Market Formation Patterns · cross-case synthesis
How immature markets came of age — and who actually made the money
The anchor cases calibrated the framework. The converging conclusion: the winner is not the inventor of the technology, but the one who sells the outcome, holds the moat in the enabling practice, and occupies the hidden layer (standard · component · consumption model · capital).
Gate G0 · 31 May 2026 · question originated in the conversation with Masha on 24 May · 2 / 12 studies45 sources
Alexey's working question
"What can I earn in this AI market today, tomorrow, and the day after — given the premises I've formulated?" The central premise: you can't sell the technology — you have to sell its outcome (Ford sold not the assembly line, but a cheap automobile). This dashboard is the bridge from history to the answer.
Key calibration finding
The converging pattern across both markets
Both markets ran on the same mechanics — and it directly confirms the thesis "sell the outcome, not the technology":
1. The technology (S0) is not sold
Edison didn't sell the lamp — he sold light; Ford didn't sell the assembly line/engine — he sold cheap mobility.
2. The moat is in the enabling practice (S1)
AC standard + grids; the moving assembly line. The practice is kept in-house, not put up for sale.
3. The model (S4) delivers the breakthrough
Insull — metered tariff; GM — GMAC consumer credit. The financial packaging finishes off the market.
The pioneer ≠ the one who profited. Edison was pushed out of GE; Olds left; the engine's inventors captured no margin. The money went to the operators of the model and the holders of the hidden layers (Insull, GE/Morgan, Matsushita; Ford→then GM/du Pont). The distinction Masha raised is confirmed twice over.
★ Digging beneath the surface
Where power and margin actually settled — by actor layer
On the surface ran the "wars of icons" (Edison vs. Tesla; Ford vs. everyone). The real levers lay in the hidden layers. This is the most strategically important cut:
L0 surface Icon inventors — known to all, but more often pioneers, not winners. Edison, Tesla · Benz, Ford
L1 capital Who decided who survived. Morgan stitched together GE (1892), pushing out Edison; du Pont/Raskob seized GM from Durant in the first downturn. Control shifts to capital, not the founder.
L2 rules/standards/pools Rent from rules, not from the factory. Electricity: UL+insurer = the gate of admission into the home; Insull's PUC bargain = a legal barrier to entry. Auto: SAE reduced 1,600 tube types → 221 and cut ~15% off the cost of every car; the 1915 patent pool removed the Selden "tax".
L3 hidden component The node that opens the market. Stanley's transformer (no AC without it) and the meter patent (a weapon against the competitor). Auto: Kettering's starter removed the "crank" and let in a far wider mass than any price cut; Fisher Body forced GM to buy it.
L5 consumption model / demand Financial packaging scales harder than technology. Electricity: load-building + installment plans on appliances. Auto: GMAC → 75% of cars on credit by 1930, and that is what gave GM the market, not a "better assembly line".
L6 knowledge/talent Who makes the technology replicable. Steinmetz made AC computable for GE; Sorensen (not Ford) was the real architect of the assembly line. Hidden know-how = the corporate moat.
Conclusion: profit and power settled in L1 (control) · L2 (rules/certification) · L3 (component) · L5 (consumption model) — not with the inventors (L0). "Don't sell the gold; sell the patented shovel and the rule of the game."
Bridge to strategy
Today · tomorrow · the day after
A direct transfer of the pattern to AI-2026 (in detail — in 03_SYNTHESIS_TO_STRATEGY.md; here — the working hypothesis ahead of Waves 1–3):
Today · 0–3 mo
Capture margin as the expert out front
The role of the early pioneer personally delivering the outcome via cowork (early adopters). Cash + raw material for case studies. Risk: getting stuck in a personal skill (like Igor) — non-transferable.
Tomorrow · 3–12 mo
The "Model T": a product you don't have to talk people into
Bring the "engines" to a transferable alpha (5-day sprint), road-test on a real business → case study → a sellable offer. The test: the team sells without the founder.
The day after · 12–36 mo
Hidden layer + capital
Occupy L2 (standard/acceptance of AI code) + L3 (the engine-shovel) + L5 (pay-per-result). Investment comes after the demo and case studies, at a normal valuation.
4 key positions from history — where to aim the moat in AI:
"UL for AI" — acceptance/certification of agents
regulatory bargain (shaping the rules)
the engine-shovel with IP/data
consumption model (a GMAC analog)
Distillate
5 rules for AI-2026 (from the two anchors)
1
Sell the outcome by unit of consumption (S4), not the technology (S0). "Power by the meter", not "a generator". For AI — a completed task / a closed ticket / a finished artifact on a clear metric.
2
Keep the moat in the enabling practice (S1) and don't put it up for sale. The assembly line and the AC standard stayed in-house. The "engines" go into the team's hands, but as a tool, not as a commodity called "AI technology".
3
Don't be first — be the one who crosses the parity line. Demand is proven by Olds/the pioneer; the margin is captured by the one who crashed the price under proven demand. Aim for a multiplier of single- to double-digit times on real economics, not "1000x".
4
Occupy the hidden layer. Standard/acceptance (L2) + the key component (L3) + the consumption model (L5) are more durable than yet another "model" (L0). That is where all the historical rent settled.
5
Healthy unit economics, without excess leverage. The Insull model was sound — what killed it was the holding-company pyramid (the 1932 crash). In an immature market, "money follows traction", but a superstructure bubble kills even the right model.
Gate G0 decision and what's next
Research program
The framework is calibrated on two anchors — it captures the essence. The next waves (per 02_RESEARCH_PROGRAM.md):
Wave 1 — GPT diffusion: PC (profit migrating into the OS/applications) · cloud/SaaS (computing utility) · smartphone (platform ecosystem). Where profit will settle in AI.
Wave 2 — caution: expert systems / AI winter (a direct mirror of the risk of "selling the technology too early") · dot-coms. Red flags for AI-2026.
Wave 3 — target zones: the history of SDLC tools · the data market · the productization of services. Which zone maps onto the winning patterns.
Wave 4 — synthesis: cross-case rules (5-whys) → choosing your "Model T" → the point for the 5-day sprint. The answer to the working question.
Gate G0 → awaiting your decision: APPROVE (proceed to Wave 1) · APPROVE with condition (close out the unverified points) · LOOP (adjust the framework).