The Last Human Edge

Second Level Thinking · Issue 50

Revisiting Annie Duke’s thinking in bets, and why, as AI makes execution cheap and abundant, the quality of your decisions is becoming the one advantage that still genuinely belongs to you.

This is the fiftieth issue of Second Level Thinking. Fifty newsletters of turning ideas over with you. When I sat down to think about what deserved the slot, I kept coming back to the subject I’ve circled more than any other, the one I devoted two early issues to, back when this newsletter was finding its feet: Decision-making, Thinking in bets, and more specifically, Annie Duke.

Back in issue #4 I gave you the machinery, that was the three Ps and a six-step framework from Duke’s How to Decide. In issue #26 I gave you the mindset: the Pete Carroll Super Bowl call that looked like the worst decision in NFL history and was in fact a perfectly sound one that simply got a bad result; hindsight bias; the truth-seeking charter; why we’re all credulous creatures who find it far easier to believe than to doubt.

I want to return to it, but emphatically not repeat it. I think there’s something genuinely new worth saying, something that has only become obvious in the last couple of years. So this is a synthesis and a sharpening, not a re-run.

Let me start with a confession, because it’s the cleanest way in.

I’ve been resulting

A few months ago I made a call in one of my businesses that turned out poorly. Money and time, both gone. And in the days afterward I did the thing I know not to do, the thing I have literally written to you about not doing: I let the bad outcome convince me it had been a bad decision. I re-ran it, winced and assigned myself the blame.

Then I did the honest audit and walked back through what I actually knew at the moment of choosing; the information I had, the odds as they genuinely stood. And the uncomfortable truth was that the decision had been sound. Given what was knowable at the time, I’d choose it again. It simply landed in the unlucky slice of the distribution.

Annie Duke has a name for the error I’d fallen into. She calls it resulting: judging the quality of a decision by the quality of its outcome. It’s the single most seductive mistake in all of human judgment, and the fact that I teach this stuff didn’t make me immune to it. That’s worth knowing. These aren’t ideas you learn once and own forever. They’re a practice.

Poker, not chess

Here’s the foundation the whole thing rests on. You’ve heard me lay it out before, so I’ll be quick.

Chess is a game of pure skill where every piece is in the open, nothing hidden, the better player reliably winning over time. Life isn’t chess. Life is poker: incomplete information and irreducible luck, where you can play a hand perfectly and still lose. Duke’s central move is to prise apart the two things we compulsively glue together — the quality of the decision and the quality of the outcome.

Two forces determine how any moment turns out: the quality of your decisions, and luck. You control exactly one of them. What I want to add this time is a sharper tool for actually using that idea, because “don’t result” is easy to say and hard to do.

So draw this: a simple two-by-two grid. Decision quality on one axis (good or bad). Outcome on the other (good or bad). Every result you’ve ever had lands in one of four boxes. The two that feel intuitive - good/good and bad/bad - teach you nothing dangerous. It’s the off-diagonal boxes where all the learning and all the traps live. Good decision, bad outcome is just variance: hold your process, don’t torch it. Bad decision, good outcome is the silent killer. This is the lucky win that quietly trains you to do something reckless again, and again, until the luck runs out. If the only question you ever ask is “did it work?”, you literally cannot tell a skilled win from a lucky one. Start plotting your own decisions on it and patterns appear fast: the person who lives in the bad-decision/good-outcome box is simply someone whose reckoning hasn’t arrived yet.

Why this suddenly matters more than it did

Now the new part, the reason I wanted to bring this back for the fiftieth issue.

For most of history, two things determined who got ahead: the quality of your judgment, and your ability to execute on it. You could have brilliant decisions and be undone by poor execution, or paper over mediocre judgment with sheer operational horsepower. The two were braided together, and both were scarce.

Execution is now becoming cheap. This is the quiet earthquake underneath everything I’ve been writing about lately — the flattening org charts, the small teams doing the work of large ones. Increasingly, the doing — the drafting, the modelling, the coding, the researching, the making — can be spun up on demand, in abundance, at a fraction of the old cost. When execution is scarce, it’s a differentiator. When execution becomes a commodity, it stops being one.

And when the cost of doing things collapses, the value migrates almost entirely to choosing which things to do. If you can execute ten times as many ideas as before, the binding constraint is no longer capacity but the quality of the bets you place with all that new capacity. Point cheap, abundant execution at the wrong decisions and you don’t win; you just make expensive mistakes faster and more efficiently than ever.

This is why I’ve come to think decision-making isn’t just a useful skill. It’s becoming the skill; the last durable human edge in a world that is automating almost everything downstream of the choice.

The operating system, sharpened

So let me pull both of those early issues together into something you can actually run. Not a summary but a six step sequence.

1. Name the bet. Every meaningful decision is, in Duke’s framing, a bet on an uncertain future. Say it out loud in those terms: “I’m betting that if I do X, Y is likely to follow.” The word “bet” does something clever, it flips on a scepticism most of us keep switched off. The moment real stakes are named, you stop nodding along to your own assumptions.

2. Widen the frame — the three Ps. This is straight from the earlier work and still the core mechanic: for each option, lay out your preferences (what outcomes you actually want, given your values), the payoffs (how big the upside and downside really are — magnitude, not just direction), and the probabilities (how likely each outcome is). A pros-and-cons list ignores size and odds. A rough decision tree, even sketched on a napkin, forces both. Don’t wait for certainty to estimate the probabilities, as the willingness to guess an explicit number is itself the skill.

3. Match your effort to stakes × reversibility. This is the piece I’d most want to add to my earlier self. Not every decision deserves the full treatment. Borrow Jeff Bezos’s distinction between one-way and two-way doors. A two-way door is reversible — if it’s wrong, you walk back through. Decide those fast; agonising over them is its own kind of waste. A one-way door is irreversible or expensive to undo. Those earn the full process, the slept-on night, the dissent. Most people get this exactly backwards by labouring over trivial reversible choices while rushing the handful that actually can’t be taken back. Spend your judgment where it’s load-bearing.

4. Recruit dissent, on purpose. Your own brain is a machine for confirming what it already believes; we covered why in issue #26, the credulous-by-default wiring that makes believing easy and doubting hard. The fix is external: people incentivised to tell you the truth, operating under what I called the truth-seeking charter. Target accuracy over comfort, real accountability, genuine openness. The practical version: when you test a big decision, bring in one person it affects and one it doesn’t, and don’t reveal your own lean until they’ve formed theirs. You want, in Duke’s phrase, two to disagree and one to referee.

5. Time-travel before you commit. You already have the 10/10/10 rule from last time — ten minutes, ten months, ten years — for dragging a decision out of the heat of the moment. Add two more, pulling in opposite directions. A premortem: assume it’s a year from now and the decision has failed badly, then work backwards and list every reason why. It surfaces the risks optimism hides. And a backcast: assume it succeeded beautifully, and work backwards to what had to be true to get there. One finds the landmines, the other finds the path, and the gap between the two is usually your real plan.

6. Decide at 80%, then commit. Waiting for certainty is just procrastination wearing a lab coat. Get to roughly eighty per cent of the knowledge you could theoretically gather, then choose — and here’s the discipline — commit to the choice even while acknowledging it might be wrong. This is Bezos’s “disagree and commit,” and it’s the operator’s version of what Sandeep Swadia keeps returning to: indecision is not a neutral, safe holding pattern. It is itself a decision — usually the worst available one — made by default, with the clock running the whole time.

The other half nobody teaches: un-making the bet

There’s a flip side to deciding well that I skipped entirely in those early issues, and I don’t want to skip it now, because it might be the harder skill.

Knowing when to quit.

Duke wrote a whole book on it, and her central finding is counterintuitive but, once you see it, undeniable: because quitting carries such social shame, we almost always quit late, never early. By the time walking away feels justified, you’re usually well past the point where you should have gone. The grit we lionise — the refusal to ever give up — quietly costs us more than the quitting we’re ashamed of.

The culprit is the sunk cost. That is, the time, money and, most treacherously, the effort you’ve already poured in, which is gone whether you continue or not, but which your mind insists on counting anyway. I know this one intimately. I can drop a monetary sunk cost in seconds; a concert ticket I couldn’t use barely registered. But a venture I’ve bled effort and identity into? Letting that go is one of the hardest things I’ve had to learn to do, even when every rational signal says the smart bet now is to fold and redeploy.

Here’s the reframe that helps me: quitting a bad bet isn’t the failure. Continuing to feed it is. A great decision-maker isn’t someone who never gets it wrong. It’s someone who notices they’re wrong faster than everyone else, and has the discipline to un-make the bet while there’s still something left to redeploy.

Your actionable takeaways

If you take nothing else from this fiftieth issue, take these.

1. Run a decision journal. Before a significant call, write down — in one paragraph — what you’re deciding, what you expect to happen, and roughly how confident you are. Date it. This single habit is the most powerful defence against hindsight bias ever devised, because it freezes what you actually believed before you knew the outcome. When the result comes in, you can finally grade the decision, not the luck.

2. Draw the two-by-two. Next time something goes well or badly, before you celebrate or self-flagellate, ask which box it’s in. Was it a good decision, or a good outcome? They are not the same, and telling them apart is where all the learning lives.

3. Sort your next five decisions into doors. One-way or two-way? Reversible or not? Speed up the reversible ones ruthlessly. Slow down and run the full process only on the ones you can’t take back.

4. Build a two-person test. For your next big call, recruit one person it affects and one it doesn’t. Ask for their read before you reveal yours. Reward them for disagreeing with you.

5. Name one bet you should already be quitting. You almost certainly know what it is. Ask the cleanest question in all of decision-making: knowing what I know now, if I weren’t already in this, would I choose to start it today? If the answer is no, you have your answer, the only thing left to decide is how quickly you act on it.

So, here’s what I’d leave you with after fifty issues: if execution is becoming free and judgment is becoming everything, what’s the single most important bet you’re making right now, and when did you last actually examine it, rather than just hope it works out?

Parting Shot

Fifty issues ago I set out to be useful. If I’ve managed to leave you with one idea across all of them, let it be this one, because it sits underneath everything else:

A great decision is not a great outcome. A great decision is the product of a sound process, and any sound process has to begin with the humility to admit, out loud, “I’m not sure.” Everything I’ve written about thinking, deciding, and betting is really just an elaboration of that single, uncomfortable sentence.

Here’s to the next fifty.