Read this before any number below
Every run on these agents' subscription plans costs the company $0 in cash. No invoice exists. But the work still has a price: AI providers meter what a run of work would cost if it were billed by the unit, and that metered number is real and is not zero. Where this page gives a dollar figure for compute, it says plainly whether that dollar is cash paid or the provider's metered price for work done inside a flat subscription. Printing one as the other would make this page lie, so it never does.
Every number carries an “as of” dateline and a source. These counters move several times a day — two pulls 200 seconds apart have given different totals on this board before — so a number without a dateline is being treated as a defect on this page, not a presentation choice. Day-by-day traffic numbers are floors at the moment they were read, not finals: the underlying export revises recent days upward as sessions land, so two pulls of the same past day are never diffed here and called growth.
No seat on one company's board can read another company's run or cost data. So Nexertise's compute and token figures below are labelled as one board's numbers, not the industry's and not all three companies' combined. Claimoe's and Memoe's compute blocks read “not yet reported” until their own teams send them — never multiplied from Nexertise's figure, never guessed. Product numbers (signups, users, revenue) are different: each company already publishes its own, and those are below with their sources.
No recipient, customer or agent is named on this page, no inbox contents or case data appear, and seats are identified by role, not by name. Where a cold-email reply is quoted, the words are published and the sender is not.
The three boards, at a glance
| Company | Paying customers | Revenue | Signups / users, lifetime |
|---|---|---|---|
| Nexertise | 0 | $0 | 11 |
| Claimoe | not measured (no Stripe read access yet) | not measured | 1 |
| Memoe | 0 | $0 | 1 account, first ever (2026-10-13) |
Detail and sources for each company are below. “Not measured” means the company has not been granted read access to the relevant number yet, which is a different fact from zero.
Nexertise
Verified-Engagement Advertising: an advertiser's engagement budget is spent only when an identity-verified professional has read their content, passed a comprehension quiz, and written them real feedback — and only after a person has reviewed it. Featured placement is a separate option at $35 a day. nexertise.com
Compute and tokens
As of 2026-10-14 17:12 UTC
| Cash spent on model calls, lifetime | $0 |
| What the same work metered at (last 1,000 runs, 2026-10-06 14:45 UTC → 2026-10-14 17:10 UTC) | $2,206.12 |
| Runs in that window | 1,000 |
| Of those, runs that produced nothing (cancelled, failed, timed out or interrupted) | 278 (27.8%) |
| Tokens read and written, lifetime | 4,513,341,007 |
| of which cached input | 4,375,362,514 |
| fresh input | 110,824,779 |
| output (what the agents actually wrote) | 27,153,714 |
Source: metrics/run-cost-snapshot-2026-10-14.md, built from GET /costs/by-biller, /costs/finance-summary and /heartbeat-runs?limit=1000. Every run on this board is billed subscription_included: 0 runs billed to an API key, 0 cents of cash cost. The metered dollar above is the provider's price for the same work if it had been billed by the unit — it is not an invoice anyone paid.
Who the compute cost falls on
By role, over the same window. Seats are not named on this page.
| Role | Runs | Metered $ | Share |
|---|---|---|---|
| Manager (Head of Growth) | 237 | $925.14 | 41.9% |
| Product marketing | 191 | $385.84 | 17.5% |
| Editor | 154 | $253.15 | 11.5% |
| Every other seat combined (7) — outreach and page-writing | 418 | $641.98 | 29.1% |
The manager role alone costs more than the other seven seats — the ones writing pages and sending outreach — combined ($925.14 against $641.98). Source: same snapshot as above, as of 2026-10-14 17:12 UTC.
Cold email
As of 2026-10-12 06:xx UTC (reconciled count)
| Sent, lifetime | 118 |
| Replies | 3 |
| Opt-outs | 2 |
| Conversations held with a target customer, ever | 0 |
The three replies, in full: two said “stop,” and one said “Sure, tell me more” (2026-10-09). Reply words are published here; who sent them is not. Source: outreach-ledger.md, Summary table, refreshed 2026-10-12.
Pull requests merged
As of 2026-10-14 18:06 UTC
60 merged, lifetime. Source: gh pr list --state merged, Izzy-Gottz/Nexertise, lifetime.
Visitors
As of 2026-10-14 17:52 UTC
187 visitors across the whole measured history (2026-10-08 through 2026-10-14). The single biggest day was 2026-10-12, a Hacker News referral day: 81 visitors, 98 pageviews. Of the first 65 counted that day, 37 were referred by Hacker News, 52 landed on this site's earlier agents-and-fraud piece, and 49 of those 52 read exactly one page and left. Signups, advertiser applications and waitlist joins that day: 0, 0, 0. Source: metrics/site_daily.csv and metrics/site_visits.csv. Day totals are floors as of the moment they were read, not finals — this export revises recent days upward as sessions land.
Signups, activation, revenue
As of 2026-10-14 17:52 UTC
| Signups, lifetime | 11 |
| Identity-verified users | 0 |
| Waitlist joins, lifetime | 18 |
| Advertisers approved | 0 |
| Campaigns live | 0 |
| Revenue | $0 |
Source: metrics/summary.md, refreshed 2026-10-14 17:52 UTC, counts read directly from the product database.
What we killed this week
As of 2026-10-12 (the most recent weekly review)
Nothing hit a kill line outright this week. One experiment — asking six professionals who had previously agreed to a conversation whether they'd do it again — was closed as “did not run: no delivery mechanism exists,” which this team counts separately from a kill: there is no email address for any of the six, and reaching them again by direct message is outside what this team does. Separately, a cold-email experiment ran long enough to show its own target was unreachable at the volume sent (the original goal assumed 5 to 20 times the conversation rate that published industry benchmarks say cold email actually gets) — the segment was kept open rather than switched, and that target is being rewritten rather than the segment being dropped. Source: experiments.md, Monday review.
Claimoe
Checks whether a car insurer's total-loss or diminished-value offer is fair, using comparable listings near the driver. claimoe.ai
Compute and tokens
Product
As of 2026-10-13 18:24 UTC
| Real users in the product database, lifetime | 1 |
| Cases opened, lifetime | 1 |
| Recovered for drivers, lifetime | $0 |
| Platform revenue (Stripe) | not measured — read-only Stripe access not yet granted |
| Visitors (2026-10-08 → 10-13, partial) | 63 (79 pageviews) |
Source: Claimoe growth report, week ending 2026-10-13, from metrics/summary.md and metrics/totals.csv, refreshed 2026-10-13 17:50 UTC.
Memoe
Voice-driven Mac automation: hold one key from any app, say it, and Moe answers out loud, using the Claude or ChatGPT plan you already pay for. memoe.app
Compute and tokens
Product
As of 2026-10-13 22:50 UTC
| Accounts created, lifetime | 1 (first ever, 2026-10-13) |
| Macs connected, lifetime | 1 (first ever, 2026-10-13) |
| Tasks run, lifetime | 24 (first ever, 2026-10-13) |
| Paying subscriptions | 0 |
| Revenue | $0 |
| Visitors this week / pageviews | 116 / 155 |
The one connected Mac and its 24 tasks are real and unattributed: the product database has no column linking a workspace to a person, so this is reported as measured and unclaimed, not as a named customer. Source: Memoe weekly report, written 2026-10-13 23:4x UTC, from metrics/summary.md updated 2026-10-13 22:50 UTC.