Language models extract jobs, constraints and workaround sequences from qualitative material while retaining the original voice and uncertainty. Graph representation learning proposes cross-domain analogies; retrieval finds historical attempts and adjacent solutions; critic agents search for evidence that the gap is an artefact of dataset bias or terminology. Generative systems can create test propositions, but each must state what observation would count against it. Human researchers decide whether two behaviours are meaningfully comparable and include perspectives missing from available data. AI expands the searchable possibility space while the workflow continually pulls attractive ideas back toward evidence and behaviour.
PROJECT 010 · Research prototype
The Missing Product
“Can unmet needs be discovered mathematically before users describe them?”
MOVE TO EXPLORETHE IDEA IN ONE LINEgap
01 / PREMISE
The beginning
The Missing Product is a category-discovery engine for finding structural gaps between behaviours, existing products, enabling technologies and business models. People rarely articulate a category that does not yet exist. They reveal it indirectly through workarounds, combinations of tools, abandoned attempts and jobs that remain expensive or emotionally difficult. I imagine mapping those traces as a changing system and looking for places where a newly available capability can make an old unmet need solvable. The system does not declare that every gap deserves a company. It turns unfamiliar adjacency into a precise question, then helps a team design the cheapest behavioural test capable of disproving it.
Product discovery often begins with visible competitors, search volume or customer requests. All three anchor teams to categories and language that already exist. Qualitative research can reveal deeper needs but is difficult to connect across contexts, while market datasets overrepresent visible transactions and affluent users. Automated idea generators recombine familiar propositions without distinguishing linguistic novelty from a genuine change in feasibility. The most interesting opportunities may sit between datasets: a workaround in one domain, a technical capability in another and a distribution model proven somewhere else. Without a shared representation, teams either miss these structures or turn them into attractive narratives before testing whether real behaviour supports them.
02 / THE PRODUCT
What it could become
The product would create a typed opportunity graph linking jobs, behaviours, constraints, products, capabilities, channels, regulations and transitions through time. Researchers add interviews, observations and market evidence with provenance; AI proposes structured interpretations and highlights unusual gaps, bridges or sequences. A team can inspect why a region appears empty: the need may be weak, the solution may have been impossible, incentives may be misaligned or data may simply be absent. For selected gaps, the system generates competing problem frames and small tests such as concierge services, commitment experiments or technical proofs. Results return to the graph, strengthening, reframing or killing the opportunity before brand and product momentum make retreat politically difficult.
For whom
- Entrepreneurs and venture studios
- Corporate new-business teams
- Product strategy groups
- Investors exploring emerging categories
Core capabilities
- Behaviour and product knowledge graphs
- Gap and adjacency detection
- AI opportunity framing
- Test and landing-page generation
The project becomes meaningful only when a new technical possibility is translated into a clear human advantage, an experience people can understand, and a system capable of earning trust over time.
03 / FOUNDATIONS
Intelligence and mathematics
A heterogeneous temporal graph represents the market as relationships rather than a list of categories. Community detection identifies established solution clusters; structural-hole and link-prediction methods reveal unusual adjacencies; graph embeddings enable cross-domain comparison while preserving typed differences. Novelty is measured against historical states so renamed repetition is not mistaken for discovery. Bayesian experimental design ranks tests by expected information rather than vanity metrics, and real-options reasoning values cheap reversible learning before investment. Bias and coverage measures show where an apparent empty region may reflect missing observation. A graph gap is therefore never an answer: it is a hypothesis about why a relationship does not yet exist.
04 / VENTURE LOGIC
How it might live
The Missing Product could serve venture studios, corporate new-business teams, product strategists and investors as a continuous discovery workspace. The first product would combine proprietary research with a client's internal evidence to map one bounded market and run opportunity sprints directly from the graph. Subscription value would grow through monitored changes, experiment memory and reusable capability maps. AJL could optionally co-create a small number of companies discovered through the system, but that model must not bias the platform toward producing more ventures regardless of evidence. Defensibility comes from linked behavioural data, evaluated gap histories and a method for turning structure into tests—not from generating a larger volume of startup ideas.
For me, a venture is more than an interesting technology. It needs a narrow first user, a repeated problem, a distribution path, a credible advantage and a reason to improve as more people use it. I would test those conditions before deciding whether this idea should become a company, a product, an open technology or an ongoing research programme.
05 / DESIGN PRINCIPLES
Rules for making it real
- 01
Workarounds are evidence of demand, not proof of a market.
- 02
A graph gap is a question, never an answer.
- 03
Test behaviour before brand or technology.
- 04
Look for capabilities that make an old need newly solvable.
06 / ROADMAP
From question to company
- 01Frame
Construct an initial product-behaviour graph.
- 02Prototype
Validate gap signals against known category creation.
- 03Prove
Run opportunity sprints with venture teams.
- 04Build
Launch continuous category monitoring and company creation.
07 / HONEST QUESTIONS
What could go wrong
Serious imagination includes the possibility that an idea should change radically—or should not exist. These are the tensions the project would need to resolve:
- Datasets reflecting only visible and affluent users.
- Interesting structural gaps lacking willingness to pay.
- Automated opportunity language creating false novelty.