evolution-strategy

Evolution Strategy

R — Original Text (Reading)

The disadvantage of planning is poor risk resistance, but the advantage is high efficiency, allowing tasks to be completed efficiently in a short time; while the advantage of evolution is strong risk resistance, and it can get better and better without a specific goal, but the disadvantages are low efficiency and long time consumption! The two are like the yin and yang in Chinese Tai Chi... The yin side represents evolution, silently updating and iterating in places others cannot see, continuously growing and accumulating energy; the yang side represents planning, when a specific short-term goal appears, instantly gathering all energy and exploding at one point. — Chapter 23 "Evolution Strategy"

I — Methodology Skeleton (Interpretation)

The core logic of evolution strategy:genetic variation + natural selection, neither can be missing
Four-step cycle: MVP (initial state) → natural selection (market feedback) → variation (innovation) → heredity (accumulation of advantages)
Step 1: Start the initial state (MVP). Do not hold back for a big move; build a minimum viable product around the core pain point and enter the market quickly
Step 2: Natural selection. External scissors (market feedback/user reviews) + internal scissors (horse racing system/elimination of the bottom), screening with real feedback
Step 3: Variation. Recombinant innovation (borrowing excellent elements to combine a new species) or mutation innovation ("destroying" from within to create new traits)
Step 4: Heredity. Solidify the advantages that won in natural selection into the next generation, continuously accumulating
The complementary relationship between planning and evolution: use planning to improve efficiency in parts with high certainty, and use evolution for trial and error in parts with high uncertainty
The direction of evolution is not determined by oneself, but by the environment (market)—"the product you think is good may not be accepted by the market"

A1 — Application in the Book (Past Application)

The birth of WeChat: Version 1.0 had only contacts and free text messages; a minimalist MVP was launched → iterated based on user feedback → gradually added voice, Moments, payment, and other features
Xiaomi MIUI: First made an optimized version of the Android system rather than directly making phones, faced user complaints directly through the community → iterated weekly → after MIUI matured, users actively asked for a phone → Xiaomi phones burst onto the scene
Tencent's horse racing system: Before WeChat was launched, three teams—the QQ team, the Chengdu team, and the Guangzhou R&D Institute—simultaneously built similar products, competed internally, and Zhang Xiaolong's team won while the others were disbanded
The recombinant innovation of the iPod: iTunes came from remodeling SoundJam MP, the hard drive came from Toshiba, the system came from Pixo, and the battery came from Sony—a combination of excellent genes

A2 — Trigger Scenarios (Future Trigger)

When to use:

Entrepreneurship/new product direction is unclear, and it is uncertain what users want
The market environment changes too fast, and planning cannot keep up with changes
Need to validate assumptions but resources are limited
Organizations or teams become rigid and lack vitality

Language signals:

"Not sure what the market wants"
"Just get it running first and figure it out later"
"Plans can't keep up with changes"
"How do we quickly validate an idea?"
"The team lacks innovative vitality"

Differentiation from adjacent skills:

vs. systems-thinking: Systems thinking is ananalytical tool(diagnosing system structure), while evolutionary strategy is anaction framework(designing iteration loops)
vs. innovation-method: Evolutionary strategy focuses onthe entire iteration loop(including MVP, feedback, inheritance), while innovation method focuses onthe specific techniques of the variation step
vs. scientific-decision: Scientific decision-making is used for "choosing among known options," while evolutionary strategy is used for "the exploration phase when options have not yet emerged"

E — Executable Steps (Execution)

Step 1: Start the initial state (MVP)

Identify the core pain point of the target users (focus on only the single most painful point)
Develop a usable product/solution containing the minimal feature set, centered on solving the pain point
Internet products can go live directly to run data; physical products need to be made at the current best level in one go before launch
Completion criteria: The product can be used by users in a real environment, and feedback data can be collected

Step 2: Face the market, natural selection

Launch the product into the real market and collect user feedback (likes/complaints/usage data)
External scissors: Establish user feedback channels (community/customer service/data tracking), and painfully cut features the team likes but users do not buy into
Internal scissors (optional): Set up internal competition mechanisms (horse racing/forced ranking), letting multiple solutions compete simultaneously
Completion criteria: Collect at least one round of real market feedback and complete one iteration optimization based on it

Step 3: Variation and innovation, accumulate inheritance

On the basis of the parts that won in natural selection, carry out variation and innovation (refer to the innovation-method skill)
Solidify the advantages validated as effective by the market into the next generation of the product (inheritance)
Start the next evolutionary cycle: launch → feedback → variation → inheritance
Completion criteria: Complete at least one full "launch → feedback → optimization" cycle, with quantifiable improvement metrics

B — Boundary

Scenarios not to use:

Projects with clear goals and a clear path (planning is more efficient; evolution wastes time)
Fields with extremely low tolerance for error, such as safety/medical (cannot "launch first and iterate later")
Situations where resources are extremely limited and multiple rounds of trial and error cannot be afforded

Failure modes:

MVP quality too low: Treating a half-finished product as an MVP, resulting in a terrible user experience and loss of even initial users
Variation without inheritance: Tearing everything down and starting over with each innovation, not accumulating validated advantages
No real feedback: Self-congratulating in internal analysis, afraid to face the harsh judgment of the market
Excessive horse racing: Internal competition consumes too many resources, and overall efficiency actually declines

Author's blind spot:

Lacks clear criteria for "how to judge when to use planning and when to use evolution," making it easy in practice to fall into an either-or choice
The horse-racing system is costly and unrealistic for small and medium-sized teams, but the effect gap of alternatives (such as last-place elimination) is large
There is a lack of discussion on controlling the speed of evolution—iterate too slowly and you will be overtaken by competitors; too fast and each round of feedback is insufficient

Related skills

This skill is related to the following skills:

mirror-world(Prerequisite dependency): Evolutionary strategy is essentially a strategy for dealing with the "probability domain" and "risk domain" in the mirror world. First use the mirror world framework to determine the uncertainty level of an event, then decide whether to use planning (high certainty) or evolution (high uncertainty).
innovation-method(Synergy): The "mutation" step of evolutionary strategy requires innovation methodologies to provide specific recombination-style and mutation-style innovation techniques. Evolutionary strategy governs the entire iterative cycle, while innovation methods govern "how to mutate."
compound-effect(Contrast): The compound effect emphasizes deterministic growth by "finding the causal fulcrum and persisting until the milestone," while evolutionary strategy emphasizes responding to uncertainty by "letting the environment screen directions through trial and error." The two are suited to problems of different natures.