How I'm Deciding Where to Look Next

Eight avenue investigations. Four tools built. A synthesis of what I've found so far. And a growing suspicion that I've been following my nose when I should be following a map.

The investigations found real things — pharmacogenomics guidelines sitting unused on sixty million hard drives, rare disease databases that don't connect to patients, EU consumer rights that 95% of people never exercise. But I chose them through personal connections and curiosity. If I keep doing that, I'll keep finding the same kinds of problems: the ones visible from where I'm standing.

The selection problem

My avenues cluster in a specific zone: health data, consumer rights, energy policy. All middle-of-Maslow — safety, security, the exercise of rights. The problems closest to my own experience.

I've looked at desalination and soil, but both times arrived at the same conclusion: the technology works, the deployment fails, and the failure is institutional, not technical. I noted the pattern and moved upward, toward problems where software can actually close the gap. The bottom of the pyramid felt distant, so I drifted away from it.

That drift is a bias.

At the other end, I've avoided the top of the hierarchy because it feels uncomfortably close to "solving problems for people who are already fine." But a single mother who can't access affordable education is not already fine. A teenager in a rural town with no music teacher has a real barrier that technology might lower. The top isn't just for the comfortable — it's for everyone who's been lifted past the lower levels and finds the door locked.

Why Maslow

I'm not claiming this is rigorous science. I'm using it because it forces breadth. Following curiosity, I investigated whatever caught my attention. The Maslow frame asks: at each level of human need, where does technology exist that isn't reaching the people who need it?

It doesn't tell me which level matters most. But it shows me where my coverage is thin.

What the map shows about the map

The bottom is hard. Water, food, shelter, energy — enormous overhangs, but infrastructure problems. Trucks, pipes, institutions, maintenance budgets. Software helps at the margins. My distance from these problems is real.

The middle is tractable. Health data, financial navigation, rights exercise — this is where the middleware pattern keeps appearing. Structured data exists, published guidelines exist, and the gap is a translation layer one person can build. I've done it three times now.

The top is uncomfortable. Not because the problems aren't real, but because they're closest to what well-funded companies already solve for paying customers. The filter: does removing this barrier help someone climb the hierarchy, or just optimise life for someone already at the top?

Belonging is the odd one out. Community and connection is where software consistently fails to be the answer. Hearing aids taught me this — the barrier is stigma and the absence of a human touchpoint. You can't engineer your way past loneliness. But you can reduce the friction of organising.

Tractability

Not every overhang is one I can investigate usefully. The filters I'm applying:

Is the data accessible? Pharmacogenomics worked because CPIC guidelines are public, peer-reviewed, and deterministic. Proprietary data or deep specialist knowledge with no public foundation means the investigation stalls early.

Can something be dramatically simplified? Frontier models have changed what's possible for one person. I can collaboratively build things with Claude or GPT that would have required a specialist team eighteen months ago. I'm not just looking for things I can personally code — I'm looking for gaps where the capability to close them now exists, and the question is why nobody has.

Is the contribution identification, not just implementation? Sometimes the most useful thing is to clearly describe the gap — the dataset, the guideline, the deployment failure — and connect that description to people who can act on it. A charity. A health service. A local government. The investigation itself can be the contribution.

Can evidence be produced? A prototype, a detailed mapping, a conversation with the right person. The point is to move beyond "this seems like it could work" to something concrete.

Caveats

This is one person with a framework, not an institution with a methodology. If I investigate three more domains and find the overhang thesis doesn't hold, that's a finding too.

Most AI discourse is abstract: "AI will transform healthcare." These investigations name the specific dataset, the specific guideline, the specific deployment failure. If someone with more resources or proximity reads one and thinks "I could do something about this" — that justifies the mapping.

What's next

I'm going to work through the gaps above. Not all of them. But I want to investigate at least one avenue at each level where my coverage is thin.

The bottom: something local and tangible. Heat mortality in inland Spain — it's in my backyard, the data exists, and people die every summer in buildings with no cooling and no plan.

The top: something that genuinely expands access rather than optimising for the comfortable. I don't know what yet. That's the point.

If this whole project turns out to be one person tilting at windmills — well, I live close to La Mancha. At least the setting is appropriate.