Evidence log: November 2026

A founder sending 25 a day is far below the bulk threshold, so only the all-sender rules bind; setting up DMARC anyway costs nothing and is what Yahoo and Outlook expect of bulk senders. The slow ramp in the send step is what Google itself advises.

The cap is not what limits a founder at 25 a day; reputation is. The step’s slow ramp is about a new domain’s standing, which no provider publishes a number for.

Reworded 2026-11-11: the sequence step now requires a postal address and an opt-out in every email to US recipients.

Reworded 2026-11-11: the rosters step now says UK sole traders need consent first, so call or write.

Reworded 2026-11-11: the DM step now says to connect without a note and send the message once they accept.

Ashby, also named by the step, was not checked in this batch.

Supports the direction of “intent decays by the hour”, not a number for email replies; the steps should not quote 21 times as if it were about email.

The sequence step now uses this talk’s under 75 words (it said 90 before 2026-11-11).

Reworded 2026-11-11 to name EU or UK traffic and the opt-in rule; the step said EU only before.

Reworded 2026-11-11: the step now states the 14-day right and the express waiver for instant access, and labels the conversion line as our view.

Plan text corrected 2026-11-10: it had said trial-to-paid, and implied retention was the same for every paywall.

Plan text corrected 2026-11-10: it had credited the figures to Cal AI.

Plan text corrected 2026-11-10, twice: the first version had 30-50% of involuntary churn with no primary source; the second had a 55% that the page does not carry. It now quotes the totals the page does state.

Do not state a budget-change percentage; Meta gives none.

The phrase first screen is our paraphrase of toward the top of the page. The above-average rating scale is not on this page and is not claimed.

Plan text corrected 2026-11-10: the 2.8x had no method behind it.

Plan text corrected 2026-11-10: it had said about 15%.

Plan text corrected 2026-11-10: the 2.7x had no source; other posts give 2.3x and 4.2x.

Plan text corrected 2026-11-10: it had called the registry canonical and said PulseMCP and Glama read it; the docs say official, in preview, and do not say Glama reads it.

Plan text updated 2026-11-10: the 2026-09-02 check read $347 for the newsletter bundle and the page now says $437, so the price has moved; the $49 scope is ambiguous.

Plan text corrected 2026-11-10: it had stated $497, a sold-out $247 tier and 400k monthly professionals as fact.

Plan text corrected 2026-11-10: it had said G2-owned since 2026 as settled fact.

Plan text corrected 2026-11-10: it had said 30-50% of tool research happens in an assistant that never sends a click. No study gives that; this is the closest real figure.

Plan text corrected 2026-11-10: it had said the six links are contractual; at most they are advertised.

Plan text corrected 2026-11-10: it had said it crossed $100k/month; the interview gives $125k over 28 days and about $80K MRR.

Plan text corrected 2026-11-10: it had said a hunter is worth roughly 15% of day-one momentum.

Plan text corrected 2026-11-10: it had said feedback requests convert about 3x better and avoid the filter.

Plan text corrected 2026-11-10: it had said the 2026 algorithm weights comment quality and responsiveness over raw upvotes, and that one comment is worth fifty upvotes.

Plan text corrected 2026-11-10: it had expected roughly 50% conversion from 20 asks.

The step already says llms.txt is cheap insurance and not a ranking lever; this is the source for that sentence.

Opened both posts: 10.7% against 2.1% (called trial-to-paid in the benchmarks post, download-to-paid in the other), and 27% against 28% one-year retention of yearly subscribers.

Opened the page: trials of 17-32 days convert 70% better than 3-day trials, 42.5% against 25.5%. The comparison group is 3-day trials here, not under four days, and the claim was corrected.

Opened the case study: 123 experiments across 46 trigger points, 160 paywall designs, 424 variants and a 31% figure are on the page.

The page does not state a 55% recovery rate (the only 55% is a sample tax rate). It says $8.2B in failed payments were recovered in 2025 and $14 for every $1 spent. The claim was rewritten to that.

Opened the page: campaigns appear only after at least 24 hours since launch, and a metric shows once it reaches 5. The sentence about a first download within 24 hours of a click could not be extracted by the browser tool; two earlier readers reported it.

Opened the page: it says you can filter by UTM source and campaign. The condition that UTM source needs the ads and referrals filter was not found in its text, so it was dropped from the claim.

Opened the report: 87%, 64%, 68.7% and 23.9%, 900 queries and 12,000 URLs are all on the page.

Opened the English Ahrefs post: it says the integration pages drove 16% of the entire organic traffic.

The pricing page rendered only a few hundred characters in Chrome, with no prices. The figures rest on two earlier readers, who disagreed on what the $49 covers.

The homepage opened but no longer shows the 700,000+ figure an earlier reading recorded, and the listing pages are still 404. No price could be confirmed.

The G2 page says it formally agreed to acquire the three platforms and is dated 2026-01-29. The SEC page returned almost no text, so the completion date could not be re-checked directly.

Opened the report: half of B2B software buyers (51%) start research with an AI chatbot more often than Google, against 29% in the earlier report, from 1,076 buyers.

The submit page opened in Chrome and advertises a dofollow backlink and a listing that lasts forever. No price or turnaround time appeared in its text, so $99 and 48 hours rest on third-party reviews.

Opened the form: it says Matt will review the submission and shows no fee or sponsored option.

Opened the FAQ: there is no advantage to having someone else post it, and who posts does not affect ranking.

Opened the DEV post: a table row gives the hunter contribution to upvotes as about 15-25% from a 30-launch sample, with no method.

Opened the FAQ: it says to invite feedback and never to ask for upvotes.

Opened both pages: how-it-works lists points, comments and time since launch; the FAQ says upvotes from established, active members earn far more points. Nothing on maker responsiveness.

Opened the page: review requests through Review Automation see a 40% open rate and an 11% conversion rate, with no method.

Opened the page: there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary. It does not name llms.txt. The statement was reworded to the page’s own phrasing.