I use AI to break down how businesses actually make money — the famous ones, the weird ones, and the ones you'd never notice.
1 breakdown every Tuesday.
I don't run these companies.
I use AI to read the parts of them nobody reads — 10-Ks, earnings calls, pricing pages, unit economics — and then I explain, in plain English, how the money actually moves.
Here's what doing that every week teaches you:
1. Most companies don't make money from the thing you think they do. The product is often the marketing. The money is somewhere quieter.
2. If you can't explain how a business makes money in two sentences, you don't understand the business. You understand its advertising.
3. Boring beats brilliant. The most profitable businesses are usually the ones nobody posts about — until someone runs the numbers out loud.
4. Pricing is the whole game. The same product at a different price, bundle, or billing cycle is a completely different business.
5. Distribution beats product more often than founders admit. The best business rarely wins. The best-distributed one usually does.
6. Every giant company was once a weird idea that sounded too small, too boring, or too strange to matter.
7. Numbers end arguments. "This company prints money" means nothing. Revenue per user, margin per sale, cost to get one customer — that's the story. And every number I post, I cite.
The AI is the research assistant, not the author. It reads 200-page filings in minutes. The judgment, the writing, and the fact-checking are the account.
Every week here:
→ Tuesday: one full breakdown — how a famous or weirdly profitable business actually makes money
→ Friday: one startup idea pulled from public data
→ Saturday: why money behaves so weirdly (psychology, pricing, incentives)
No guru stuff. No "I built a 7-figure business" claims — I didn't, and this account will never pretend to. Just the numbers, cited and explained clearly.
Follow if you've ever looked at a business and thought: wait, how do they actually make money?
That's the only question this account answers.
— Bifurcation AI
elvis died in 1977 and his catalog is still finding new ways to get paid. 274 songs in a training lawsuit now, and Sony, Universal and Warner are suing Anthropic too. training data's about to have a price tag
The owners of Elvis Presley’s music are suing Anthropic, the company behind Claude, accusing it of using 274 Presley songs to train its AI models without permission, per The Guardian.
The estate’s lawyer accuses Anthropic of building its business on copyright infringement at an unprecedented scale.
Anthropic has also been sued by music publishers including Sony, Universal and Warner.
@fortworthchris "you no longer have a lock on your door" is a wild way to describe an encryption problem. also accidentally the best sales pitch post-quantum security startups will ever get
@JoePompliano the 'no baseball experience' part was the feature, not the bug. they just ran it like a portfolio — buy low, trade guys before the expensive years, never fall in love with an asset
@nickgerli1 28% in 2.5 years and that's before commissions. throw in HOA and taxes on a downtown tower and that seller probably bled $400k all-in to own a condo for 30 months
@jolieganache wait so the $500 headline can quietly turn into $1k+ once usage and whitelisting stack on top? kinda breaks the whole '20 deals a month' math then
The whole model rests on one number: a creator with 10k followers averaging $5k a month in brand deals. At typical rates for that size — $150 to $500 for a UGC video or post — that's 10 to 30 closed deals per creator, every month. Most niches don't generate that many offers for accounts this small. The agent makes chasing deals nearly free. It doesn't make the deals exist. Managers skip sub-100k creators not because the labor costs too much, but because the deal flow is too thin to take 20% of. Automating the chase doesn't fix the flow.
You could easily pull $25k/month running brand deals for small creators with Grok Bot.
Silicon Valley keeps saying the next big companies are services businesses run by AI agents.
This is one.
A hungry creator with 10k followers who takes every deal, UGC included, can make $5k to $10k a month.
Bigger ones make $20k to $30k.
The deals are there. The labor of chasing them is the problem.
Managers won't touch anyone under 100k because the work costs more than their 20%.
This is the opportunity
Agents just made that work almost free.
1. Sign 50 hungry creators in one niche. Tell them you're powered by AI agents, so you take 10%, not 20%.
2. Give each creator their own Grok Bot. Connect their Gmail and WhatsApp and X, and give it their rate floors.
3. It’s routine: "Check email and WhatsApp and X and LinkedIn every morning, midday and late afternoon. Only show me threads where the next move is mine, with a reply drafted. Never send without my OK."
4. Add a Sponsorship agent: "Study this creator's niche. Find brands already paying creators this size, from UGC to dedicated posts. Draft a pitch for each."
5. Add a daily money check: "Match every payment to its invoice. Flag any finished post that was never invoiced."
6. You approve the drafts. You take 10% of every deal.
50 creators averaging $5k a month is $250k in deals.
Your 10% is $25k a month.
I run my own brand deals this exact way now.
There's an obscene amount of money sitting in deals nobody has time to chase.
@HHGF99@jasonlewris ok that's fair — if the cut's been sitting at 7-7.4% all year, then 44.8% of listings cutting isn't a wave, it's just everyone taking the same turn eventually
@ConnorAbene revenue per employee is the sneaky one tho. you can fake it for like a year just by not backfilling roles and quietly burning out whoever's left
@girdley southwest in the 'sold for parts' tier is harsh but they spent 2 years deleting everything people liked about them — free bags, open seating, all of it. not much left to defend
@DallasAptGP going south and west before north is a weird tell. 81 sq miles and it still basically hugs the core — wonder if that's where the demand is or just easier roads to learn on
@TurnerNovak@MenloVentures@mmurph the $750M fund could only write $50-100M of it so they stitched together a $500M SPV, and LPs wired in without seeing a single financial. that's not due diligence, that's a reputation compounding for like 20 years
@shawngorham 30 min of donuts and coffee before 6am turned into a $350k settlement covering five years of employees. most expensive free breakfast in history