The steal file
Nine teams, nine signature moves - deconstruct, re-skin, ship your version.
- nine teams
- signature move
- transferable principle
Nine teams worth following closely. Each one has a signature move. Under each move sits a principle you can transfer to your own product. Most of these plays were built by very small teams.
Anthropic
Customer at the center, product in the background - the campaign is about the people using it.
Steal: make the user the hero, not the feature.
Lovable
Censors its own announcements - blurred a partner name, let comments guess for hours.
Steal: withhold the reveal to manufacture comments.
Ramp
Makes the pain physical - Kevin filing 600K expense reports in a glass box. 110M+ views.
Steal: show the "before state," don't describe it.
Torq
In a lookalike category, looking different wins - flaming skeleton, a fictional "hype monkey" intern.
Steal: build a recurring character, not one-off posts.
tl;dv
Runs the company page like a stand-up account; role-specific jokes. 0→2M users on comedy.
Steal: narrow the joke to one role at a time.
Clay
Cult by design - a hard-to-earn "Certified" badge, a 20K public Slack doing support.
Steal: give best users something scarce to belong to.
Cluely
Provocative position people fight about for free; publishes the playbook including flops.
Steal: pick a fight; be honest about failures. (Study the edge, don't copy it.)
Ahrefs
Free premium tools + internal data; a teaching channel that never advertises. Bootstrapped to $149M ARR.
Steal: give away insight you could charge for.
Decart
Every product ships as a playable public link on day one. A tiny team out-marketing labs many times its size.
Steal: replace the demo form with a link people can play with.
Do this first
- 01Pick the one team above whose audience most resembles yours. Save their last ten posts in a folder.
- 02For each post, write the mechanic in one line - what made it spread, stripped of the brand.
- 03Re-skin one mechanic for your product and ship it this week. Copy the structure, never the joke.
The anti-pattern list
Mistakes a whole room signed off on while somebody stayed quiet.
- nine ways to blow it
- stay benign
- punch up
Mistakes that a whole room signed off on. The meta-lesson: build a culture where a junior can say "this is NOT good" to the boss without getting fired - that's the real job.
✕Fake a "loved by" logo wall
Only spotlight customers you actually have, even if the list is short.
✕Break the platform's rules for a joke
A fake job title to dunk on a rival got a 100K-follower account shadowbanned. Run the joke without breaking the rules.
✕Call your last version "cringe" in public
Trashing the old version insults every customer who used it. Praise the new; pretend the old was great too.
✕Punch down
An upstart swinging at giants is expected; the giant mocking small competitors is a bad look. Never trash a rival mid-deal.
✕Argue with angry customers in public
The lowest-scored comment in Reddit history is a corporate reply defending a policy. Change the policy or stay off the keyboard.
✕Joke on sensitive ground
Religion and politics will burn you. Benign violation only works when it stays benign.
✕Hold a gift hostage
Asymmetric gifts only work when the gift is bigger than the favor. A padlocked bottle "unlocked" by a 30-min call is extortion; a single sock returned for a 30-second review is charming.
✕Target "every industry" in one message
Readers see "every industry" and decide it isn't for them. Pick one, run five specific versions, let data choose.
✕Promise "unlimited" without doing the math
Model the top 5% of users before any all-you-can-eat promo. If the unit economics break there, the promo dies before launch.
Three ways a hyper-growth product loses the people it won
Even a rocket collects scar tissue. None of these are growth failures. All three are trust failures.
- Silent regressions. A small edit quietly breaks working logic. Ship deterministic tests, or the product degrades under its own success.
- Lock-in backlash. If the frontend exports but the backend stays trapped, power users rebuild from scratch the day they outgrow you. Then they post about it.
- Opaque billing. Retry loops burn paid credits. Support falls behind. Trust drains faster than any campaign can refill it.
Steal: the anti-patterns that kill you are the ones your own growth creates. Audit for silent breakage, exit paths, and billing surprises before you audit your ads.
Do this first
- 01Read the list above against your live site and last month's campaigns. Flag anything that fails: fake logos, borrowed claims, unmodelled "unlimited" offers.
- 02Name one person who is allowed to say "this is not good" before anything ships, and tell them out loud that it's their job.
- 03Audit your own growth for trust debt: silent breakage, missing exit paths, billing surprises. Fix the biggest one before your next campaign.
Operating cadence
Score it on ICE, nail it before you scale it, standardize only what already works.
- ICE
- nail it first
- standardize four things
- test more
How to run all of the above as a repeatable practice instead of a lucky streak.
01Prioritize experiments, don't just build
Score every idea before you build it - Impact × Confidence × Ease.
see detail →hide detail ↑
High-ICE, low-effort tests ship first. The failure mode is simple: you fall in love with building, and skip the validation the score would have forced.
02Nail it before you scale it
Get one thing working with real people, then systematize.
see detail →hide detail ↑
Especially with humor, don't lock structure before you've seen a vivid response. Start small - learn on a series of posts, not a Super Bowl spot.
03Standardize the four things (only after it works)
Topics, formats, voice, content rules - locked in once, then repeated.
see detail →hide detail ↑
Topics
The handful of things your audience has strong feelings about, that you keep returning to.
Formats
Repeatable shapes (a first-person story, one meme layout) your audience recognizes on sight.
Voice
2–3 adjectives (dry, deadpan, warm) you hit every time.
Content rules
Your always/nevers - always real numbers, always end on a one-liner. Consistency is the fingerprint.
04Test more, pre-judge less
The editor arriving too early is what holds teams back.
see detail →hide detail ↑
A boss in the room means nobody risks anything. Say it all first, cut later. Perseverance substitutes for talent. When a format doesn't land, change the format - don't repeat the same thing louder.
The whole model in one line
Find the wedge, make people feel seen, build loops not funnels, show the product instead of describing it, and make one real person feel you made it only for them. Nail it before you scale it.
Pre-launch checklist
- Is there a wedge? A sharp, contrarian angle that makes this easy to spread - not a better-mousetrap claim.
- Does it make someone feel seen? An unspoken truth, an inside joke, a shared pain.
- Is there a loop? What does this produce that feeds the next turn - comments, content, users, data?
- Can you show it instead of telling it? A playable link or a visible "before state" beats a description.
- Right surface for the goal? Differentiate → ad; reduce friction → 404; connect → email; awareness → careers page.
- Does it sound human? Read it out loud; kill formality, adjectives, and banned words.
- If funny: specific + surprising + benign, survives deadpan, punches up not down.
- Is every claim honest? No fake logos/titles, no trashing the old version or a rival, unit economics modeled.
- Small enough to be a real test? Nail it before you scale it - and score it on ICE first.
Do this first
- 01Put every idea currently in flight into one list and score each on Impact, Confidence and Ease, 1-10. Kill the bottom half today.
- 02Run the top-scoring idea as the smallest real test you can - one post, one page, one segment - and set the number that decides whether it worked.
- 03Write down your topics, formats, voice adjectives and content rules only for the things that already worked. Repeat those weekly.
The pricing loop
Free tier is the top of the loop; the paid gate is a growth mechanic, not a finance decision.
- free tier as top of loop
- meter what scales
- gate the graduation
Pricing usually gets handed to finance and treated as a separate discipline. It isn't. What you give away decides how many people enter the loop. What you gate decides who exits it. Both are distribution choices.
Meter what scales with value
Separate the credits people spend building from the ones burned by live, running workloads. Revenue then tracks how hard people use the product, not seat count. The bill also makes sense to the person paying it.
Gate the graduation, not the aha
The free tier has to reach a real outcome. Put the wall where success creates new needs: custom domains, code export, higher throughput, integrations. Upgrading should feel like graduating, not like paying a toll.
01Price the free tier as a marketing line item
Free deploys that carry your name are cheaper reach than ads, so budget them like media spend.
see detail →hide detail ↑
Model the top 5% of free users before you widen the tier (§10). If the unit economics survive there, generosity is acquisition. If they don't, the loop eats you.
02Make consumption legible
A meter people can't predict is a trust problem disguised as a pricing problem.
see detail →hide detail ↑
- Show the balance before the action, not after.
- Never let a retry loop or a background job spend silently.
- Refund fast when it happens anyway.
- Make planning cost zero. Exploring should never feel expensive.
The test
Ask of pricing what you ask of every tactic. What does this produce that feeds the next turn? A free tier that produces public artefacts, templates and word of mouth is a loop. One that only produces cost is a discount.
Do this first
- 01Write down the outcome a free user must reach to feel the product work. Check that your free tier actually gets them there.
- 02Move your paywall to the moment after that outcome - export, domains, throughput, collaboration. Anything gating the aha comes down.
- 03Model your top 5% of free users before you widen anything, and show the meter before the action so nobody is surprised by a bill.