Account bootstrap (day 0)

The SOPs (x-account-sop, reddit-sop etc.) describe a running machine. This is the checklist for standing an account up — the layer from-sop-to-operation §5 only itemized. It stops where the SOP's steady loop begins (first-10-artifacts-for-learning-not-reach is the SOP's job, industrialization-needs-a-mechanical-arbiter).

The thesis — there is no "warming up"

The single most load-bearing finding, checked across all five arenas: no platform documents a "warmup" ritual, an account-age gate, a probation clock, or a numeric trust score that unlocks after N days. The "drip activity for two weeks before you post a link / age the account / 养号 seven days" advice is folklore — and it traces, on four of five platforms, straight to vendors selling proxies and anti-detect browsers who profit from making new accounts sound fragile. Xiaohongshu operators report the opposite: new accounts get a cold-start 流量扶持 (traffic boost), and content quality dominates account age.

What is real is behavior-based anti-spam that disproportionately catches new accounts because that is where the bots are. So the day-0 job is not "age the account" — it is three things:

1. Don't look like a bot (§1) — the genuine kernel under all the warmup superstition.
2. Lock the irreversible fields correctly, once (§2) — several day-0 choices can never be undone.
3. Wire measurement before the first post (§4) — or the SOP's experiment loop starts blind.

Everything else on day 0 is positioning, which the SOP's Phase 0 already owns (audience=buyers, one reprintable format, faceless-vs-persona — x-account-sop Phase 0).

§1 — Don't look like a bot (the real kernel)

The behaviors that trip documented anti-spam on day 0, all arenas:

• Mass follows / connects / likes in a short window. LinkedIn's documented invite restriction fires on volume + ignored-invite ratio + automation-suspicion (not on age — official, [help/a551012]); Reddit's low-karma spam filter throttles new+low-karma accounts and earning any karma clears it, Premium does not bypass (official — but the popular "~1 action / 10 min" number is not published, only the mechanism); X caps follows at 400/day and posts at 2,400/day and bans "bulk, aggressive, or indiscriminate" following (official, help.x.com/x-limits + /x-automation).
• Immediate outbound links. No documented "links unlock on day 14." Link deprioritization for non-Premium accounts is not in X's Help Center — the only official acknowledgment is Musk's own post ("put the link in the reply… stops lazy linking"), framed generally, not as a non-Premium rule. What is documented verbatim: the X API charges $0.200 to create a post containing a URL vs $0.015 without — ~13× — a link penalty priced into the judge itself (docs.x.com/x-api pricing).
• Datacenter / VPN IP or a geo-mismatch at signup. A stable, residential-consistent egress reduces verification challenges — the legitimate residue of the "keep a consistent IP" folklore. (Note: the specific "X rejects VoIP numbers" claim is not in any official X page — it's third-party; X's Help Center is silent on VoIP acceptance.)
• Bulk-identical posts across accounts (coordinated inauthentic behavior) — this, not mere plurality of accounts, is what gets penalized.

§2 — Irreversible day-0 choices (lock these once, correctly)

Several handle/identity fields cannot be changed or change only under a hard cooldown. Get them right before the first post.

The two genuine traps for this column's stack: Reddit's username is permanent, and Xiaohongshu's RED ID changes exactly once ever — both are decisions, not defaults.

§3 — The account-path decision (per arena, once)

• X / short-video: personal account is the default; faceless-format is viable and a persona is a standing cost (x-account-sop Phase 0.3). No business-account decision blocks day 0.
• Reddit: the account is not the brand — the operator is a participating person, and the founder discloses rather than hides the affiliation (outsider-founder-earns-voice, reddit-industrialize-and-die). No "company account."
• Xiaohongshu: the real fork. 企业号 (蓝V; official fee 600 RMB per review, valid 1 year, annual 年审 required, non-refundable on fail; needs a 营业执照; one subject certifies up to 20 employee accounts — the "1 license = 2 accounts" figure is folklore) vs personal (posts fine, 实名 lets it sell/live-stream at smaller scale). Overseas/港澳台 entities must certify on PC (not the app), upload a color scan of the registration document with company seal, and a stamped 申请公函 (physical 鲜章 + handwritten signature — e-seals rejected) whose name matches the docs; non-Chinese/English materials need a chopped 翻译件. No deposit is in the official cert docs — the agency-blog "20k–50k RMB 保证金" figures are unverified. Default personal until commerce is actually needed. See xiaohongshu-redlines.
• The operating tell (Xiaohongshu, non-resident): IP属地 displays the operator's country on the profile AND every comment, with no off-switch ("不支持手动开启或关闭", overseas shown to country; official 2022-04-15). It is derived from the connecting device's carrier IP (not GPS — disabling location does nothing), lags ~6–24h after a sustained IP change, and every "change your 属地" method is VPN/proxy spoofing (folklore + ToS risk). This is the single biggest structural tell of operating from abroad — design around it, don't try to hide it.

§4 — Wire measurement before the first post

from-sop-to-operation §5: the experiment loop is blind if measurement is hooked up after posts exist. Day-0, before artifact #1:

• Branded-search tracking and referrer/UTM tagging on any owned destination — so the being-seen signal is attributable (being-seen-reliably).
• The per-post metric fields the SOP declares as its {metrics} (context-derives-from-the-sop) exist and are writable — this is what the utter-manual diary lints against on day 1.

§5 — Credentials & multi-account hygiene

The genuinely documented risk is account linking, not weak passwords: a phone/email/device reused across accounts is the cleanest cross-account link the platforms have, and reusing a suspended account's number/email/device inherits the suspension (X locks a deactivated account's number + email for 30 days, official; Reddit's Content Policy bans ban-evasion — this is Rule 2 "participate authentically", not Rule 4 which is about minors — and links accounts by unspecified "account signals", not an officially-named IP/fingerprint mechanism; Xiaohongshu binds one phone to one account and offers only 换绑, not 解绑 — in-app-enforced, no public doc).

• One unique real email per account (not name+1@, name+2@ aliases — trivially correlated).
• No phone-number reuse across accounts on the same platform. Carrier SIM, not VoIP, where a challenge can fire.
• App-based TOTP (time-based one-time password) 2FA over SMS everywhere it's offered — SMS ties account security to a number you may need to isolate per account, and is SIM-swap-exposed.
• Credentials live outside the product repo (utter-manual AGENT contract), never committed.

§6 — The superstition ledger (actively discount)

Folklore with no platform documentation on any arena — do not build the bootstrap around any of these: the multi-week warmup calendar; a fixed probation clock or numeric trust score; karma/follower-by-day curves; aged-account value; bio keyword-stuffing as a recommendation-reach lever (the narrower search/matching effect is real — LinkedIn headline, Xiaohongshu nickname/bio, X people-search bio, Instagram Name field — but YouTube states its tags field "plays a minimal role" in discovery and that recommendations are ultimately decided by per-video viewer engagement, not metadata; note the honest limit — titles/descriptions do feed YouTube's search-relevance matching, so it's tag-stuffing specifically that's dead, not all metadata); precise reach multipliers ("10×/3×"); Xiaohongshu 养号 session quotas; LinkedIn's ~100/week invite cap as an official number (LinkedIn's own Help says Support "cannot disclose" it — no figure is published) and SSI as a reach input (SSI is real and official, but LinkedIn never states it's a feed-ranking signal and even disclaims it "correlates with measurable outcomes").

Where it feeds the machine — stage 0, history 0

Two framings converge on this note, and both are load-bearing for utter-manual:

Stage 0 of the growth ladder. the-growth-ladder runs Chen's five stages from cold-start (stage 1: manufacture the first dense pocket by hand) upward — but cold-start already assumes the account exists. Bootstrap is the genesis stage before stage 1: no pocket to seed yet, no network, nothing to be stage-blind about. Its whole content is the one-time irreversible setup that stage 1 then builds on. A cell that reads its stage from metric history (growth-stages) has to know that the stage-0→1 transition is "the account is stood up and measurement is live," not a metric threshold.

History 0 of the diary. The utter-manual diary is a history (its atomic unit is one-cell-one-day). This is the genesis entry that history extends from — the root the daily loop is anchored to. When a stateless agent reads back through the diary to reconstruct state (git-as-agent-trust-substrate), history-0 is what it reads back to: the constitutional day-0 record of which handle/RED-ID was locked (§2), which account-path was chosen (§3), what measurement was wired (§4), and the positioning it inherited. Everything after is a delta on that genesis. The irreversible-field locks and the measurement wiring are therefore the natural genesis lint — a cell's diary cannot open its day-1 entry until history-0 is recorded and valid.

This is from-sop-to-operation's "day 0–30 is not the steady state," made concrete: the steady loop is the SOP's job; the genesis that the loop and the history both start from is this.