Reject Change, Sometimes
Kent Beck · email · 2026-09-02 Reader 🖥️
Because a growing product's payoff structure differs by 3X phase — roughly symmetric win/loss during Extract, sharply upside-skewed during Explore, and ruin-inclusive (win/die/switch) during Expand — the correct exposure to volatility differs sharply too: Extract calls for the disciplined, rebalancing 'Shannon's Demon' strategy of protecting revenue and keeping changes reversible even at some cost to upside (i.e., deliberately rejecting some change), Explore calls for going all in on volatility despite that same strategy's mathematical dominance in symmetric bets, and Expand calls for a third move — investing directly to reduce the probability of catastrophic failure — so 'embrace change' is not a universal software-development virtue but a phase-specific one.
mixed · mixed — A narrative parable (three traders/Shannon's Demon) establishes a mathematical/theoretical result about rebalancing strategies under uncertainty, then the essay argumentatively maps that result onto Beck's own prior 3X (Explore/Expand/Extract) framework to prescribe a different, practical volatility posture per phase. An appendix supplies runnable code that makes the theoretical claim empirically checkable.
Structure
| Part | Role |
|---|---|
| Prudence/Reckless/Shannon's-Demon (SD) parable | narrative hook establishing the puzzle |
| "What?" section: naming Shannon's Demon and worked 2-day sequences | expository/mathematical explanation |
| 3X: Extract | argumentative application |
| 3X: Explore | argumentative application with a modified model |
| 3X: Expand | argumentative application with a further-enriched model |
| Appendix: Simulator | empirical/falsifiability device |
Key terms
- Shannon's Demon (SD)
A strategy of rebalancing a stake evenly between a safe holding and a symmetric double-or-half bet after every round, which compounds value over repeated rounds even though each individual bet has zero expected value change. - long volatility / short volatility
Being 'long volatility' (Reckless) means benefiting from large swings by staying fully exposed; being 'short volatility' (SD) means benefiting from calm/rebalancing and giving up some upside to avoid downside. - 3X: Explore / Expand / Extract
Beck's pre-existing three-phase model of how a growing idea/product moves from finding a growth loop (Explore) to scaling past bottlenecks (Expand) to harvesting profit (Extract); this essay assigns each phase its own volatility posture. - Prudence / Reckless / SD
Three named trader-personas embodying three strategies: never bet (Prudence), always bet everything (Reckless), rebalance after every bet (SD). - rebalancing
The mechanical act of resetting the split between safe and risky holdings after each round; the specific action that produces Shannon's Demon's edge and that Beck maps onto 'keep changes reversible as much as possible' in Extract.
Propositions
- explicitGiven a symmetric double-or-half coin flipper, a strategy that rebalances half the stake into the flipper after every round (Shannon's Demon) outperforms both an all-safe strategy (Prudence) and an all-in strategy (Reckless) over repeated rounds, even though each individual bet has zero expected value change.conf: high
- explicitIn a payoff structure skewed toward a much larger win than loss (e.g. +300%/-33%), the all-in 'Reckless' strategy outperforms Shannon's Demon-style rebalancing, because the occasional large win dominates the outcome even after strings of losses.conf: high
- explicitSoftware product development's three phases (Extract, Explore, Expand) present three structurally different payoff regimes: Extract resembles the symmetric double-or-half flipper, Explore resembles the skewed (+300%/-33%) flipper, and Expand adds a third, catastrophic ('death') outcome alongside a growth win.conf: highextendsCanon 3X: Explore/Expand/Extract · p4 — Elaborates the prior 3X essay's claim that each phase requires its own distinct approach by supplying the specific volatility/risk-strategy dimension (symmetric-flip/short-vol for Extract, skewed-flip/long-vol for Explore, ruin-avoidance for Expand) that the earlier essay left unspecified.
- explicitBecause Extract's payoff is roughly symmetric, the correct strategy is Shannon's-Demon-style rebalancing: protect the revenue stream, take some growth bets, lower costs, and keep changes reversible as much as possible — i.e. deliberately reject some change even at the cost of some potential upside.conf: high
- explicitBecause Explore's payoff is asymmetric (large upside, tolerable downside), the correct strategy is to go 'all in' on the flipper rather than rebalance or hedge — 'Reckless' behavior is not actually irrational in this regime.conf: high
- explicitBecause Expand introduces an irrecoverable losing outcome ('death') that erases all future upside, its correct strategy is neither pure rebalancing nor pure all-in betting, but investment directly aimed at reducing the probability of the catastrophic outcome and/or increasing the probability of clearing the current growth bottleneck.conf: mediumextendsCanon 3X: Explore/Expand/Extract · p7 — Adds a specific mechanism — investing to reduce the probability of a catastrophic 'death' outcome, alongside increasing the probability of clearing the bottleneck — to the earlier essay's more general claim that Expand's core risk is failing to scale and the response is focus on the current growth bottleneck.
- implicitThe right degree of embracing versus rejecting change/volatility in software product development is not a universal constant ('always embrace change') but a function of which of the three 3X phases a given product or team is currently in.conf: high
- explicitThe interesting strategic question about a risky bet is not whether to take it at all (an ROI/expected-value calculation) but how to play it (bet sizing, rebalancing cadence, hedging structure).conf: high
- implicitA runnable simulation showing the median outcome over many repeated trials makes an otherwise counterintuitive mathematical claim (that a zero-EV-per-step rebalancing strategy beats both extremes) directly checkable rather than merely asserted.conf: medium
Arguments
In Extract-phase software product work, teams should favor a Shannon's-Demon-style strategy — protecting revenue, taking modest bets, lowering costs, keeping changes reversible — over embracing change indiscriminately. strength: medium
Explicit premises:
- Extract's change payoff resembles a symmetric double-or-half flipper (some changes gain customers/revenue, some lose them, at comparable magnitude).
- Given a symmetric flipper, rebalancing after every round mathematically outperforms both all-in and all-safe strategies over repeated rounds (shown via the worked sequences and the simulator).
Implicit premises:
- Product changes in a mature, monetizing product line are, in aggregate, close enough to symmetric-payoff events for the coin-flipper analogy to transfer.
- The goal in Extract is compounding value over many rounds, not maximizing any single round's payoff.
Inferences:
- Analogical mapping from a trading strategy proven optimal for symmetric zero-EV bets to a software-strategy prescription for Extract-phase changes.
In Explore-phase software work, teams should go all in on volatile, potentially-failing bets rather than hedge or rebalance. strength: medium
Explicit premises:
- Explore's payoff is asymmetric: wins are much larger (e.g. +300%) than losses (e.g. -33%).
- Under a sufficiently upside-skewed payoff, an all-in strategy outperforms rebalancing, and a string of losses is an acceptable, low-cost outcome ('who cares... we'll move on').
Implicit premises:
- A team can survive/afford repeated small losses long enough to reach the rare large win — i.e. Explore bets are individually cheap or bounded even though each one is played 'all in'.
- Moving on to a new idea after a losing streak carries negligible cost relative to the foregone upside of not swinging fully at winners.
Inferences:
- Same analogical transfer as the Extract argument, applied to the skewed-payoff case.
In Expand-phase software work, the priority strategy is investing to reduce the probability of catastrophic failure ('death'), not rebalancing or all-in betting. strength: low
Explicit premises:
- Expand's payoff structure includes a third, irrecoverable outcome (death) that erases all future value, unlike Extract's and Explore's payoffs.
- Neither Shannon's-Demon rebalancing nor all-in betting is designed to manage an irrecoverable-loss outcome; a different strategy — reducing the probability of death, or increasing the probability of clearing the growth bottleneck — is needed instead.
Implicit premises:
- Once an irrecoverable-ruin outcome is added to a repeated-bet model, the earlier bet-sizing logic (rebalance vs. all-in) no longer applies because the game cannot be repeated indefinitely after ruin.
- Engineering and operational investments (performance tuning, resourcing, backup & recovery) causally reduce a scaling product's probability of death, though this causal link is asserted rather than modeled.
Inferences:
- Extension of the model by adding a third possible outcome, loosely evocative of ruin/ergodicity reasoning (Taleb-adjacent) though not cited or derived mathematically as the first two cases were.
Narratives
- The Prudence/Reckless/Shannon's-Demon (SD) parable of three traders at an ancient exchange choosing daily between a strongbox and a double-or-half coin flipper.
→ Establishes, via a simple concrete scenario, the counterintuitive core finding that a rebalancing (partial, reversible) strategy beats both a fully cautious and a fully reckless strategy over time — the analogy the rest of the essay maps onto software. - The worked 2-day win/lose and lose/win numeric walkthroughs starting from 100 coins (Reckless ends flat either order; SD ends at 112.5 either order).
→ Makes the abstract Shannon's Demon claim concrete and checkable by hand before generalizing to skewed payoffs and the 3X mapping. - The Appendix simulator: runnable JavaScript computing the median outcome over 1000 trials of 100 daily coin flips for Prudence, Reckless, and Demon strategies.
→ Offers falsifiable, runnable proof of the parable's claim and signals Beck's own epistemic style ('I program so I can understand the world').
Questions addressed
- ✅ Why would a strategy (SD) with the same expected value per bet as an all-in strategy (Reckless) end up producing more value over repeated bets?
Rebalancing 'banks' part of a win before the next round's downside is applied to it, so a loss following a win costs less in absolute terms than the win gained — a volatility-harvesting effect shown numerically and via the simulator. - ✅ Is 'embrace change' (long volatility) always the right stance for software product development, as earlier posts in this series seemed to imply?
No — it depends on the 3X phase; Extract specifically calls for going short volatility (rejecting/hedging change) rather than embracing it, which is the essay's titular point. - ✅ What determines whether an all-in ('Reckless') strategy or a rebalancing (SD) strategy is better for a given bet?
The payoff skew: symmetric payoffs favor rebalancing; sufficiently upside-skewed payoffs favor all-in. - ✅ How should a team behave when a bet carries risk of irrecoverable ruin, not just symmetric or skewed win/loss?
Beck proposes a third strategy — direct investment in reducing the probability of the catastrophic outcome, or increasing the probability of clearing the bottleneck — but does not quantify it or show it optimal the way the first two strategies are shown; he explicitly calls this part of the model still unripe. - ⭕ Exactly which real software-development activities correspond to a 'coin flip' in each phase, and what tells a team which payoff regime a specific proposed change actually falls into?
Gaps
- incompleteThe essay never specifies how a team is supposed to determine, for any actual proposed change, which of the three payoff regimes (Extract's symmetric flip, Explore's skewed flip, or Expand's win/die/switch) applies — the entire prescriptive value of the framework depends on this classification step, which is left as an exercise.
Read next: Watch for a follow-up Beck post operationalizing the classification (a checklist, heuristic, or worked real-world case); or pressure-test the framework against a real proposed change in your own team's Extract-phase product to see whether 'which regime is this?' has an obvious answer in practice. - incompleteExpand's strategy ('invest to reduce probability of death, or increase probability of clearing the bottleneck') is asserted without the same numeric/simulated grounding given to Extract and Explore — no payoff numbers, no simulator branch, no worked example of the tradeoff between the two sub-strategies.
Read next: Watch for Beck extending the Appendix simulator to model a three-outcome (win/die/switch) game, which would let the Expand claim be tested the same way the Extract/Explore claims already are. - uninformedBeck imports the volatility/rebalancing framing directly from Kunal Bhalla ('pointed out the exception... introduced me to... Shannon's Demon') but gives no independent check that the coin-flipper model captures the actual statistical structure of real product changes, versus being an elegant trading analogy that happens to sound applicable.
Read next: Look into the empirical distribution of 'did this change help or hurt' outcomes for real product changes (e.g. from experimentation/A-B-testing literature) to see whether they're closer to symmetric, skewed, or something else entirely — and whether that distribution actually differs by company stage the way Beck assumes. - incompleteThis essay finally explains the 'long volatility' claim that an earlier post in this series ('The Beginnings of an Idea: XP is Long Volatility') left undefined — but it delivers the general 3X (Explore/Expand/Extract) framing, not XP specifically; it never closes the loop back to concrete XP practices (pairing, TDD, continuous integration, spikes) as instances of 'short' or 'long' volatility bets the way the original teaser promised.
Read next: Watch for Beck to explicitly re-connect this volatility/Shannon's-Demon framework to specific XP practices (is TDD a 'rebalancing' practice? is a spike a 'Reckless' Explore bet?) to fully deliver on the original 'XP is long volatility' framing.
Closes gaps elsewhere
- The Beginnings of an Idea: XP is Long Volatility · g1 — Supplies the definition of 'long volatility' (Reckless's all-in strategy) versus 'short volatility' (Shannon's Demon's rebalancing strategy) that g1 flagged as missing, and maps both onto software product development via the pre-existing 3X (Explore/Expand/Extract) framework — closing most of the explanatory debt from the earlier teaser post, though it explains the general 3X model rather than XP-specific practices (see this extraction's g4).