5 Bizarre AI stories straight from 2050

Why does it matter? Because AI just made one person more powerful than a pharma company, a PE firm, and a research institution — simultaneously.

The old gates are gone. Not crumbling — gone. One guy cured his dog's cancer with off-the-shelf AI tools. PE firms are paying software multiples for service businesses. A company accidentally became the world's most valuable pedestrian-data holder. And 13-year-olds are making 21-year-old founders feel like the old guard.

Walk away knowing:

• The bottleneck in personalized medicine is now paperwork, not biology
• Anthropic did $6B in a single month — more than Snowflake does in a year — and is still only valued at $400B
• Service businesses are getting repriced upward in real time, and most owners don't know it yet
• The most accessible high-upside business model right now requires no product, no funding, and no team

One guy cured his dog's cancer with ChatGPT, AlphaFold, and Grok — and the paperwork was the hard part

An Australian entrepreneur heard there was nothing he could do about his dog's cancer. He decided that was someone else's problem.

He used ChatGPT to figure out the approach: sequence the tumor's DNA, compare it against the dog's healthy cells, isolate the mutation driving the growth. He got that sequencing done at a lab by framing it as research. Then he fed the results into AlphaFold — DeepMind's protein structure prediction tool — to figure out the physical shape of what he was attacking. Then he used Grok to design a single-shot custom vaccine.

Shaan's framing: "he decided to enter the high agency Olympics and manually go cure his dog's cancer."

He found a lab. Convinced them to manufacture it. The thing that almost stopped him wasn't the science — it was the approvals. He eventually got clearance under a researcher's ethics umbrella, administered the vaccine, and cured the dog.

The reusable principle here is blunt: the biology is no longer the hard part. Any reasonably technical person with access to the same tools this guy used — all public, all free or cheap — can attempt treatments that didn't exist before. The constraint is bureaucracy. That's the layer the next wave of startups is going to attack.

Anthropic did $6B in one month — and at $400B, it might still be undervalued

$6 billion. In February. One month.

To put that in context Shaan borrowed from the All-In pod: that's more revenue than Snowflake or Databricks — "two of the greatest software companies of the last 20 years" — generate in an entire year. Anthropic did it in 28 days.

And they've done this 10x growth thing not once but three or four years running: zero to $100M, $100M to $1B, $1B to $10B, and now tracking toward $100B. The current valuation sits at $400B, which Shaan calls "a steal, shockingly."

The wrinkle: Anthropic's CEO Dario Amodei isn't out there flexing about trillion-dollar data centers. He's being careful. Because the same math that makes the growth look miraculous also makes the risk existential — if you invest for another 10x and only get 5x, you go bankrupt. He said that explicitly.

Sam's observation cuts to the real shift: when he and Shaan were coming up in San Francisco, the golden standard was "triple, triple, double, double." That was the blueprint for a once-in-a-generation company. Now the benchmark is 10x every single year. They're not even in the same ballpark anymore. Founders and investors who are still calibrated to the old standard are measuring with the wrong ruler.

PE firms stopped buying SaaS — they're buying service companies and pricing them like software

A year or two ago, Sam's friend Rine wrote a blog post calling this. He titled it "Service as Software" — the inverse of the famous SaaS acronym — and the thesis was simple: service businesses trade at lower margins and lower multiples for two reasons. They require lots of people, which kills gross margin. And they require skilled people, which kills scalability.

AI eliminates both problems.

One person can now do what five people did before. Gross margins in AI-enabled service businesses are heading toward 75% — up from the 40-50% that used to be the ceiling. And the more AI handles the repeatable work, the more the business can scale without proportionally adding headcount.

The market has noticed. Sam and Shaan talked to PE firms who've already made the move: they've shifted budget away from buying SaaS companies and are buying service companies instead — valuing them the way they used to value software.

That repricing has started. The window where you can acquire or build a services business before the multiples fully adjust is closing. If you own one, Sam's framing is direct: you can either decide your business is over, or decide your multiples are about to go up significantly. Those are genuinely different paths with very different outcomes.

Pokémon Go's real asset was never the game — Niantic kept the data and is now licensing it to AI companies

Niantic sold Pokémon Go for somewhere around $2-3 billion. The game had a massive hype cycle, faded, changed hands.

But Niantic kept something the buyer didn't get: every photo and video captured by roughly 100 million players walking around cities worldwide with their cameras open. Street-level footage of sidewalks, intersections, storefronts, and pedestrian paths in nearly every major city on earth — captured from human eye height, not from a car-mounted camera doing a sweep every decade like Google Street View.

Delivery robots need exactly this. To navigate sidewalks, they need to recognize terrain, obstacles, curb cuts, and street context from a pedestrian vantage point. Niantic has the only dataset like this in existence. They're now licensing it to AI companies, and they're the only ones who can.

Shaan's line on data-as-oil cuts through the usual noise: it's not that all data is valuable — it's that very specific types of data are like rare crude. "You might have a specific type of oil that nobody else has."

The question worth sitting with: what behavioral or physical-world data has your business been quietly collecting that seemed like exhaust at the time? Because a few companies are waking up right now and finding a lottery ticket in their back pocket.

The AI transformation consultant is the new SMMA — and the playbook is identical

Remember SMMA? Social media marketing agency. The 2015-era "business in a box" where the pitch was: businesses know social media matters, they don't know how to use it, and you bridge that gap for a monthly retainer.

Shaan says that exact same moment is happening right now with AI.

Every business owner knows AI is important. Almost none of them know where to start or who to trust. The gap between "I should be doing something" and "here's what you should actually do" is enormous — and it's widening every week.

The playbook: pick a niche. Learn the AI tools relevant to that industry on nights and weekends. Go to business owners with a no-risk offer — "I'll do an audit of your business and tell you where AI can help." Charge nothing for the audit. Sell the implementation. Do it once with one dentist, then do it with 50 more dentists.

Shaan demonstrated this live: he fed his brother-in-law's retail shopping center business into Claude, got 10 AI use cases, grabbed the best one, threw a prompt into Perplexity's computer, and had a live working dashboard tracking which retailers were expanding and contracting — in minutes. His brother-in-law's immediate response: "Can we do a call today?"

That reaction is the market. It's everywhere. The only thing missing is the person who shows up to bridge it.

13-year-olds are making 21-year-old founders feel ancient — because role models, not talent, were always the binding constraint

Sam had dinner with eight young entrepreneurs — oldest was 25 — who were treating him the way he treats seasoned veterans. He asked them what 21-year-old founders are into these days. Their answer: hiring, scaling, girls. "The new gym, tan, laundry."

But the more striking part of the conversation was this: those same 25-year-olds told Sam they feel about 13- and 14-year-olds the way he feels about them. In awe. Slightly intimidated. Already a little behind.

Shaan's explanation doesn't invoke talent or genetics. It's purely mimetic: "we're all mimetic creatures." When he was 13, the only things he had to model were the NBA and skateboarding — "that was what was the cool thing you could go try to emulate." He didn't meet an actual entrepreneur until he was roughly 18.

Now a 13-year-old with a phone can fall into a rabbit hole at 9pm and by midnight have watched 40 minutes of a 19-year-old founder explaining how they sold a company doing $30M ARR. That becomes the new normal. The new reference point for what's possible.

Human potential for entrepreneurship hasn't changed. The training data has.

'High AI exposure' doesn't mean your job disappears — it means the market for your job explodes

Andrej Karpathy — OpenAI co-founder, ran Tesla's self-driving program for years, now just posting AI experiments on Twitter "like Peter Levels with six brains" — built a visualization at karpathy.ai/jobs. Every US job plotted by size, colored by AI disruption risk. Red means high exposure to AI reshaping. Green means relatively safe.

The counterintuitive call: software developers scored high-red — and that's not bad news.

Karpathy's own caveat is the key: high AI exposure doesn't mean replaced, it means reshaped. For software developers specifically, AI expands the demand for software faster than it displaces the people writing it. Each developer becomes more productive, which makes software cheaper, which creates demand for more software, which needs more developers.

"The red doesn't mean replaced, but it does mean reshaped."

The mistake is treating disruption risk and job-elimination risk as the same thing. They're not. The question isn't whether AI touches your field — it touches almost everything. The question is whether AI grows the underlying market or shrinks it. In high-demand fields, the answer is almost always grows.

The bottleneck keeps moving — and it's always one layer above where everyone is looking

Every story in this episode points at the same thing: the hard part is no longer the hard part. Biology wasn't the obstacle for the dog cancer guy — bureaucracy was. The science of AI isn't what's stopping businesses from benefiting — implementation is. Service company margins weren't the ceiling — headcount was. The data's value wasn't obvious — the use case hadn't arrived yet.

Where things are heading: whoever figures out how to compress the gap between "AI can theoretically do this" and "AI is actually doing this inside your business" wins the next decade. That's a regulatory problem, an education problem, a trust problem, and a distribution problem — all at once. None of it is a compute problem anymore.

---

Topics: AI, entrepreneurship, business models, cancer research, data monetization, job disruption, youth entrepreneurship, service businesses, Anthropic, Pokémon Go