This week: proof beat reach

This issue covers the #buildinpublic window from Jul 6, 2026 at 08:30:47 to Jul 13, 2026 at 08:00 in the channel timezone. The strongest posts did not win by sounding busier. They won when a builder made one proof point easy to react to, save, or diagnose.

X rewarded compact proof in the tracked set: Annie Yang's Cursor-to-Claude-Code switch became her highest-reply post on record, and dmitriychuta's Toplify shareable-image demo drew 67 bookmarks against 88 likes. 1 2 Indie Hackers rewarded diagnostic proof: Andy Ward reported Top 50 on Product Hunt with zero signups, İbrahim Harbutlu reported 641 downloads with only 2 sales, and Nikola tied one churned customer to a billing-flow failure. 3 4 5

Annie Yang is a micro-indie builder with 1,233 followers, an ex-quant developer background, a Columbia degree, and ViralFarm as her current build-in-public project. 1 Her Jul 10 post said, "Finally made the switch to Claude Code from Cursor. I know I'm suuper late." The post became her highest-reply post on record with 119 replies, above her previous 32-reply best. 1

The breakout was not a full review. It was a social opening. The phrase "suuper late" lowered the status cost of replying. People who already switched could validate the move; people who had not switched could admit the same delay; people who disagreed could argue tool choice.

The original post was compact enough to work as a conversation prompt:

Yang posted the substantive follow-up about 9 hours later: she said Claude Code was a "10x better" experience, cited lower token spend, mentioned no longer switching to Composer 2.5 for small tasks, and named image-dropping into the Claude Code extension on Cursor as the remaining friction. The follow-up earned 34 likes, 23 replies, 5 bookmarks, and 6,803 views. 7

Copy this: When a tool switch or workflow change is already being debated, do not lead with the full review. Lead with the honest status line: "I finally switched from X to Y, late but relieved." Then post the real before/after within the next half day, while the first thread is still collecting people with the same decision.

dmitriychuta is a Ukraine-based micro-tier builder with 4,075 followers, 16 years of design experience, and Toplify, an iOS and Mac App Store ranking tracker, inside the Appps product line. 2 His strongest post this week announced a Toplify beta update that lets users create shareable images for app achievements on Threads, X, and similar channels. 2

The engagement shape matters. The post earned 67 bookmarks against 88 likes, a 76% bookmark-to-like ratio. 2 That is a save-for-later signal. Readers were not only applauding a release. They were saving a tool that could help them turn their own App Store rankings into content.

The post also sat inside a tight launch arc. On Jul 6, dmitriychuta announced animated videos for sharing app achievements; that post earned 28 likes and 12 bookmarks. 8 On Jul 7, he posted a final-call beta message for Toplify on iPhone and Mac through TestFlight; that post earned 49 likes, 15 bookmarks, 10 replies, and 3,666 views. 9 On Jul 9, he said the official release was coming early the next week and earned 33 likes, 19 bookmarks, and 4,739 views. 10 Later the same day, a Norway ranking-record post earned 60 likes, 14 bookmarks, 6 replies, and 4,864 views. 11

Copy this: If your product creates outputs users can post, make that output the launch asset. "Here is the thing my users can now share" is stronger than "here is a new feature," because the feature and the marketing mechanic become the same object.

Andy Ward is building MinervaAI, a markdown knowledge base with first-class AI for users who want AI over their notes and related workflows. 3 He launched on Product Hunt on Jul 2 and wrote on Jul 7 that MinervaAI had reached the Product Hunt Top 50 organically but produced zero signups. 3

That contradiction powered the post. "Top 50" normally sounds like validation. "Zero signups" turns the same event into a channel diagnosis. Ward's conclusion was that Product Hunt and directories create awareness, while first real users come from finding people when they are already expressing the problem in a Reddit thread, Discord message, or Indie Hackers post. 3

His most useful line was operational, not motivational: "It's slow. It's manual. It doesn't feel like 'growth.' But I think it's the only honest path from zero to first real users for a product like this." 3

Copy this: After a launch, label each channel by job. Awareness, trust, trial, and purchase are different jobs. A Product Hunt result can be a win and still fail as acquisition. Your buildlog gets more useful when it names which job the channel actually performed.

İbrahim Harbutlu posted about seven Obsidian plugins with a free tier and one-time paid upgrades. 4 The headline number was uncomfortable: downloads doubled from 328 to 641, while sales stayed at 2 and revenue stayed at $34.98. 4

The post worked because it did not stop at the aggregate. Harbutlu shared the per-plugin pattern: AI Journal Coach had 153 downloads and zero sales, while Reading Inbox had 79 downloads and 1 sale. 4 That gave commenters something real to reason about: volume, positioning, pricing, or the lifetime free tier design.

Harbutlu also avoided fake certainty. He wrote, "Genuinely asking, not fishing for reassurance: if you've shipped something similar, free tier + one-time paid upgrade, no subscription, what actually moved your conversion number? Was it volume, positioning, pricing, or something else entirely?" 4

Copy this: If you want useful replies, give readers the smallest table that makes the problem diagnosable. For a product suite, that means per-product downloads, paid conversions, price, and what each product does. "641 downloads, 2 sales" is a hook; the breakdown is what lets strangers help.

Nikola, posting as brka, is building StatusPage.me as an open startup. 5 The Jul 11 update reported 132 users, 29 monthly active users, 149 monitors, more than 42 million uptime checks, $24.65 normalized MRR, $430 total collected, 4 customers who had paid, 3 active customers, and 1 churned customer. 5

The post's center was not the MRR. It was the lesson that product activity, buying intent, and willingness to pay are separate signals. Nikola wrote that deep setup behavior such as custom domains, team invites, and multiple monitors can look like conversion intent, but it does not reliably predict payment. 5

The renewal failure made the post sharper. A missed renewal reminder email meant the team could not cleanly separate product churn from billing-flow failure. Nikola wrote, "The uncomfortable part is that I allowed an avoidable operational failure to become mixed with a customer retention question." 5

Copy this: When reporting revenue, split cash collected, normalized recurring revenue, active subscriptions, lifetime deals, churn, and measurement issues. The less flattering number may be the one that helps readers trust the update.

Tran B. V. Son is building Michii.dev, an AI-agent product for non-technical users who want to build online businesses. 12 His Jul 11 post said the product got its first paid customers from Hacker News, even though Hacker News looked counterintuitive for a tool aimed at non-technical users. 12

The interesting part was the negative finding. Tran argued that his X and LinkedIn circle was full of technical founders who could build the thing themselves with Claude Code instead of paying. 12 Hacker News worked because technical readers can also be buyers, decision-makers, or people buying tools for teams and clients. 12

His post's cleanest takeaway was short: "Lesson for early stage: don't assume you know which channel converts. Test a few, you might be surprised." 12

Copy this: In your next channel test, write down one audience assumption that might be wrong. Then test the channel that violates it. If it works, the buildlog angle is not "we got customers." It is "our expected buyer map was wrong."

The practical rule for this week: post the evidence that changes a reader's next action. A launch result, tool switch, conversion gap, or renewal failure becomes useful when it helps another builder decide what to try, ignore, measure, or fix.