The most useless advice I ever got, and I got it constantly, was "just be yourself." Here's the lesson this essay exists to deliver: social convention is a protocol, the protocol is undocumented, and if you weren't born with a working implementation, the answer isn't intuition — it's reverse engineering. You can extract the rules, write them down, and run them mechanically. I know because I eventually had to.
Advice from people whose defaults match the room
"Just be yourself" is advice from people whose factory settings happen to match the environment. It's the API designer telling you the interface is intuitive. Of course it's intuitive — to them. They never had to read documentation that doesn't exist, because they are the documentation, compiled at birth and patched continuously through a thousand playground interactions that installed correctly.
For the rest of us, every room runs a protocol with real, enforceable rules — about pacing, about how long a message gets to be, about when correcting someone is helpful and when it's an attack, about how much enthusiasm is charming and at what point it becomes suspect. None of it is written down anywhere. The rules differ between rooms that look identical. They ship breaking changes without a changelog. And violations don't return an error message — they return silence, distance, and eventually a disconnect, usually without a single line of diagnostics you could use to fix the bug.
I spent years filing that under "people are irrational." That was wrong, and it was also the exact kind of wrong that keeps you stuck. The protocol isn't irrational. It's just proprietary.
What reverse engineering it actually costs
Here's what nobody tells you about learning social rules by trial and error: the sample efficiency is terrible and every failed experiment costs you something real.
I knew some of the rules abstractly for years — you can read every book on this and I probably did. It didn't transfer. Thinking about it privately was not enough, because I didn't have enough examples for my brain to build the pattern. Some things you genuinely cannot learn from description; you need volume, labeled data, lived instances — and every training example is an actual interaction with an actual person where getting it wrong has actual consequences. Imagine debugging a protocol where every failed handshake burns a peer that may never reconnect. That's the tuition. I paid a lot of it, and I want to be honest that some of those bills don't get refunded no matter what you learn later.
The other cost is subtler: when you finally do push through and act freely — and it goes fine — nobody updates their model of you. And the first time it doesn't go fine, everyone treats that as confirmation of what they suspected all along. The protocol's enforcement is asymmetric: compliance earns you nothing visible, and violations are remembered. That asymmetry is why "keep trying things and see what works" is such an expensive learning algorithm here, compared to literally anywhere else I've applied it.
So if trial and error is brutally priced and intuition isn't installable, what's left?
Mechanical rules beat intuition you don't have
At some point I stopped trying to acquire the thing other people have — that continuous, ambient, real-time read of the room — and asked a different question, the one I'd ask about any system I can't see inside: what's the minimal ruleset that produces acceptable output without requiring internal state I don't possess?
That reframe changed everything, for one specific reason: intuition has to run in real time, in the moment, under load — exactly when I don't have it. A rule runs before the moment. You write it once, calmly, with full information, and then the 11 PM version of you with something urgent to say doesn't get a vote. It just follows the protocol. I am extremely good at following protocols once they're written down. The entire problem was that nobody had written them down. So I wrote them down.
People hear this and think it sounds robotic, like giving up on authenticity. It's the opposite. The rules don't govern what I think, what I care about, or what I say when someone actually wants the depth — they govern packet size and timing, nothing else. Same person, same signal, delivered in a form that survives contact with a receiver. What's actually robotic is spending every interaction white-knuckling a self-monitoring loop that fails exactly when it matters. I've done both. The rules are freer.
The ruleset
These are mine — the real ones, extracted from my own failure data, not adapted from a book. Yours will differ, but the shape is the point: concrete, checkable, and executable without any social intuition at all.
1. The four-sentence rule. Anything longer than four sentences doesn't get sent to a person — it gets written up somewhere permanent, and the person gets a link and one line. Rationale: past four sentences, a message stops being communication and becomes an assignment. Publishing converts the same content from an imposition into an offer.
2. One message, then silence. After I send something, I send nothing until they reply. No follow-up thought, no "also—", no clarification of the thing I just said. If it matters, it survives until they answer. Rationale: intention cannot fix a pacing problem, because pacing failures happen precisely when intention is most inflamed. Only a rule that doesn't consult my intention can.
3. The overnight rule. Anything written while upset gets saved, not sent, and reread the next day. No exceptions clause, because every exception I ever granted myself was the message I most needed this rule for. Rationale: the delete key costs nothing in the morning and everything at night.
4. Two weeks read-only. In any new community, I say nothing for two weeks. I'm not lurking — I'm documenting their protocol: pacing, humor, what gets corrected and by whom, what thanks looks like. Rationale: every room runs a different undocumented spec, and I respect protocols once they're documented. So the first contribution I make to any room is documentation, even if I'm the only one who ever reads it.
Notice what these rules are not. They're not "be less enthusiastic" or "seem more normal" — vibe-goals you can't verify and will always feel like failing. Each one is binary. Did the message exceed four sentences? Did I wait? Anyone can audit their own compliance, which means the rules actually run, which is more than intuition ever did for me.
The transferable part isn't my ruleset. It's the permission slip: if the room's protocol was never documented for you, you're allowed to treat that as the engineering problem it is. Collect your own failure data — the specific moments things went sideways, not the general feeling of being wrong. Look for the mechanism, not the moral. Write the smallest rule that would have prevented it. Run the rule until it's boring.
Boring is the goal. The protocol was never going to be intuitive for you. It can still be solved — and solved beats intuitive, because solved is written down, and things that are written down can be kept, improved, and handed to the next person who was told to just be themselves and left to guess what that means.