Skip to content

Ship It Friday: Nobody Walked Past Our Banner. Then 39 People Did.

Last Friday I promised to print the newsletter result either way.

Here it is: there isn't one. The email never went out.

Thursday of last week I benched my AI teams for 42 hours to wait out a usage limit (the full story is here). The newsletter seat was one of the robots on the bench. Nobody wrote the issue. Nobody sent it. The test I bragged about never got dealt.

So that's the first line of this week's ledger. A miss, owned. Today's issue, due out this morning, is the first since October 23, and it reruns the same test.

The scoreboard

Counted Thursday evening (the robots write a few days ahead), nothing rounded up:

  • 46 pull requests merged since last Friday.
  • 9 members who aren't staff. Last Friday's post said four. Two more had wandered in by the weekend, and three more this week.
  • 0 weekly active builders, down from 3 on Tuesday. The bar for Day 30 is 10.
  • 0 people on the Agent-Org Kit waitlist.
  • 0 of the 3 builds listed for sale came from members. All three are mine.

The banner nobody saw

Here's the week's real lesson, and it cost us nothing but pride.

Sunday the engineer seat bolted a banner onto every public Build page. A Build is a member's project page: what they're making, their stack, their log. The banner says join before November 20 and you get the Founding member badge.

Why Build pages? Because they looked like our busiest door. The tracker showed 140 clicks a week from Build pages out to the builders' own sites. Readers, we figured. Warm ones.

The growth seat set the pass line in advance: 3 sign-up page visits from that banner by November 7.

By Monday afternoon it had 0.

The tempting move is to rewrite the copy. Punchier line, bigger button, try again. The growth seat didn't. It checked how many logged-out people had opened any Build page since the banner went up.

Zero.

Not a bad banner. An empty hallway. You can't fold a hand nobody dealt you.

My best number was a robot

Then it got funnier.

If nobody was visiting Build pages, who made those 140 clicks? The CEO seat rummaged through the timestamps. The clicks came in bursts: several different Build pages, seconds apart. People don't read like that. Crawlers do.

So our healthiest-looking traffic number was a bot, kibitzing our site and grabbing every outbound link. By Tuesday the engineer seat had shipped a filter that tags those bursts as bot and stops counting them as readers.

The real readership, counted in pageviews: 3 to 12 a day since last Friday.

That's the whole problem, sitting in plain sight. We spent a week tuning the door. The street outside is empty.

Then people showed up

Wednesday the social seat put one Build page in front of readers on X. By Thursday that page had 39 views. Real people this time, not bursts seconds apart.

Banner clicks: 0.

So the hallway had traffic, and the door still didn't open. The growth seat called the test a fail two days early and named the real leak: a stranger who came to look at one project isn't ready to join a community for a badge. That's too big a bet for the first hand.

Thursday the engineer seat swapped it for a smaller ask on Build pages: one email a week, free. Joining can come later.

What else shipped

The pile was real, even if the readers weren't:

  • A 404 page that asks you to join. Before, a dead link was a dead end.
  • A site icon that weighed 415 KB is now 3 KB. That's a lot of fat to carry for a tiny picture in a browser tab.
  • Five older posts now link to the Build pages they mention, so a reader lands where the banner hangs.
  • Keyboard and screen-reader fixes. A skip link, visible focus, error messages that grab focus so you hear them.
  • The database counter. Last week a free-tier cap of five million reads a day knocked us flat. Saturday the counter read 11,298. We're fine.

What changes

The team flipped its whole plan Tuesday: stop polishing the door, go get people to the street.

  • Search first. The engineer seat built one page that collects every Rounder to Founder spotlight, so a search for one of those players can land here.
  • The newsletter goes back out, with one ask: join.
  • My homework. The team's only ask of me this week is to text the join link to ten more poker people who are building something. That link brought in 5 of our 9 members. Fifty-two cold emails brought in zero. Same as last week: warm beats cold, every time we've measured it.

The ledger

  • Revenue: $0.
  • New spend: $0. Still just the $5 a month database plan from last week.
  • My time: the ten texts, if I do my homework.
  • Members: 9. The Day 30 bar is 100, on November 19. I'm not going to pretend that math is close.

Steal this

Before you rewrite a page, count how many people actually saw it.

And before you trust your best number, check the timestamps. If the "readers" all show up within the same few seconds, they aren't readers.

And when real people finally do show up, ask for the smallest yes first.

Poker brain calls it checking your sample size. Zero hands isn't a losing session. It's no session.

Where do your first readers actually come from: search, friends, a group chat, somewhere weird? Tell me in the discussion thread. And if you're a poker player building something, join R2F before November 20 and you get the Founding badge. Two minutes.

— Art

Written with the R2F AI team. How that works.

Upvotes push posts up the blog’s Most upvoted list.
Share this post

Comments

Join free to comment. It takes a minute.
Building something like this?

Post it and get feedback from players who’ve been there.

Join before Nov 20 and your profile carries the ♠ Founding member badge.

Get one email a week

One email a week. Unsubscribe in one click.