Small Bets

Table of content

Daniel Vassallo left a $500K/year job at Amazon to sell info products and run experiments. He didn’t start a VC-backed company. He made small bets.

The philosophy: instead of going all-in on one big idea, run many cheap experiments. Most will fail. A few will work. The portfolio survives because no single failure kills you.

One Big Bet vs. Many Small Bets

The default path looks like this: quit your job, raise money, build for 2 years, hope for product-market fit. If it fails, you’re broke and burned out.

Small bets flips it:

One Big Bet	Small Bets
2-year runway	Weeks per experiment
Need to be right once	Can be wrong repeatedly
Success = exit	Success = sustainable income
Investors pick winners	You discover winners through volume

Nassim Taleb calls this “barbell strategy” in Antifragile. Most exposure in safe assets, small exposure in high-upside experiments. You can’t lose more than the bet. You can win multiples.

How Small Bets Actually Work

Vassallo’s own portfolio includes:

The Good Parts of AWS: ebook, $300K+ revenue
Everyone Can Build a Twitter Audience: course, $400K+ revenue
Small Bets Community: paid community, 7,600+ members
Userbase: open-source product (pivoted/killed)
Various SaaS experiments: some failed, some work

Not one company. A portfolio. Some products took a weekend to ship. Others evolved over months. The key: each bet had limited downside and required validation before heavy investment.

The $1,000 Test

Vassallo’s heuristic: before building anything serious, try to make $1,000 with a small project first.

This filters out most bad ideas quickly:

Can you explain it in one sentence?
Will someone pay before it’s perfect?
Can you ship a version this week?

If yes to all three, run the experiment. If not, move on.

Small Bets in AI Projects

The philosophy maps directly to AI development:

Old way: Build one complex agent, spend months on architecture, hope it generalizes.

Small bets way: Build 10 single-purpose tools. See which ones get used. Kill the rest.

Examples:

Big Bet	Small Bet Equivalent
Full AI assistant	One slash command that summarizes threads
Autonomous coding agent	Script that runs tests and reports failures
Universal memory system	Plain text files with grep

Parallel sessions are small bets in action. Run five experiments simultaneously. The ones that work get expanded. The ones that don’t get abandoned without guilt.

Learning in public compounds small bets. Each blog post is a bet on audience building. Each TIL is a bet on future-you needing that reference. Volume creates optionality.

Why People Avoid It

Small bets feels less serious. We’re trained to respect the founder who raised $10M and works 80-hour weeks. The person shipping weekend projects looks like a dabbler.

But survival rate tells a different story. Most VC-backed startups fail. Most indie hackers with portfolios have sustainable income within a few years.

The ego hit is real though. “I have 12 small projects” doesn’t sound as impressive as “I’m building the future of X.” You have to decide what you’re optimizing for.

What You Can Steal

Run experiments in parallel. Don’t finish one before starting another. Overlap keeps you moving when individual projects stall.
Set kill criteria upfront. “If this doesn’t hit 100 users in 30 days, I’ll shut it down.” Then actually shut it down.
Ship embarrassing v1s. The goal is learning, not perfection. A weekend prototype that gets feedback beats a month of planning.
Keep the day job longer. Vassallo advises: don’t quit until your portfolio generates enough. Small bets works because you’re not desperate.
Reuse everything. That failed project’s auth system becomes your next project’s auth system. Bets compound when components transfer.

Next: Learning in Public shows how sharing your experiments creates compounding returns.