Here's a thought experiment from the enforcement side. Two accounts each post twelve comments this week, same communities, same quality, same disclosure. Account A: twelve comments spread across seven days, mornings and evenings, a quiet Sunday. Account B: twelve comments between 9:14 and 10:41 on Tuesday, fifteen minutes apart, then silence.

Same words. One of these accounts is getting rate limited this month, and it isn't A.

Almost everything founders are taught about Reddit is about content: write value, disclose, don't spam links. All true, all covered earlier in this series. But the layer nobody manages deliberately is when, and the platform weighs when heavily, because timing is the one signal content can't fake. Spam looks like a schedule. People look like tides.

Why pace outranks content

Reddit's enforcement (post 2 covered the layers) has an easier job than reading your prose: it can read your intervals. Burst posting, metronome regularity, and round-the-clock activity are patterns that predict spam far more reliably than any single comment's text, and they're computable at platform scale for free.

The specific shapes that read as machine:

Bursts: many actions in a tight window, then nothing. The classic founder-on-a-lunch-break pattern is indistinguishable, at the timestamp level, from a bot that woke up.
Metronomes: comments at suspiciously even intervals. Fifteen minutes apart, twelve times, is a schedule confessing itself.
No nights: a human account has quiet hours. An account active across all 24 reads as either a bot or a team, and communities treat both the same way.
Volume cliffs: an account that did 2 comments a week for a month suddenly doing 15 a day is announcing that something changed, and the filter's guess about what changed is not charitable.

The bitter part: every one of these patterns is natural founder behavior. You batch your marketing hour, you catch up at midnight, you sprint when a launch lands. Reddit's filters read your calendar honesty as automation.

The story-spam counterexample, steelmanned

Before the principles, the strongest objection deserves its hearing, because there's a playbook that appears to refute all of this and every founder eventually sees it work.

The playbook: take one founder story, rewrite it ten ways, and post it relentlessly across every startup community until the name sticks. We've watched companies run this. Honestly: it works, short-term, on its own metric. Everyone in those communities did remember them. Repetition plus receipts plus shamelessness is a real distribution strategy, and pretending otherwise would be dishonest.

Now the ledger's other side. The same communities that remembered them got tired of them, said so publicly, and the reputation followed the brand outward: threads about them in adjacent communities open with the spam reputation before the product gets discussed, and every future post from anyone at that company starts in a hole. Reddit's memory is long, searchable, and it compounds in both directions. They bought recall and paid in trust, in a channel whose entire value is that surviving content is trusted.

The honest read on the trade: if you need recall this quarter and never intend to need the community's trust again, the spam playbook is available and it functions. For a founder building a company that intends to exist in three years, it's selling the moat to buy a billboard. This series is for the second founder.

The principles of a human pace

The fix isn't a magic number, it's a handful of principles any schedule can satisfy:

1. Set a daily ceiling per account, and keep it boring. A handful of quality comments a day, sustained, beats fifteen on Friday. The ceiling that never varies dramatically is itself the signal of a person with a life. When you feel the urge to do more today, that's tomorrow's budget arriving early.

2. Space actions irregularly. Not evenly (metronome), not all at once (burst). Real gaps: twenty minutes, two hours, forty minutes. If you batch your Reddit time, batch the reading and drafting, then spread the sending.

3. Keep human hours. Pick the waking hours of your actual timezone and stay inside them. An account that never sleeps is one of the cheapest tells the platform has, and quiet hours cost you nothing: the threads will still be there at 7am, and post 7 covers why the morning sweep is fine.

4. Ramp changes gradually. Coming back from vacation, launching, hiring someone to help: whatever the reason for more volume, phase it in over weeks. Volume cliffs get reviewed; slopes don't.

5. One account, really. The tempting workaround (spread the volume across three accounts) is the worst of all options: coordinated accounts are the pattern Reddit's sitewide layer (and its ban-evasion detection) exists specifically to catch, and it converts a community-level mistake into a platform-level one. One account, seasoned properly, at a human pace, is the entire sustainable configuration. Operators who run the disposable version anyway (buy aged accounts, accept bans as routine, keep spares warming) say the quiet part themselves: a ban isn't bad luck on that road, it's the eventual outcome, and it takes the account's whole compounding history with it.

6. Treat pushback as a circuit breaker, not a speed bump. Post 2 said it and it bears repeating at the cadence layer: a rate limit means days of backing off, not minutes of waiting. The accounts that die are the ones that resume full speed the moment the timer clears.

One more piece of testimony worth more than any theory: an operator who documented his automation attempts described trying VPNs, private proxies, and every combination of them, and watching every account die within days of its first automated post. His conclusion matches everything this post has said: the IP is the least of what the platform reads. Browser fingerprints, the timing of actions, what the account does either side of posting, how many accounts have touched the same setup. He kept believing he was one configuration away, and he never was. Content quality was never the variable; the shape of the behavior was.

The uncomfortable summary

Content mistakes get your comment removed. Cadence mistakes get your account removed, and account removal is the expensive one: it takes the seasoning weeks (post 3), the community standing, and the entire compounding history with it.

This is also, candidly, the layer where doing it by hand is hardest. Nobody's judgment fails at "write a helpful comment." Everybody's judgment fails at "it's 11pm and I have twenty minutes, I'll just knock out eight replies." Pacing is a discipline problem, and discipline problems are what schedules and systems are for, whether the system is a calendar rule you actually obey or software that enforces the ceiling for you.

Thread Otter treats pace as a hard constraint, not advice: per-account daily ceilings, irregular spacing between sends, quiet hours in your timezone, gradual ramps for young accounts, and automatic multi-day backoff the moment a platform pushes back. The judgment about what to say stays with you; the discipline about when is enforced. However you implement it, implement it: cadence is the ban vector, and it's entirely controllable.

This is post 6 of the 8-part Reddit for founders series. Next up: listening, triage, and the 90-minute window where the first useful comment wins.