Automate Build in Public: The 5 Levels of Content Automation
TL;DR
- Build-in-public automation is a spectrum, not a binary. Five levels map the range from "manual recall" to "fully agentic publication."
- Most solo founders should target Level 3 (AI-drafted, human-approved). Level 4 (auto-publish with override) gains 15% conversion at the cost of brand-equity risk.
- The level you can sustain is more important than the level you aspire to. Level 0 done consistently beats Level 4 abandoned at week 5.
Automation in build-in-public is the difference between a practice that survives past month 2 and a practice that collapses around week 5. This cluster sits inside our GitHub-to-content pillar and is the implementation companion to turn GitHub commits into tweets.
Why automation is the constraint
The single most predictable failure pattern: a founder commits to daily build-in-public posting, runs it manually for 2-4 weeks, hits a deep-work week or a low-energy stretch, misses 3 days, and never recovers the cadence. The practice does not die because the founder lacks discipline; it dies because the friction of manual posting exceeds the willingness to absorb the friction every single day.
Automation lowers the friction. Done correctly, the post lands in your approval queue within seconds of the work that justifies it — you approve in 60 seconds and the post ships. Done incorrectly, the automation produces spam that damages brand equity faster than no posting at all.
The five levels map the trade-off space.
Level 0 — Manual recall
Setup: None. You write posts when you remember to.
Outcome: ~10% conversion rate from content-worthy work to actually-shipped posts.
Where it works: For founders shipping infrequently who do not need a sustained cadence. For occasional retros and major launches where the work is its own reminder.
Where it fails: Daily/weekly cadence. The "remember to post" cognitive load gets reallocated to building within 2-3 weeks.
Level 1 — Manual reminder
Setup: Daily calendar block (e.g., 5pm: "write today's build post"). You open your editor at the scheduled time, review the day's commits, and write the post manually.
Outcome: ~30% conversion rate.
Where it works: Founders with strong reflection habits who want the journaling effect of writing posts. Pre-tooling baseline.
Where it fails: The "review the day's commits" step still costs 5-10 minutes of context reconstruction because the commits already aged out of the 30-minute window. Many of the day's most interesting commits do not get posts because the why is gone.
Level 2 — Triggered draft
Setup: GitHub webhook that fires on commits matching simple rules (feat:, fix:) and creates an empty placeholder in your notes (Notion, drafts folder, etc.). You write the post into the placeholder when you have time.
Outcome: ~50% conversion rate.
Where it works: Founders comfortable with light tooling who do not want AI in the loop. Better than Level 1 because the placeholder fires at the commit moment, capturing context.
Where it fails: You are still writing every post from scratch. The 5-10 minutes of writing time per post is too high to sustain at 4-7 posts per week alongside building. Drift back to Level 1 is common.
Level 3 — AI-drafted, human-approved (the sweet spot)
Setup: Webhook fires on commit, classifies through a signal-to-noise filter, packages the context (commit + diff summary + relevant prior posts), AI drafts a post in your voice, the draft lands in your approval queue. You approve in 30-60 seconds.
Outcome: ~70-80% conversion rate.
Where it works: The sweet spot for ~90% of solo founders. The friction is low enough to sustain past month 6; the human approval gate prevents the off-voice posts that damage brand.
Where it fails: Founders who cannot resist editing every draft heavily — the per-post time creeps from 60 seconds back to 5 minutes, killing the conversion advantage. The cultural shift is approving fast and trusting the draft for low-stakes posts.
Dev Cards + Loudy is the buildinpublic.so implementation of Level 3.
Level 4 — Auto-publish with human override
Setup: Same as Level 3 but posts schedule for publication on a delay (e.g., 30 minutes). You can override / reject within the delay window. If you do not, the post ships.
Outcome: ~95% conversion rate but ~3-5% of posts ship without thorough review.
Where it works: High-volume teams with multiple posts per day where the approval queue would otherwise back up. Larger accounts where the brand can absorb occasional off-voice posts because the surrounding volume is high enough to dilute any single mistake.
Where it fails: Solo founders building brand equity. One off-voice post that goes viral can damage trust built over months. The 15% extra conversion is not worth the 3-5% off-voice risk for small accounts.
Level 5 — Fully agentic publication
Setup: An agent decides what to post, when, on which channels, with what variations. The founder reviews aggregate analytics monthly but does not see individual drafts.
Outcome: ~99% conversion rate, but the posts are no longer authored by the founder in any meaningful sense.
Where it works: Brand accounts (not founder accounts) where the voice is institutional. Large publishers who treat content as a feed rather than as personal expression.
Where it fails: Founder accounts almost universally. The point of build-in-public is the founder's voice and judgment; removing both produces content that operators detect as automated and disengage from. Level 5 for a founder account is the same as Level 0 from the audience's perspective — neither feels like the founder is actually present.
The level you can sustain beats the level you aspire to
The most common mistake: trying to jump from Level 0 to Level 4 in a single move, abandoning the system within 30 days, and reverting to Level 0 with the additional confidence that "build-in-public does not work for me."
The healthier path:
- Start at Level 1 (calendar block). Run it for 2 weeks. Notice the conversion rate.
- Add Level 2 triggers if conversion is low. Run for another 2 weeks.
- Layer Level 3 tooling when you are confident the practice will stick. Most founders should land here permanently.
- Only consider Level 4 after 6+ months of Level 3 consistency, and only if your post volume is high enough that approval queue management is a real cost.
- Do not run Level 5 as a solo founder. It is not the voice the audience came for.
The brand equity argument
The reason this matters: build-in-public is a multi-year trust accumulation game. Each post adds or subtracts from the trust account. Voice-matched posts add; off-voice or templated posts subtract. The math favors approval-gated automation over auto-publish almost universally for solo founders because the cost of a subtraction is higher than the cost of a missed addition.
A founder running Level 3 for 12 months with 5 posts per week ships ~260 voice-matched posts. A founder running Level 4 for 12 months with 7 posts per week ships ~365 posts but with ~15-20 off-voice posts mixed in. The trust account at month 12 is typically higher for the Level 3 founder despite the lower volume.
Sibling clusters
- Turn GitHub commits into tweets — Level 3 operational deep dive
- GitHub to content — the pillar
- Commit messages as marketing — the source-side discipline
- Claude Code build-in-public skill — agent-native trigger patterns
- AI ghostwriter for developers — the drafting layer
FAQ
What is the right level for my stage? Pre-launch: Level 0 or 1 is fine — you are figuring out the voice, automation premature. Launch through month 3: Level 2 or 3 — the volume picks up. Month 3+: Level 3 is the sweet spot for most. Month 12+ with sustained cadence: consider Level 4 only if approval queue cost is real.
How do I know if I am at the wrong level? Too low: you are skipping posts because writing them takes too long; cadence is below 3/week. Too high: posts are shipping that you would have edited or rejected if you had seen them; voice drift across weeks; operator engagement declining month-over-month.
Can I use Level 3 tooling without paying for it? Yes — the components exist as separate open-source tools (GitHub webhook to your own service, custom inference call to OpenAI / Anthropic, approval queue in Notion). Integrating them yourself takes ~1-2 weekends. The buildinpublic.so stack collapses the integration work into a single onboarding flow, which is the value prop.
Will Level 3 automation make my content feel templated? Only if your voice rules are templated. If the AI Brain has captured your authentic voice (lowercase, specific, opinionated, no banned words), the output stays authentic. Templated feel comes from generic prompts to generic LLMs without voice persistence — exactly what AI Brain solves.
What is the right balance of automated vs hand-written posts? Empirically: ~70% automated (commit-driven daily posts, scheduled weekly demos), ~30% hand-written (monthly retros, longer threads, posts about something the automation cannot see like a conversation or a decision). The hand-written 30% is what keeps the account feeling human.
Building is no longer the bottleneck. Visibility is. buildinpublic.so is narrative infrastructure that runs inside your building workflow — Dev Cards implements Level 3 trigger + classification, Loudy implements Level 3 drafting, and Vibey implements Level 3 scheduling with the human approval gate preserved.