How to build in public on X and LinkedIn without wasting months (2026)

The short answer

Building in public works when your posts reach the people who'd buy your product and you turn their attention into something you own, like an email list or a community. It doesn't work as a follower-growth hobby. Most accounts stay small, and the famous examples you've seen are survivorship bias: you hear about the handful who made it, not the many who posted daily for a year and got nothing.

So treat it as a sales channel with a scorecard. Post where your buyers already read, share things that are useful to them, send every post back to a signup, and count signups per post. If that number stays at zero after a couple of months, change the audience or the format, not the posting frequency.

This guide covers what X and LinkedIn say about their own feeds, what to share and what never to share, a weekly cadence you can keep up, and how to measure it. It sits alongside Getting your first customers without ads and Where to launch a startup.

Why most build-in-public accounts stay small

Most accounts stay small because attention on social platforms is concentrated in a few accounts. It's worth knowing the base rates before you commit months to it.

• In Pew Research Center's 2019 study of 2,791 US adult Twitter users, the most active 10% of tweeters wrote 80% of all tweets. The median user in the other 90% had 19 followers and tweeted about twice a month.
• Even the median user in that top 10% had 387 followers, and wrote 138 tweets a month.
• Pew's 2021 follow-up found the most active 25% produced 97% of tweets between 12 June and 12 September 2021. That group averaged 65 tweets a month and received an average of just 37 likes and one retweet a month.

These are US adults on Twitter in 2019 and 2021, not founders on X in 2026, so don't treat them as a forecast. They do show the shape: lots of posting, little reach, for most people.

Survivorship bias is the other half. The founders held up as proof that building in public works are the ones whose products took off, and their audience often grew because the product did, not the other way round. You never see the matching set of founders who posted the same updates and quietly shut down. Be wary of any claim that building in public brings "three times more customers" or similar: we couldn't find a primary study behind those multipliers, so we don't repeat them.

None of this means don't do it. It means set an honest goal (signups and conversations from buyers) and judge it on that.

What the platforms say about their own feeds

The platforms have published more than most "algorithm hacks" posts admit, and it's better to read the source than a summary of a summary. Here's what each one says, with dates.

X: the open-sourced For You code

X's current For You ranking code is public in the xai-org/x-algorithm repository on GitHub. Its README, checked on 26 October 2026, says:

• The feed blends in-network posts (recent posts from accounts you follow) with out-of-network posts found by ML retrieval, cluster similarity and a shared list of popular posts.
• A model called Phoenix predicts how likely the viewer is to take each action on a post: like, reply, repost, quote, share, clicks (including link and profile clicks), dwell time and follow, plus negative actions such as "not interested", mute, block and report. Those predictions are combined into one weighted score.
• After scoring, posts after an author's first get a decaying multiplier (author diversity), posts from accounts the viewer doesn't follow are multiplied by a factor below 1, and posts from authors whose impressions are below a threshold are "lifted toward a target position".
• Filters remove posts older than 48 hours, posts already seen, and replies and reposts from accounts the viewer doesn't follow.
• The README added notes on 13 and 14 August and 18 September 2026, including a correction that weights scale each viewer's *predicted probability* of an action, not raw like or report counts.

What that means for you: your followers see you first, strangers see you at a discount, each post has a short window, and a post that makes people stop, reply or click gets scored higher. And replying to a big account won't put you in its followers' For You feeds: the code filters out replies from accounts the viewer doesn't follow. The README lists link clicks as a predicted action and doesn't describe a link penalty, so ignore confident claims either way unless they cite the code.

The older twitter/the-algorithm repository is still up and describes the earlier system (Home Mixer, a heavy ranker neural network, SimClusters). Its README links to X's 31 March 2023 engineering post, which now returns an error on X's own site, so we read the Internet Archive copy. That post said each request pulled about 1,500 candidate tweets, the For You feed was 50% in-network and 50% out-of-network on average, and a roughly 48 million parameter network ranked them for positive engagement such as likes, retweets and replies. Treat those 2023 figures as history: the 2026 repository describes a different, transformer-based ranker.

LinkedIn: engineering posts and help pages

LinkedIn describes its feed in plain words on its own pages:

• Its engineering post Engineering the next generation of LinkedIn's Feed (12 March 2026) says the feed serves more than 1.3 billion professionals and that "the most valuable content is timely, relevant to their professional goals, and grounded in trust". Its ranking model reads more than a thousand of a member's past interactions, including long dwells, likes, comments and shares, and relies on "professional signals and engagement patterns, never demographic attributes".
• A November 2025 post, Putting members first, says LinkedIn tests whether any group of creators is systematically ranked lower, and that different engagement on similar posts doesn't by itself mean bias.
• An older post, Community-focused Feed optimization (25 June 2019), says the feed takes into account "timely feedback to content creators", because creators care about prompt responses.
• The help page Distribution of your content on LinkedIn says your posts go to a subset of your connections and followers, based on connection strength and their notification settings. Its tips: write content your audience can relate to, cover lessons learned and industry insight, "avoid being overly promotional", complete and verify your profile, and react to and comment on other people's posts.
• LinkedIn relevance groups the signals into identity (your profile), content (how often a post is viewed or engaged with, and whether it comes from someone you follow or are connected to) and activity (what you react to, comment on and share).

Bluesky and others

Bluesky's blog post Algorithmic Choice with Custom Feeds (27 July 2023) says it offers "an open marketplace of feeds" that users choose between, so there's no single algorithm to write for. We didn't find a Threads page that says anything specific enough to include.

How the two main platforms compare

Make the point an owned channel

The point of building in public is to move people from a feed you rent to a list you own. A platform can change the rules or your account at any time. X's Terms of Service (effective 9 October 2026; we read them through the r.jina.ai text reader because x.com blocks direct fetching) say X may "limit distribution or visibility of any Content" and, to the extent the law allows, terminate an account "for any other reason or no reason". Neither platform's feed is something you control.

An email list or a community is. Every post should give a reason to join: the weekly update in full, the template you mentioned, early access, or the numbers you don't post publicly. Keep the signup one click from the post, on a page that says exactly what people get and how often.

Get consent the right way. In Australia, the ACMA says you must have consent before sending marketing emails, keep a record of who gave it, when and how, identify yourself, and honour unsubscribe requests within 5 working days. You can't send an electronic message to ask for consent, because that message is itself marketing. Other countries have their own rules, so check where your readers are.

What to share and what never to share

Share what helps your buyers and what you'd be fine seeing on the front page of a newspaper. Keep anything about customers, unverified money claims and other people's secrets out.

Customer data and privacy

Treat anything that identifies a person as off limits unless they've agreed. The UK's Information Commissioner's Office says personal data is any information relating to an identifiable person, including a name, an identification number, location data or an online identifier. A screenshot of your Stripe or CRM dashboard usually contains several.

In Australia, the OAIC's guidance on APP 6 says an entity can only use or disclose personal information for the purpose it was collected, unless an exception applies, such as the person's consent. It counts as disclosure when you make information accessible to others outside your business, and even when the recipient already knows it. The Privacy Act generally doesn't cover a small business with annual turnover of A$3 million or less, but there are exceptions (health service providers and businesses trading in personal information, among others), and your customers' trust doesn't come with a turnover threshold. Blur names, emails, avatars and company names before you post, and ask in writing before you quote a customer.

Revenue screenshots and income claims

Post only revenue you can prove, and think twice before you post it at all. Revenue screenshots are easy to fake and easy to misread (gross vs net, one good month, refunds not shown), and readers know it.

The legal risk grows if you later sell a course, coaching or a "how I did it" product. The US Federal Trade Commission's Business Opportunity Rule guide says it's illegal to make an earnings claim for a business opportunity unless you have written materials that back it up, and an earnings claim statement must give the dates the earnings were achieved and the number and percentage of buyers who got at least that result. In October 2021 the FTC sent a notice to more than 1,100 businesses, including business coaches, warning that false or unsubstantiated earnings claims, such as saying represented profits are typical, could bring civil penalties of up to US$43,792 per violation at the time. If your revenue posts are part of a sales funnel for advice on making money, get proper advice first.

Competitors' and other people's confidential information

Don't post what isn't yours to share. That covers a competitor's pricing or roadmap you heard privately, details from a former employer, a partner's terms, or what a prospect told you on a call. It's a fast way to lose trust with the exact buyers you're trying to reach.

A simple weekly cadence

A steady rhythm of four useful posts a week is enough, and each one should point to your email capture. X's code filters For You candidates older than 48 hours, so one post a week has a short life; LinkedIn's 2019 post notes it values timely feedback to creators, so reply to comments the same day.

Write for your buyers, not for other founders. A post about "my MRR journey" mostly reaches founders; a post about the problem your product solves, told through what you learned building it, reaches customers. Repurpose the same week's material across both platforms if your buyers are on both, but don't post the same thing twice on one.

Reply to every comment early, and spend ten minutes a day commenting usefully on posts your buyers read. LinkedIn's own help page lists this as a way to build reach, and X's code weighs predicted replies.

If you're launching on a set date, plan the build-in-public posts as the lead-up and use Where to launch a startup for the launch week itself.

Measure signups, not followers

The number that matters is signups (and the conversations they lead to) from each post. Followers and likes are inputs, not results.

Tag every link you post with UTM parameters. Google Analytics' own guide says to always use utm_source, utm_medium and utm_campaign, uses utm_content to tell apart two links in the same place, and warns that values are case sensitive (google and Google count as different sources), so it recommends lowercase. A link might look like:

https://yoursite.com/?utm_source=linkedin&utm_medium=social&utm_campaign=bip-2026-w44&utm_content=lesson-pricing

Then mark your signup as a key event in Google Analytics, using the steps in Google's key events help page, so reports show which posts led to signups. Each week, log posts, signups and any sales conversations in a simple sheet. After four to eight weeks, drop the formats that bring no signups and do more of the ones that do.

Some people will sign up later without clicking your tagged link, so add a "Where did you hear about us?" field to the signup form. It catches what UTM tags miss.

Common mistakes

• Writing for other founders. Founders like and repost each other, which feels like traction, but they rarely buy. Write for your customer.
• Chasing followers. A bigger account with no signups is a hobby. Track signups per post from week one.
• No owned channel. If the platform limits your reach or your account, the audience goes with it. Send people to a list.
• Posting customer screenshots. Blur every name, email and logo, and ask before quoting anyone.
• Unverified revenue posts. If you can't back a number with records, don't post it. If you sell advice, the FTC's earnings-claim rules apply.
• Believing algorithm folklore. Read the X repository and LinkedIn's own pages before changing your posting around a rumour.
• Copying survivors. The famous examples had a product people wanted. Building in public amplifies demand; it rarely creates it.
• Bursts then silence. X's For You code ignores posts older than 48 hours. A steady weekly rhythm beats a big week and a quiet month.

Your build-in-public checklist

1. Write down who your buyer is and which platform they read most.
2. Set one goal: signups (or sales conversations) per month from posts, with a review date eight to twelve weeks out.
3. Set up an email list or community with a clear promise, and a one-click signup page.
4. Make sure every signup is consent-based, records when and how consent was given, and has a working unsubscribe.
5. Add UTM tags to every link and mark the signup as a key event in Google Analytics.
6. Add "Where did you hear about us?" to the signup form.
7. Decide your publish-or-not list: which numbers you'll share, and which you won't.
8. Blur personal data in every screenshot and get written permission before naming or quoting a customer.
9. Post four times a week: a progress update, a number, a lesson and an ask, each pointing to the list.
10. Reply to comments the same day and comment usefully on your buyers' posts.
11. Review signups per post each week and keep only the formats that bring signups.
12. At the review date, decide plainly whether to keep going, switch platform or put the time into another channel from Getting your first customers without ads.

If you'd like other founders to sanity-check your plan, it's free to request an invite to Remote Builders, which is invite-based and runs builder dinners in Bali and Australia. See also Communities for remote and bootstrapped founders compared.