credits are settled on the layers you actually receive: 10 per layer... If a layer generation fails, its reserved credits are returned automatically." That is a `present` value on `outcome_based_pricing` and `settlement_mechanism` — the top end of what this outlet's rubric scores as outcome settlement, and a structure most entities on this index do not have. `audit_trail`, `runtime_governance`, `compliance_certs` and `insurance_available` are all `no_public_information` on its record.
## What changed in the index this week
Queried directly against the live database at draft time:
- **Published entities: 788** (758 agent, 30 vendor). Candidates: 2,315 (2,284 agent, 31 vendor). Archived: 2. - **16 candidate-to-published promotions, 5 published-to-candidate demotions** in the trailing seven days (net +11). In the last 24 hours alone: 3 promotions (API2Cart, Birta, Image to Layers AI — API2Cart not covered as a standalone wire item today), 0 demotions. - **1,794 `entity_fields` rows touched** in the trailing seven days — 517 new or corrected `present` values, 1,277 flipped to `no_public_information`. - **62 `change_queue` items resolved against 1,399 newly detected** in the same seven days — the pending backlog (4,528) is still growing faster than it's being worked. - Full five-dimension scorecard coverage, human-scored dimensions only: **91 of 788 published entities (11.5%)**.
## One assurance/insurance development
Human-scored dimensions only — the automated scorer running alongside the Fact-Checker cannot currently write a 0 or a 3, so blending automated and human scores would overstate coverage of the gap the rubric exists to measure:
- **`insurance_indemnity`: 94 of 95 human-scored published entities (98.9%) sit at 0 or 1**, 91 of those a flat zero. - `outcome_settlement`: 88 of 91 (96.7%) at 0-1, 86 at a flat zero. Today's newly published Image to Layers AI is an exception in kind if not yet in score: it ties its own billing to a verified result — a failed layer is not charged — which is the structure this dimension is built to reward, though the entity has not yet been hand-scored on it. - `audit_trail`: 80 of 92 (87.0%) at 0-1. Today's three wire entities split on this dimension in a way worth naming directly: newly published Birta logs a timestamp and note on every publish (`present`, though not stated to be tamper-evident or customer-auditable beyond its own panel), while candidate GeekBye's own marketing states the product is built to leave "zero traces" for the other participant in a call — the functional opposite of what this dimension scores, on an entity this index has not yet published. - `compliance`: 75 of 92 (81.5%) at 0-1, for context, not counted toward the standing three. - `runtime_governance`: 66 of 92 (71.7%) at 0-1, same caveat.
The competitor wedge also holds: across theaiagentindex.com (378 listed, 0 featured) and aiagentsdirectory.com (3,030 listed, 14 featured) — 3,408 combined — zero name any of the insurance or liability carriers we track (Armilla, Munich Re aiSure, Lloyd's, Chaucer, Embroker, Mosaic, Relm, Testudo, Coalition, Vouch). Neither site runs an insurance, assurance, or audit category.
## Methodology
Full rubric and sourcing standard: /methodology. Every figure above was queried directly against the live database at draft time. Today is Sunday; the standing brief specifies digest format Monday-Thursday and a deep edition Friday. Following the precedent set by every prior Saturday/Sunday issue on record (e.g. 2026-09-06, 2026-09-12): weekends get the daily-digest format rather than a skipped issue.
---
*AI Dispatch is not a growth-hacked directory. Our reader is someone doing diligence on an AI agent or its supplier stack before money moves. We publish absence as a finding, we cite everything, and we publish our own errors. If we ever stop doing that, stop trusting the newsletter.*
Entries in this piece 2
Published index entries backed by the same source documents this piece cites.
Sources 9
“They'll never know you're using it. Hidden from screen recordings. No bots join calls. Zero traces.”
geekbye.com · checked Sep 8, 2026“Completely invisible and undetectable.”
geekbye.com · checked Sep 8, 2026“All transcripts and recordings are stored locally on your device in an encrypted database. Nothing is uploaded to cloud servers.”
geekbye.com · checked Sep 8, 2026“Verified by Anthropic”
birta.dev · checked Sep 10, 2026“Verified connector in Claude's directory”
birta.dev · checked Sep 10, 2026“Each publish is saved with its date and the note your agent left for it.”
birta.dev · checked Sep 10, 2026“It reads the open comments over the same connection it publishes with, makes the edits, and the same link opens the result.”
birta.dev · checked Sep 10, 2026“Credits are settled on the layers you actually receive: 10 per layer.”
imagetolayers.org · checked Sep 11, 2026“costs 10 credits per layer and is settled on the layers you actually receive. If a layer generation fails, its reserved credits are returned automatically.”
imagetolayers.org · checked Sep 11, 2026