okaihub

Building in Public for AI Products: What to Share and What Not To

A practical guide to building in public as an AI founder: what to post, how often, which numbers to share, and where the approach backfires.

Published 2026-10-21 · 3 min read

Building in public works for AI products and it fails for AI products, usually for the same reason: founders share the wrong layer. Metrics and milestones are the least interesting part. The decisions, the things that did not work, and the specific detail of how the product behaves are what people read.

This is a guide to the practice, including where it backfires. It is not universally good advice, and treating it as such is why some founders spend months posting to nobody.

What is worth sharing

Share thisWhy it worksExample
A decision and its reasoningIt is transferable — the reader can apply it even if they disagreeWhy we capped the free tier by cost per query rather than by seats
Something that failedRare, credible, and specific enough to be usefulThe onboarding step we removed and what happened to activation
A concrete technical detailDemonstrates the product without claiming anything about itHow we cut streaming latency, with the before and after
What you learned from usersOriginal information nobody else hasThe phrase three users independently used to describe the problem
A real number with its contextNumbers alone are noise; numbers with a cause are a lessonSignups doubled after we changed the first sentence, and why we think so
The common thread: each of these is information the reader could not get elsewhere.

What is not worth sharing

  • Bare milestone posts. "We hit 100 users" is not content. It becomes content when you say what produced them and what you would change.
  • Revenue screenshots without context. They read as either a brag or a plea, and both reduce trust. If the number is relevant to a lesson, share the lesson and the number together.
  • Daily activity logs. "Today I refactored the auth module" has no audience. The version that does is "here is what was wrong with our auth and what we learned".
  • Engagement-bait questions. Asking your followers what to build is fine once, as a genuine question. As a recurring format it becomes obvious.
  • Anything you would not say to a customer's face. Building in public has a long memory, and future customers read it.

The cadence question

There is no correct frequency, but there is a correct relationship between effort and consistency. Two substantive posts a week that you can sustain for six months beats daily posting that stops in three weeks — the audience forms around the handle that is still there.

A workable pattern: one longer piece every week or two about a decision or a lesson, and a handful of short observations in between. The short posts are cheap and they keep the handle visible; the long one is what people share.

Where it backfires

  • It becomes a substitute for selling. Posting feels like progress and is more comfortable than individual outreach. If your user count is flat and your follower count is rising, that is the diagnosis.
  • It attracts other founders. An audience of builders is pleasant and does not buy B2B software. If most of your engagement comes from people making similar products, you are writing for peers rather than customers.
  • It dates your product. Public pivots, abandoned features and revised pricing are all permanent. This is manageable but worth knowing before you post a roadmap.
  • It creates a support channel you did not staff. Public problems get public replies. If you cannot answer them, silence is visible.

How it connects to the rest of your distribution

Building in public is not a channel on its own — it is the raw material for the others. A decision you wrote about becomes a paragraph in your product description. A failure you described becomes a post in a community where someone has the same problem. A technical detail becomes the thing an AI assistant cites when someone asks how your category works.

That is the version of this that compounds. The version that does not is posting into a void and waiting, because an audience built on personality alone is slow to form and does not transfer to a product.

More in Content & building in public