LLenny's Podcast
← All frameworks
StrategyAmol Avasare

Exponential Bet Allocation

If AI is your core value, shift the growth portfolio from micro-optimizations to large bets

Difficulty
Advanced
Time to result
~months to results
Steps
3
Confidence
85%

Traditional growth teams spend ~70% of effort on small-to-medium optimizations and ~30% on big swings, because a normal product's value grows maybe 30-50% over two years — a range you can capture with high-conviction small wins. Amol argues that when AI is the central value prop, product value grows 100-1000x, so the incremental market from large bets dwarfs the optimizations. Anthropic therefore flips or evens the ratio toward big swings — but only for genuinely AI-first products.

Origin

Amol Avasare's framing at Anthropic, grounded in the company's cultural fixation on 'the exponential' and log-linear thinking; explicitly contrasted with the small-bet-heavy playbook he ran at prior traditional-growth roles.

Core principles

  • 01A team's bet mix should track how fast the underlying product value is compounding
  • 02Linear-value products (30-50% growth over 2 years) reward high-conviction small optimizations
  • 03AI-first products (100-1000x value growth) reward large bets that unlock entirely new markets
  • 04Small optimizations still matter and compound — you don't abandon them, you reweight
  • 05The rule only applies when AI is the central element of value, not a side feature

How to run it

  1. 1

    Diagnose whether AI is your core value

    Ask honestly: is the primary value your product delivers underpinned by AI as a central element, or does it have some AI features bolted onto a non-AI core? This single question decides which playbook applies.

    Pro tip Companies like Cursor and Lovable pass this test; a grocery-delivery app with an AI feature on the side does not.

    Watch out Misclassifying a non-AI-core product as AI-first will over-invest in speculative bets.

  2. 2

    Estimate the two-year value differential

    Project how much more product value you'll deliver in two years. For a linear business that's ~30-50%; for an AI-first business it can be 100-1000x. The size of this gap sets how much upside big bets can capture.

    Pro tip Anthropic literally plots everything on log-linear scales; if your charts only make sense linearly, you're probably a linear business.

  3. 3

    Reweight the portfolio toward big swings

    For an AI-first product, move from the traditional ~70/30 (small/large) split toward 50/50 or a large-bet majority. Keep doing the small optimizations because they compound and no one else will, but don't let them crowd out the forest for the trees.

    Pro tip Back conviction bets even when they're research-heavy and unproven — Anthropic's growth team built the Chrome extension because one engineer was bullish and 'no one else is doing it.'

    Watch out At scale a 1% optimization is worth enormous absolute dollars, which makes it tempting to tally up small wins and skip the hard big bets.

In the wild

Building the Chrome extension on conviction

An engineer on Anthropic's growth team was bullish on a research-heavy Chrome extension that no competitor was building. Rather than optimize something safer, the growth team backed the big, uncertain bet purely on conviction.

The Chrome extension now underpins multiple use cases across co-work and Claude Code — a large-swing outcome Amol says he wouldn't have pursued at a traditional company.

Common mistakes

Tallying micro-optimizations to look productive

At Anthropic's scale a 1% win is worth another billion, so it's easy to rack up small optimizations and claim big impact — while missing the exponentially larger markets that only big bets unlock.

Applying the big-bet playbook to a non-AI-core product

If AI is a side feature rather than the core value, product value grows linearly and over-indexing on large speculative bets wastes resource that high-conviction small optimizations would have captured more reliably.

Is it for you?

Best for

Growth and product leaders at AI-first companies where model capability is the central value driver and future value is compounding exponentially

Not ideal for

Products where AI is a peripheral feature, or resource-constrained teams whose value grows linearly and who benefit more from high-conviction small wins

From the transcript

if the primary value that your product delivers is underpinned by AI as as a central element of it, then I think you should operate…

31:00

the the product value that we will deliver in two years time is probably like a thousand X a hundred to a thousand X what…

29:30

I think that we we index a lot more towards larger swings as opposed to smaller optimizations.

From the episode

Head of Growth (Anthropic): “Claude is growing itself at this point”

Amol Avasare