1 · Intellectual lineage
The adjacent possible and its effective subset are established ideas.
Stuart Kauffman's adjacent possible describes possibilities reachable from a present state, with realized novelty opening further possibilities. The concept has since spread across innovation studies and other fields.
More directly, Josef Taalbi's recent innovation research already distinguishes the universe of possible recombinations from an effective adjacent possible constrained by search complexity, absorptive/search capacity and resources. Agalmic Research therefore does not claim the term or distinction as its invention.
James March's exploration/exploitation framework adds another essential constraint: organizations must allocate scarce resources between exploring new possibilities and exploiting what is already known.
The local contribution sought here is synthesis and application to machine-amplified cognition, not ownership of the adjacent-possible framework.
2 · Candidate abundance is not progress
A larger menu can still leave us hungry.
Large language models and other cognitive tools can produce candidate hypotheses, designs, arguments, representations and implementation paths at low marginal cost. That increases the supply of possibilities visible to a user.
But a possibility becomes practically valuable only through additional functions: recognition of significance, lineage checking, validation, selection, coordination, handoff and realization. When those capacities do not scale with generation, more ideas can increase epistemic congestion rather than innovation.
This is why the programme treats candidate production as an input rather than an output measure.
3 · A recursive frontier
Realization changes what becomes reachable next.
The adjacent-possible tradition emphasizes that actualizing one possibility can expose further possibilities. A useful innovation therefore has value not only in its immediate payoff but in the new paths it enables.
Discovery-production loop
state → search → recognition → validation → realization → new state → new adjacent possible
The important object is not the idea alone. It is the capacity of the whole system to continue this loop productively.
For Agalmic Research, this suggests evaluating projects partly by frontier leverage: whether a successful result creates reusable capability, lowers the cost of later exploration, exposes valuable new branches or enables other actors to reach possibilities that were previously inaccessible to them.
“Frontier leverage” is local operating language, not yet a validated economic quantity.
4 · What cognitive tools can change
Search is only one lever.
Cognitive tools may alter the effective adjacent possible by changing several different capacities:
- generation: producing or recombining more candidate possibilities;
- representation: making latent structure easier to perceive;
- recognition: helping identify which candidates are unusual or significant;
- evaluation: lowering the cost of criticism, simulation, calculation or comparison;
- routing: identifying the expertise or institution needed for the next epistemic step;
- realization: reducing the cost of code, design, coordination or implementation.
The most important complement may differ by domain. Cheap generation can make evaluation scarce; cheap analysis can make physical experimentation scarce; cheap invention can make coordination or capital scarce.
5 · A programme criterion
Judge research by capacity created, not pages produced.
The practical question for an Agalmic Research project becomes:
If scarce attention is spent here, what worthwhile future becomes more reachable?
A research object can answer well even if it is small. A literature map can reduce search cost. A negative result can close an expensive branch. A provenance protocol can enable distributed research. A prototype can reveal whether an idea survives reality. A handoff can connect a low-authority insight to competent validation.
Conversely, a sophisticated paper that mostly adds another attractive idea to an already overflowing candidate pool may have low frontier value.
6 · Research questions
Turn the meta aim into something testable.
- How should the effective adjacent possible of a human-machine research system be operationalized without pretending to enumerate unknown possibilities?
- Which cognitive tools enlarge search breadth versus recognition, evaluation or realization capacity?
- When does increased candidate generation reduce effective innovation because selection and validation cannot keep pace?
- Which interventions create the greatest reusable frontier leverage per unit of scarce attention or capital?
- Can provenance and handoff systems measurably increase the fraction of promising low-authority candidates that reach competent evaluation?
- How should institutions balance exploration of new possibilities against exploitation and realization of existing ones under machine-amplified search?
7 · Reflexive application
The project should embody the criterion.
Agalmic Research now separates its large preserved possibility set from a deliberately small Active Frontier. Ideas can remain valuable as options without being promoted into papers or projects. Public notes are expected to explain what capacity their publication creates.
The meta aim is therefore operational as well as rhetorical:
Preserve abundant possibilities. Allocate scarce commitment toward the branches that most improve the capacity to reach worthwhile futures.
References & antecedents
- March, J. G. (1991). Exploration and Exploitation in Organizational Learning. Organization Science, 2(1), 71–87. doi:10.1287/orsc.2.1.71
- Björneborn, L. (2020). Adjacent Possible. In The Palgrave Encyclopedia of the Possible. Reference.
- Taalbi, J. (2026). Long-run patterns in the discovery of the adjacent possible. Industrial and Corporate Change, 35(1), 123–149. doi:10.1093/icc/dtaf028
- Weitzman, M. L. (1998). Recombinant Growth. Quarterly Journal of Economics, 113(2), 331–360. doi:10.1162/003355398555595
- Cohen, W. M. & Levinthal, D. A. (1990). Absorptive Capacity: A New Perspective on Learning and Innovation. Administrative Science Quarterly, 35(1), 128–152. doi:10.2307/2393553
This is a synthesis note, not an exhaustive adjacent-possible literature review. The “effective adjacent possible” is explicitly attributed to existing innovation research rather than claimed as local novelty.