Investing in AI when it moves too fast: a builder's read
Two VCs gave advice on investing in AI when everything moves too fast. Here's how I'd translate it for a Sri Lankan engineer placing time and money bets.
How to invest in AI when everything moves too fast is a question I used to think only applied to people with a fund. After reading TechCrunch's writeup of its StrictlyVC evening in Los Angeles, I think it applies to every engineer in Sri Lanka deciding what to learn next, what side project to start, and where to spend a thin monthly budget.
The investors were talking about cheques. But the underlying problem is the same one I have: how do you commit to anything when the ground shifts weekly?
🔍 What the two investors actually said
The panel paired Carter Reum, co-founder of M13 ($2.5 billion under management, with stakes in 17 unicorns), and Chang Xu, a partner at Basis Set Ventures, an AI-focused fund that started in 2017 and is now investing from its fourth fund of nearly $1 billion.
Xu named the trap directly. Growth curves are unreal right now (she cited ChatGPT reaching $40 billion in revenue in six months), which makes sky-high valuations look almost reasonable. Her warning:
"If you price every single deal to that math, there's no way that will work out well."
Reum's point was that we have seen this movie before, with cloud and with the iPhone. What's different now is that a new startup competes against "the ten largest tech companies on the planet" from day one.
Key takeaway: Fast growth is real, but pricing your bets as if the fastest case always happens is how you lose. That's true for a fund, and it's true for the 80 hours you're about to pour into a side project.
⚡ Velocity markets vs depth markets
The most useful idea for a builder was the split between two kinds of markets:
| Market type | What wins | Risk for a small builder |
|---|---|---|
| Velocity market | Raw speed, shipping first, riding the model of the week | A bigger player or a model update erases your edge overnight |
| Depth market | Solving a genuinely hard problem that stays hard | Slower to show traction, but the moat lasts |
If you are a solo developer in Colombo with no funding, a pure velocity play is brutal. You are racing the same ten companies Reum mentioned, and they ship faster than you can. The thin-wrapper-over-an-API project that looks clever today is a feature in someone's app next month.
Depth markets are where a small team can actually hold ground, because the hard part stays hard regardless of which model is on top.
🛠️ "Friction as a moat" is good news for Sri Lankan builders
Reum's phrase that stuck with me was friction as a moat: regulated, messy, slow-moving industries like healthcare and emergency services where the hyperscalers don't bother to move quickly. The friction that makes a market annoying is exactly what keeps the giants out.
For a Sri Lankan builder, this maps almost perfectly onto local problems that global tools ignore:
- Tax, EPF/ETF, and payroll rules that change with every budget and that no US startup will ever localise.
- Government forms, fees, and processes that need someone who actually understands the system.
- Sinhala and Tamil language handling that big models still do badly.
That's the entire reason I build local-context tools instead of generic ones. A US fund will not fund a Sri Lanka tax calculator, and a US startup will not build one. The friction is the opportunity.
Don't compete where the giants are fast. Build where they refuse to go.
💡 "Cocktail napkin math" for your own time
The investors said they still rely on "cocktail napkin math" to sanity-check fundamentals before getting swept up in the hype. You can run the same check on your own bets, except the currency is your hours.
Before committing to a project or a tool subscription, I ask:
- What does the fastest case require to be true? If it needs me to out-ship Google, cut it.
- Does this survive the next model release? If a single OpenAI or Anthropic update kills it, it was a feature, not a product.
- Can I price the real cost? API calls add up fast. I run the numbers on our AI model comparison and the free-tier comparison before I write a line of code, so I know whether the idea works on a learning budget at all.
That last one matters more here than in LA. When your runway is a student stipend or a freelance month, a wrong subscription is a real loss.
Key takeaway: The napkin math isn't about being negative. It's about refusing to let a 1% outcome justify a 100% commitment.
🌐 The microscope and the telescope
Reum's closing image was that a founder needs "a microscope in one eye and a telescope in the other" — execute on today while watching for the shift that makes today irrelevant.
For me that means two habits I'd recommend to any engineer here:
- Microscope: ship something small and real every week. Free tiers and open-source models mean you can build and learn without spending. The execution muscle is the part nobody can take from you.
- Telescope: spend an hour a week just tracking what shipped. The model that changes your plan will arrive with no warning.
You don't need capital to apply any of this. You need to pick depth over velocity, build where friction protects you, and refuse to bet your hours on the fastest case happening.
What this means for you
The VCs in that room were deciding where to put millions. You and I are deciding where to put evenings and a small budget, which honestly feels riskier because we can't diversify across 17 companies.
So borrow their discipline. Avoid velocity races you can't win against the ten biggest firms on earth. Find the friction that keeps them out, which in Sri Lanka is everywhere. Run the napkin math on your hours before you commit. And keep one eye on the long view, because the thing that resets the board is already being trained somewhere.
Moving fast isn't the skill. Knowing which fast things to ignore is.
Original source
How to invest when everything is moving too fast