2026-08-28·by Sijie Wang#market#awareness#x-court

nikita-bier-sell-the-spike

b. 1990, UC Berkeley. The consumer-app operator whose stated physics deny durability — and whose machine is built accordingly: manufacture a spike, monetize it, sell before decay. tbh: launched Aug 2017, sold to Facebook Oct 16, 2017 — nine weeks in ($30M by later press consensus; contemporaneous TechCrunch said only "<$100M"). Gas: launched Aug 2022, sold to Discord Jan 2023. Both died inside their acquirers (tbh shut July 2018 "for low usage"; Gas didn't retain in Discord either) — which is not a refutation of the model but its confirmation.

The iteration asset

tbh was app #15. Fourteen predecessors died ("map apps, chat apps, to-do lists"); the first took ~1 year to build, the last ~2 weeks. The compounding asset was never any single app — it was the shrinking cycle time.

The seeding machine (tbh, Aug 2017)

  1. Launch at one Georgia high school — Georgia schools start earlier; the academic calendar is a first-mover resource.
  2. Create a school-specific private Instagram account; follow every student whose bio names the school; post nothing but mystery.
  3. At ~4pm, flip the account public with a download link and accept all pending follows.
  4. Result: 40% of the school installed within 2 days; spread hopped to 3 nearby schools on its own.
  5. Roll out geofenced, state-by-state — the locked states' clamoring is manufactured waitlist scarcity, i.e. free distribution.

Nine weeks later: 5M downloads, 2.5M DAU (daily active users), #1 US App Store. Gas repeated the shape with one monetization twist — "God Mode" (~$6.99) selling hints about who complimented you: 7.4M installs, ~$7M consumer spend at acquisition.

The doctrines (his words)

  • Audience physics are quantitative: "the number of invitations sent per user drops 20% for every additional year of age from 13 to 18… If you build for adults, expect to acquire every user with ads."
  • Test at network density, not user count: "you need to get an entire school to adopt, just to know if everyone had 10 friends, would they actually derive value."
  • Time-to-value: "if you can't demonstrate value in the first three seconds, it's over."
  • Latent demand: Sarahah hitting #1 in the US while entirely in Arabic was "one of the strongest signals… that people want something" — demand shows up as users suffering through a distorted process to get a value.
  • The honest core: "Building a retentive social product is a black swan event."

The account as career capital

~500K X followers at hiring, ~1.1M by 2026. His floor formula: one subject "you know more about than anyone else," one insight under 5 sentences, daily, for 6 months. His joke ladder (2022): "0–500: Reply Guy… 25–50K: Shitposts… >100K: Get Cancelled." In 2022 he publicly pitched himself to run Twitter product; July 1, 2025 it happened — "I've officially posted my way to the top." The terminal rung of the collab ladder (airrack-the-collab-ladder) turns out to be the judge's own product seat.

In office, his moves read as a war on low-effort arbitrage: aggregator payout cuts (60%, then a further 20% deduction), reupload detection that redirects revenue to original authors, Communities shut down (0.4% usage, 80% of spam reports). The A4-tier (a4-the-language-arbitrage) of X is being taxed by a man who knows exactly how it works.

Caveats

The $30M is unconfirmed consensus; Gas revenue varies $6M–$7M across tellings. Dark-pattern critiques are documented (engineered FOMO, notification loops, contact-graph exploitation of minors' invite behavior); Gas survived an October 2022 viral sex-trafficking hoax that triggered police warnings despite the app having no messaging or location features.

PRINCIPLES

  1. Sell the spike: if retention is a black swan, the peak is the product — convert (exit, launch, monetize) before decay; time-to-sale is a design parameter, not an outcome.
  2. Iteration speed is the real asset: 15 apps, 1 year → 2 weeks per attempt; the portfolio compounds even when every unit dies.
  3. Seed at network density: the adoption unit of a network product is the network (a school), not the user; 40%-of-one-school beats 0.1%-of-everywhere.
  4. Manufactured scarcity is free distribution: geofenced rollout converts the excluded into a waitlist.
  5. Audience physics are measurable: invite propensity decays 20% per year of age — pick audiences by their sharing mechanics, not their spending power.
  6. The account is career capital: daily posting compounded into distribution, hiring reach, and finally the judge's own seat.

Sources: 15 failed apps (Startup Archive) · seeding mechanics (Synergy Labs) · tbh acquisition (TechCrunch 2017) · Gas numbers (Appfigures) · Discord acquisition · doctrine quotes (Lenny's transcript) · black-swan tweet · X hiring (SF Standard) · posted my way to the top · reupload detection · hoax debunk (Washington Post)

about this entry

One of sijie's wiki entries. The AI on this site is grounded in the same corpus and answers in sijie's voice, with citations back to entries like this one — answering costs sijie money, so it waits behind a code: enter an access code →