Build-in-Public Week 4: Channel-Pivot Pilot Verdict — +196v Catalog Lift, +182v Quora 30d, and an Honest First Read on LinkedIn-First

#buildinpublic #indiehackers #marketing #android

Originally published on LinkedIn, May 30 2026, as a native long-form post — this dev.to canonical lands at +7d per the reverse-channel-order rule. X long-post ran June 1 (+48h). Same content asset, three surfaces, inverted order.

TL;DR: The channel-pivot rule (Quora-first → X long-post → dev.to canonical at +7d) just finished its second full 7-day evaluation window at +196 catalog views (+60.3%) — more than double the +92v (+39.5%) it produced in window one. Quora 30-day views added +182 (+19.8%) and crossed the 1K monthly milestone. The secondary experiment — publishing Build-in-Public to LinkedIn first — produced honest, smaller numbers: 8 impressions on the LinkedIn native post, 1 view on the X long-post. The deciding read is the Day-1 view count on this very article. Build-in-public means publishing the verdict either way.

For eight weeks, my dev.to catalog was a museum exhibit. Articles got published, decent ones earned a slow trickle of views from search, and the 1-day post-publish window — the part of dev.to traffic that's supposed to matter — was almost always a flat zero. I was reading the analytics weekly, and the weekly story was: "no measurable change."

Then I changed the publish order. Not the writing, not the topics, not the platforms. Just the order.

This is the two-window verdict on that change, plus the secondary experiment that this article is itself the test instrument for.

The pilot in one paragraph

I'd been writing dev.to as the canonical surface. The article published first, then a Twitter thread, then a Quora answer, then LinkedIn — pillar to derivatives, the standard waterfall. The Week-3 scorecard compared two real numbers from that approach: on Quora, our "Spying" answer climbed from 135 views to 447 views in 24 hours. On the matched-content dev.to publish, Day-1 traffic was zero. Same words. Same week. Wildly different surfaces.

So I flipped it. Starting 2026-05-19, the rule became: Quora first, X long-post at +72h, dev.to canonical at +7d. Pillar content lives on the surface where its audience is already searching. The waterfall reverses. The dev.to canonical becomes the durable backstop, not the launchpad.

Window one (5/19–5/24): catalog +92 views (+39.5%) — the first material lift since the feed-pickup throttle started. Window two (5/24–5/31): +196 views (+60.3%), from 325 to 521 — the largest 7-day catalog lift in our scorecard era, and a clean doubling of window one's record. Two-week rolling: +288v dev.to (+88.6%) and +568v Quora 30d (+106.8%), both starting from a flat 8-week pre-rule baseline. As of this writing the catalog sits at 614 cumulative views — up from 253 on May 19, +142% in 18 days.

That's the pilot. Now the details.

What "+196 views" actually means after eight flat weeks

A throttled dev.to account looks like this: your existing catalog keeps slow-trickling views from search (good), but any new publish hits a wall on Day 1. We confirmed it across thirty articles in early May. Three consecutive newsjacks hit identical 0–2 view ceilings inside 96 hours of publish, while 4-day-old pieces in the catalog continued to accumulate views normally. The pickup mechanism had quietly shut off for new submissions on this account.

The conventional fix is "wait it out and reduce publish cadence." We did that for a week. The throttle didn't lift. The new theory came out of looking at where our content actually moved people: not at dev.to discovery, but at Quora's intent-graph. A question like "What information does a free Android security camera app collect about my home?" has the exact shape of intent we want — anxious, specific, ready to act — and Quora routes those questions to answer-writers who have published in the topic. The audience is already on the page. The reader doesn't need a feed to find us; the question itself is the feed.

The pilot moved the pillar to where the audience already was. The dev.to article stopped being the publish event and became the destination link people followed after they'd already engaged with the content somewhere else. The throttle didn't lift — but the throttle stopped mattering, because the new traffic was external referrals onto an existing catalog, and dev.to's referral-traffic surface is not throttled the same way feed-pickup is.

Window two's composition tells the story even better than the total. The +196 concentrated in three places: a newsjack on the Kimwolf botnet bust (id 3747395) that climbed 0 → 32 views across its first seven days; the Texas/Meta newsjack adding +20 more before saturating at 52v; and a canonical (Six Signals) breaking 0 → 20 on Day 4. Every one of those was co-published with a same-week Quora pillar on the matching anxiety question. The pieces that weren't paired stayed flat. Cross-channel pairing is the variable. It's now confirmed across two consecutive windows with monotonically increasing magnitude.

What "+182v Quora 30d" actually means

Quora was a side surface in our marketing plan. The account had 16 answers and 64 monthly views as of the 2026-05-10 scorecard. Then a single answer on the "Is my phone spying on me?" question shape moved 135 → 447 views in a 24-hour window, and we rebuilt the production line around what it taught us.

Window two: 30-day content views moved 918 → 1,100 (+182, +19.8%) — crossing the 1K monthly milestone — on the largest single-week publish volume ever (+17 net-new answers, 30 → 47). That's roughly 23 views per answer on the catalog average, far above any other channel we operate. The viral "Spying" answer crossed 1,000 views on its own and is still compounding.

The shape that works, every time: the question begins with "Is [thing happening] to me?" or "How do I know if [thing] is doing [bad behavior]?" The reader is anxious. The reader does not want a polished comparison list. The reader wants the diagnostic ladder — a step-by-step way to check the thing for themselves, starting with the lowest-effort test. Our app comes in at the end as the option whose architecture removes the question entirely, not as the first-paragraph plug.

That structure works on Quora specifically because Quora's recommender treats answer length plus question-specificity as signal. Comparison lists are good for search-engine acquisition; diagnostic ladders are good for Quora's internal recommender. We pivoted the writing to match the surface.

The Texas/Meta newsjack as a control variable

A newsjack is a special case of pillar content: the topical urgency is the discovery vector. We published id 3723597 on 2026-05-22 in reaction to Texas AG's investigation into Meta AI Glasses (the "always-enabled mode + facial-geometry collection" complaint). The article used a three-architectural-questions frame: opt-in vs. always-on, local vs. cloud, on-device vs. off-device inference. Our app is a worked example on each.

That article did something no dev.to publish on this account had done since the throttle began: it broke double digits on Day 1. It compounded to 52 views by Day 7 and holds there today, still the catalog's best performer.

The Texas/Meta NJ is meaningful as a control because we did not change the dev.to-side publishing mechanics for it. Same API, same tags, same byline. The variable that moved was the cross-channel pairing: this NJ ran alongside a Quora pillar the same week, and the Quora pillar referenced the dev.to article inline. External Quora traffic landed on the dev.to canonical via the cross-link. Then the Kimwolf NJ repeated the pattern (0 → 32v) under the same pairing rule, beating Texas/Meta at the same days-since-publish offset. n=3 NJs now confirm it: cross-channel pairing, not topic narrowness, is what clears the throttle. The pairing rule is now a hard gate in the content engine — no newsjack ships without a same-week Quora companion on the matching frame.

The LinkedIn-first experiment: honest numbers, verdict pending

The pilot above is the headline finding. The secondary experiment is the one this article exists to measure — and the early numbers deserve to be reported as plainly as the good ones.

The hypothesis came from Week 3: when we cross-posted that week's Build-in-Public to LinkedIn on a 4-follower account, the post earned 34 impressions. On dev.to, the same content, published the same day as the canonical, earned 0 views on Day 1. That ratio said: for the build-in-public format specifically, LinkedIn is the carrier and dev.to is the archive. So Week 4 — the piece you're reading — reversed the channel order. LinkedIn published first (Friday May 30, native long-form). The X long-post followed June 1 at +48h. This dev.to canonical lands today, June 6, at +7d.

Here are the measured numbers, pulled this morning before publish:

LinkedIn native post (5/30): 8 impressions. Lower than Week 3's 34 — going native-first did not, by itself, buy more LinkedIn reach on a 5-follower account.
X long-post (6/01): 1 view. The long-post format continues to underperform on our X account, where reply-driven engagement has always beaten timeline posts.
dev.to canonical (this article): publishing now. The Helm-locked decision criterion: if this piece records ≥1 view on Day 1 — something Week 3's canonical could not do — the LinkedIn-first runway hypothesis survives. If it lands at 0, the rule is falsified and the E-archetype channel mix gets rethought.

Two honest observations while the verdict is pending. First, the LinkedIn-side lift the hypothesis predicted has not materialized — 8 impressions is a real number on a 5-follower account, but it is not a runway. Second, even if Day 1 comes in at zero, the experiment cost nothing: the content was written once and published three times, and the dev.to canonical was always going to be the backstop, not the carrier. That asymmetry — cheap experiments with one-directional downside — is the only reason a solo developer can afford to run a pilot like this at all.

The peer-dialogue compound: from comments to disclosure

One quiet result from the window deserves its own section, because it produced the catalog's first comments — a column that had read zero since launch.

We've been leaving thoughtful, non-promotional comments on third-party dev.to posts in our topic area for six weeks. On 2026-05-18, a comment on a peer's end-to-end-encryption article landed on the AI-snippets surface within 24 hours — faster than our own articles index. On 2026-05-22, a third party (@privacyfish, from the privacy.fish mail service) joined the same thread, replying to our reply. And on 2026-05-31, a developer we'd been in a multi-turn dialogue with (FuriousOfNight) disclosed his own unreleased desktop camera project in a reply on our Week-3 build-in-public thread — the first verbatim author-disclosure the comment graph has produced, and the reason the Week-3 canonical now carries 2 comments.

Dev.to's article-level signals (reactions, comments, follows) are noisy, but the thread shape itself is the kind of signal that algorithms pick up — depth of replies, breadth of participants, time-to-second-response. We're maintaining six waiting-for-turn-back peer-dialogue nodes in the comment graph. It's the slowest channel we run and the only one that produces relationships instead of views.

What we're keeping, what we're changing

Keeping

The channel-pivot rule for content with anxiety-question shape: Quora-first, X long-post at +72h, dev.to canonical at +7d. Two windows of evidence, +288v dev.to and +568v Quora on the two-week rolling read.
The diagnostic-ladder writing shape for Quora pillars: low-effort tests first, high-effort tests last, app as architectural existence proof at the end.
The newsjack pairing gate: no NJ publishes without a same-week Quora companion on the matching frame. n=3 confirmations.

Changing

Publish-cadence watch: window two shipped 8 net-new dev.to pieces in 7 days — the same regime that triggered the original feed-pickup throttle. A deliberate gap day is now scheduled weekly to keep the rolling cadence at or under 1.2 articles/day.
Quora wave 3 pivots from privacy-anxiety to subscription-shrinkage framing — the dominant competitive moment right now is camera-app vendors squeezing free tiers (AlfredCamera's 2-cam/24h watermark squeeze, Arlo Secure's $4.99 → $7.99 hike, Eufy's per-camera cloud-fee creep). Same question shape, new anxiety.
UTM tagging moves to the daemon level — a five-scorecard gap between deciding to do it and doing it consistently. Still fixing.

Open questions

Does this canonical land above zero on Day 1? (The LinkedIn-first verdict — this article is the instrument. Answer lands 2026-06-07; the full +7d read on 2026-06-13.)
Does the diagnostic-ladder shape carry from privacy-anxiety to cost-anxiety questions, or was it topic-specific? (Wave-3 batch is the test, running now.)
Does the canonical-cohort 10-view plateau (three pieces stuck at exactly 10v for days) represent a real ceiling or a view-counter artifact? (Watching through 6/08.)

Where the app actually is

Numbers as of this morning's API pull:

dev.to: 62 published articles, 614 cumulative views, 4 reactions, 2 comments. Up from 253 views on 2026-05-19 — +361v / +142% in 18 days, the largest sustained lift since launch.
Quora: 47 answers, 1,100 monthly content views (up from 64 in early May), 11 followers. The highest-yield organic surface in the system.
Google Play: live on production. The app is privacy-first and local-only by architecture, with an embedded Ktor server for LAN streaming and optional YouTube Live for opt-in cloud streaming.
The system: 15 scheduled tasks across content generation, distribution, analytics, strategy, and community pulse — built on Claude Code, the same way the app itself was built across 75+ AI-assisted sessions.

If you're building something similar — a solo product with a marketing problem that has to be solved without a team — the lesson from these two windows is the one I keep relearning: the surface determines the format, and the format determines the order. We optimized two of those three for eight weeks before fixing the third. The third was the one that mattered.

The app: Background Camera RemoteStream on Google Play

The website: superfunicular.com

Cross-links for further reading:

Originally published on LinkedIn May 30, 2026. X long-post June 1. dev.to canonical June 6. Same content, three surfaces, inverted order — and this article is the measuring instrument for whether that order works.