The questions we have not answered.
A research page that lists conclusions is a marketing page. This one lists the problems we consider open, because that is the honest state of the work.
Vision open. Mechanism protected.
We intend to be genuinely open about the thesis, the framing, the reasoning and the questions — and closed about implementation. That line is not a marketing posture; it is where the actual difficulty sits.
Public
- The thesis and the design horizon
- How we frame problems and choose what to work on
- The structure of each ecosystem as a real-world system
- The capability layers of the platform, at the level of what they are for
- Where we are from, and why the design environment is the right one
- How we think — the reasoning, in enough detail to be argued with
- The questions we consider unsolved, and what would settle each one
- What we have got wrong, including results that went against us
Not public
- Network design and mechanics
- Economics and incentive structure
- Onboarding and operating strategy
- Operating geography — where work is concentrated, how presence is arranged, in what order
- Models, data design and implementation
- Anything whose disclosure would only help someone skip the hard part
This boundary protects system mechanics from competitors. It is not a licence to be opaque with the people inside the system. Anyone whose activity a system of ours records is owed a clear account of what is being recorded, a real choice about it, and protection of what is held — and none of that is traded away to keep an architecture private. The two are separate obligations and we intend to hold both.
What we are actually stuck on.
Each of these is unresolved. Several may turn out to be unresolvable in the form stated, which would itself be a finding.
- Q 01
Can informal trust be represented without being destroyed?
Local trust is contextual, relational and partly unspoken. Every attempt to formalise something like it risks converting a living judgement into a score that behaves differently. What is the minimum representation that is useful and still true?
- Q 02
What is the smallest viable unit of a physical economic record?
Too coarse and it is useless for decisions. Too fine and it cannot be captured honestly at the point of action. Where is the line, and does it move by ecosystem?
- Q 03
How does identity survive migration?
Households move between economies seasonally and permanently. Standing, obligation and history need to travel with them — across regions, ecosystems and years — without becoming surveillance.
- Q 04
What makes a thin market thick enough to function?
Low-density markets lack the liquidity most market designs assume. What is the actual threshold, and what can substitute for volume below it?
- Q 05
Where should a decision be made?
Some decisions need global context, some need local knowledge that never reaches a server, and some need both. There is no general rule yet for placing a decision correctly.
- Q 06
Can local knowledge be captured without being extracted?
A great deal of economically valuable rural knowledge is undocumented. Making it legible creates value; the open question is whether the person who holds it ends up better off or merely recorded.
- Q 07
What does an existing institution lose in transition?
Transforming the function of a local institution changes it. Some of what changes is the point; some of it may be the thing that made it work. We do not yet know how to tell those apart in advance.
- Q 08
Does coordination designed for fragmentation generalise?
Our claim that this travels beyond rural India is a hypothesis. It has not been tested, and we would rather label it than assert it.
- Q 09
Does new capability actually accumulate, or is it captured?
The argument for designing an access layer assumes a loop: capability produces income, income produces capital, capital produces further capability. That loop can leak at every hop, and value can be captured by whoever holds the interface, the market access or the credit. If it leaks, the result is participation without accumulation — a far weaker claim than the one we are making. This is the most important open question we have.
- Q 10
If capability is reached without ownership, who owns the thing it is reached through?
Separating access to a technology from ownership of it lowers the capital barrier. It also makes ownership of the intermediating infrastructure decisive. The same structure can reduce dependency or manufacture it, and we do not yet know what determines which.
- Q 11
Is there a minimum viable set of conditions for conversion?
Infrastructure, skills, capital, trust, market access and a physical interface all appear necessary for a technology to become economic capability. Which of them are strictly necessary, and which can be substituted — by intelligence, by coordination, by someone else’s presence? The cost of the whole architecture depends on the answer.
- Q 12
Which parts of this arrive on their own?
Some of what we propose to build may be delivered anyway by falling costs, better models and public digital infrastructure. We cannot yet distinguish durable system gaps from temporary diffusion lags, and building for the second kind is wasted effort.
- Q 13
Can complementarity with frontier technology be made real rather than rhetorical?
We describe our work as complementary to frontier technology development. Whether that becomes genuine interoperability, or remains a positioning statement, is untested.
Three of them, stated properly.
A question is easy to list and hard to argue with. These three are set out so they can be attacked at a specific joint — including the condition that would tell us we are wrong, written before we have any result.
We intend to hold every open question in this form eventually. Stating the falsification condition in advance is the part that matters: it is nearly impossible to add afterwards with a straight face, and without it a research claim can absorb any outcome.
Does new capability accumulate, or is it captured?
- Current position
- If an access layer is built well, capability produces income, income produces capital, and capital produces further capability — a loop that compounds for people it previously excluded.
- Assumptions
- That value created at the edge is retained at the edge; that whoever operates the interface does not absorb the surplus; that credit and market access do not re-concentrate it.
- What would falsify it
- Participants whose measured activity rises over several cycles while their retained income, assets or optionality do not. That would mean we produced participation without accumulation — a far weaker claim than the one we are making.
- Strongest counterargument
- Every intermediating layer in economic history has captured a share, and the ones that captured least were usually competed into it rather than designed that way. Design intent is weak evidence.
- Status
- Open. Untested. We consider this the question most likely to invalidate the thesis.
Can informal trust be represented without being destroyed?
- Current position
- A minimal representation exists that is useful to a system and still true to how trust actually works — closer to a record of what someone has actually done than to a score of what they are worth.
- Assumptions
- That the useful part of local trust is partly observable in behaviour; that a representation can stay contextual rather than collapsing into one number; that the people represented can see and contest what is held.
- What would falsify it
- Any representation faithful enough to be useful turning out to change behaviour in the direction of the metric — people managing the record instead of the relationship. If the act of representing reliably corrupts what is represented, the minimal version does not exist.
- Strongest counterargument
- This is what credit scoring already did. It formalised something relational, the formal version became the reality, and the thing it originally measured stopped mattering.
- Status
- Open. No prototype has been tested against this.
Which parts of this arrive on their own?
- Current position
- Most of the conversion layer is a durable system gap. Falling model costs and public digital infrastructure will deliver some of it, but not identity that persists across ecosystems, not enforceability, and not presence where physical action happens.
- Assumptions
- That these gaps are structural rather than a matter of timing; that no incumbent has sufficient incentive to close them; that public infrastructure stops short of the physical layer.
- What would falsify it
- Any substantial part of the layer becoming ambient within a few years without anyone designing it deliberately. If the gap closes on its own, building for it was wasted effort — and we would rather discover that early than defend it late.
- Strongest counterargument
- Almost everyone who has claimed a durable infrastructure gap in a fast-moving technology has been wrong about the timeline, and often about the gap.
- Status
- Open, and time-sensitive. This one gets more answerable every year whether we work on it or not.
Tell us where this is wrong.
This is the mechanism behind a claim we make elsewhere on the site — that we intend to be comfortable being wrong in public. A disposition without a channel is just a sentence.
The questions above are the useful place to start, because each one names the condition that would defeat it. A challenge that lands on one of those conditions is worth considerably more to us than agreement is.
- An assumption we have not noticed we are making
- A reason a falsification condition is the wrong test
- Evidence that one of these questions is already answered
- A counterargument stronger than the one we stated ourselves
- A place where the reasoning on this site does not actually hold
What we commit to: a serious challenge gets a written answer from a person. If it changes our position, the revised thought is published with the change noted and the challenge credited, unless you would rather it was not. We are not promising to answer everything, and we would rather state a narrow commitment we can keep than a generous one we cannot.
What we will not do is discuss mechanism — the operating model, the economics, the local architecture or the geography. That boundary does not move, and a challenge aimed at it will get a polite version of this paragraph.
Three thoughts, one version each, and no evidence log yet — because there is no evidence yet. When there is, it appears here, including the results that went against us. We would rather this section stay visibly thin than pad it into looking like an active research programme.
One paper.
Written work as it is released — system analyses, design notes and findings from prototypes, including the ones that failed. Free to download, quote and circulate, with attribution and without asking.
The Conversion Problem
Why technological progress does not automatically become human capability — and what would have to be designed for it to do so.
Sets out the thesis in full: the two questions, the 1850 thought experiment, both forms of the legacy trap, the progress–access–capability–opportunity chain, the two futures, rural as proving ground, and the relationship to frontier technology. Section 10 publishes six open questions with the condition under which each would defeat the thesis. Section 11 states what the paper does not claim.
Download PDF →
It claims no results, because there are none. It contains no citations, because there is not yet a literature to cite and a paper with invented references is worse than one with none. Our own selection bias is stated inside the argument rather than left for a critic to find.
Version 1.0 · September 2026 · no evidence collected