Improbable Collabs

[Improbable Collabs @ImprobableColla]

@davidrgoldberg Skill issue

[Improbable Collabs @ImprobableColla]

@businessbarista @tenex_labs That comp range is nuts. My friend at NVDA would crush this, those golden handcuffs might be tight tho lol

[Improbable Collabs @ImprobableColla]

@its21blackjack Same bro. I feel like a spokesperson for draft dawg but that’s been my go-to game that I actually enjoy recommending to people

[Improbable Collabs @ImprobableColla]

@OldJradd @PenguinHoops facts. there are actual good ones out there though, like draft dawg

[Improbable Collabs @ImprobableColla]

@jurajm_ @realapp @splashsports @BatterUp_app @hottakes_app @Stathead @TheSteinLine @statmuse Yo this is awesome! You tried? 82-0 or draft dawg? those are my go-to games rn

[Improbable Collabs @ImprobableColla]

@IbrahimS15 @stripe I’m guessing you’re a new grad? By next year you won’t give a shit about what normies think about where you work lmao, trust me.

[Improbable Collabs @ImprobableColla]

@ecomchigga SYSTEM

[ecomchigga @ecomchigga]

i make $8,247/month selling a PDF on a twitter account that has no name or face attached to it. the whole business runs in 14 minutes a day.

6 months ago the account didn't exist. here's every step.

1. picked a niche where people were already spending money on guides and courses. not what i found interesting. what people were actively paying to learn.

2. searched twitter and reddit for "how do i" and "struggling with" inside that niche. looked for the same question being asked by different people in different words.

3. checked gumroad and whop to see if products already existed in that space. competition isn't a warning sign. competition is confirmed demand.

4. picked the one problem i understood well enough to explain clearly to a stranger in one sitting.

5. opened google docs and wrote the complete answer like a long text to a friend who genuinely needed help. didn't format. didn't design. just wrote until it was done.

6. organized the mess into: what the problem actually is, why most free advice doesn't fix it, the specific step-by-step solution, and real examples with real numbers.

7. kept it between 11 and 22 pages. short enough to finish in one read. detailed enough that the reader doesn't need to google anything else.

8. added screenshots wherever a step needed proof or visual clarity.

9. exported as PDF. made a cover in canva with a free template. 8 minutes.

10. created a gumroad account. free until your first sale, then 10% per transaction. uploaded the PDF. wrote a title that describes the outcome, not the file.

11. priced it at $39. low enough for an impulse buy. high enough that buyers actually respect it and open it.

12. pulled 2 pages from the guide and turned them into a free preview. this is the lead magnet that feeds everything.

13. created a free telegram group. pinned the free preview at the top. pinned the $39 product link right below it.

14. set up the telegram so new members land in a room where other buyers are posting screenshots, asking questions, and sharing progress. the community does the selling. not you.

15. created a faceless X account. profile picture with no face. bio explains who i help and what result i deliver in one line. bio link goes straight to telegram.

16. used an aged account (6+ months old) to skip the algorithm's new-account suppression. fresh accounts get buried for the first 60-90 days regardless of content quality.

17. found 10 accounts in my niche between 10K and 50K followers. screenshotted their 50 best-performing tweets. studied the hooks, the formats, and the structures that consistently pulled views.

18. set up a Claude project with 3 reference files: a bank of 170+ proven viral tweets with view counts, a voice profile with rules that strip every AI tell, and a formatting guide that controls rhythm and sentence variation.

19. used Claude to generate 3 tweets per day. every tweet checked against the reference bank and scanned for AI patterns before posting.

20. scheduled everything through TweetHunter. posting times locked at 8am, 1pm, and 8pm with 4-6 hour gaps. spacing defeats the algorithm's session decay penalty.

21. never posted more than 3 times per day. the algorithm runs a penalty called the AuthorDiversityScorer that exponentially cuts your reach for every additional post from the same author in someone's feed. your 4th tweet of the day gets roughly 20% of the reach your 1st one got.

22. made 1 of the 3 daily tweets a CTA. structure: first 70-80% is pure value strong enough to bookmark on its own. then "comment [KEYWORD] and i'll send you the free guide. must be following + RT."

23. set up TweetHunter's silent auto-DM. when someone comments the keyword, they get the telegram link automatically. no public reply. silent only. public auto-DM replies get flagged as spam behavior and tank your reach score.

24. replied publicly to every commenter for the first 30-60 minutes after every single post. each author reply fires a 75x engagement weight in the algorithm. this is the single highest-leverage action available to any creator on the platform and almost nobody runs it on purpose.

25. that 30-minute window exists because X's engagement cache refreshes every 5 minutes for new tweets but only every 10 minutes for tweets older than 30 minutes. engagement velocity in the first half hour propagates through the scoring system at 2x speed. after that the window closes and the same engagement is worth half as much.

26. spent 20 minutes every morning in DMs answering questions from new telegram members. kept answers short. when someone asked something the guide covers in depth i'd say "i cover this inside the full system, i don't normally go this deep in DMs but i can show you what's inside if you want." not a pitch. a boundary.

27. used yes-stacking in longer DM conversations. 4 questions the lead can only say yes to before the product ever comes up. by the time price is mentioned they've already decided without realizing it.

28. for anyone who said "let me think about it" i replied "honestly i don't think you're ready for this yet, let's revisit in a few months." took the sale away. 30-50% of stalled conversations closed same day from the near-miss psychology alone.

29. posted proof constantly. every gumroad notification screenshotted. every testimonial shared. every result documented. nothing on earth sells like receipts from real people.

30. raised the price $5 after every 20 sales. $39 became $44 became $49. same product. growing proof. more perceived value with every bump.

31. self-reposted top performing tweets at 12-24 hours with no penalty. re-posted them fresh after 48+ hours when the algorithm's cache resets and the tweet re-enters the candidate pool as brand new content. one great tweet performs 3 separate times if you time the reposts.

32. wrote one long-form X article per week. articles pull 300K-1M+ views in 2026 because they trigger dwell time (weighted +10 in the scoring) and bookmarks (weighted 10x). highest-reach format on the platform right now.

33. never posted off-topic. not once. one viral meme tweet feels good but mathematically drifts your content vector in the algorithm's embedding space. every on-niche post after it reaches fewer people because the system is less certain what your account is about.

34. stayed consistent even when growth felt invisible. the algorithm scores your posts based on recent engagement history before anyone sees them. going silent collapses that baseline. recovery takes 5-10 consecutive strong posts before reach returns. consistency isn't a motivational poster. it's a mechanical input to the scoring system.

35. added a $497 comprehensive course behind the $39 front-end product. 2-4% of $39 buyers upgrade within 30 days without being pitched.

36. added a $5K 1-on-1 partnership behind the $497 course. one person applies every 2-3 months. this single tier changes the entire math of the business overnight.

37. batched all content creation on sundays. one 90-minute session writing the full week of tweets. the system runs the remaining 6 days and 22.5 hours without me.

38. tracked what actually drove replies and bookmarks and cut everything else. likes are weighted 1x in the algorithm. replies are 13.5x. bookmarks are 10x. reposts are 20x. optimizing for likes is optimizing for the cheapest signal. the scoreboard is public knowledge and most creators have never looked at it.

the tools:

TweetHunter: $49/month (scheduling + silent auto-DM)
Claude: $20/month (tweet generation with reference system)
Gumroad: free until first sale, then 10% per transaction
Telegram: free
Canva: free
Tally: free (application form for $5K partnership tier)

total monthly cost: $69. total daily time: 14 minutes.

the timeline:

week 1-2: setup, first tweets, $0
month 1: 800 followers, finding the rhythm, $340
month 2: 2,100 followers, CTA tweets clicking, $1,800
month 3: 4,400 followers, telegram compounding, $3,200
month 6: 11,000 followers, backend kicking in, $8,247/month consistent

the first $340 took 6 weeks. the jump from $340 to $8,247 took 4 months. the system compounds because the telegram community grows every day and never shrinks. more members means more proof. more proof means higher conversion. higher conversion on the same traffic means the revenue climbs without the effort climbing with it.

i put all 38 steps into a full system. way more detail than fits in a tweet. module-by-module walkthroughs, the exact Claude prompts, the DM scripts word for word, the telegram setup, the pricing framework, the algorithm breakdown, and every template i use.

comment "SYSTEM" and i'll send you the link. must be following + RT.

[Improbable Collabs @ImprobableColla]

@kingpriyom @FranWalsh73 You don't understand how $SPYI works.

[Improbable Collabs @ImprobableColla]

@ThinkingUSD how does this indicate they called the top? (I'm a noob)

[Improbable Collabs @ImprobableColla]

@kingpriyom @FranWalsh73 If you could guarantee 10%/year people would be retiring a lot earlier...

[Improbable Collabs @ImprobableColla]

@ramit time to buy your book :)

[Improbable Collabs @ImprobableColla]

@p_millerd Seeing the number of swings at-bat you all are taking makes me realize that I'm not thinking big enough.

[Improbable Collabs @ImprobableColla]

@riskmaxxing here I was thinking a milly before 30 was a big deal

[Improbable Collabs @ImprobableColla]

@ramit Did you put the money in an HYSA? I'm debating throwing cash in VOO for purchases that are >3 years out, tho I know some would argue that's unwise for anything less than 5 years out

[Improbable Collabs @ImprobableColla]

@eastvillageguy Lease takeover is superior but yea this is what I did to ultimately overpay for a 1Bd in Williamsburg

[Improbable Collabs @ImprobableColla]

@MediJo20 This and Draft Dawg are the two best NBA games out there IMO

[Improbable Collabs @ImprobableColla]

@yafriendray Lorimer L > Bedford L

[Improbable Collabs @ImprobableColla]

@ShaanVP Shaan bro, you gotta try draft dawg. Way better than 82-0 IMO