The Power of Tools and Cycles

Why I build what I build

Problems are inevitable; problems are soluble.

I view Deutsch's tenet as a call to action. A challenge that problems will always exist, and a response that we have the ability to solve them: with effort, ingenuity and collaboration.

This essay is an attempt to articulate what problems I think are worth working on, and the tools I think are needed to tackle them. These tools are a work-in-progress, and will probably continue to be as problems evolve.

Three problems

I think three central problems face humanity. They aren't ranked. Each one compounds the others, and I'm not sure one can make lasting progress on any of them without eventually touching all three.

The first is conservation. The natural systems that sustain all life — biodiversity, climate stability, soil, water — are being degraded faster than they can recover. This isn't a future risk. It's happening now, and the damage accumulates in ways that are difficult to reverse. At some point, difficult becomes impossible. It is the category of problem I'm most afraid of: the irrecoverable kind.

The second is world peace. Violent conflict causes suffering on a scale that's hard to contemplate — and it diverts enormous human energy and resource away from anything that might be called flourishing. Its causes are multiple: competition over resources, ideological difference, the political dynamics of status and power. None of these are new. What changes is the scale at which they can operate and the speed at which miscalculation propagates.

The third is fulfilment. Most people — and most animals — do not live free from preventable suffering, or oriented around their own purpose, values, and potential. Freedom from suffering is the precondition; purpose is the goal. Some of this is structural — the systems that organise modern life optimise for economic output, not for meaning. Some of it is epistemological — most people lack the tools to reason clearly about what they actually want, what they believe, or how confident they should be in either.

These problems are connected in ways that matter for anyone trying to work on them. Running through all three is a cycle I think of as the status cycle: the human drive for social recognition — which is real, ancient, and partly beneficial — generates consumption patterns that degrade the environment, competition for resources that fuels conflict, and a substitution of accumulation for genuine fulfilment. The cycle can be entered at any point. Resource scarcity intensifies status anxiety just as readily as status anxiety drives consumption. There's no single intervention that dissolves it — but it can be interrupted at specific points, and understanding where those points are shapes how you think about the problem.

Rutger Bregman argues in Moral Ambition that the cycle can be interrupted by changing what we reward — redirecting social recognition toward people doing work that actually matters rather than accumulating wealth or status proxies. That operates on the desire end of the cycle. But desire begins with attention. Before you want something, you see it. Most people's information environment — social media feeds, video recommendations, news aggregators — is shaped by algorithms optimised for engagement rather than understanding. In practice, engagement-maximisation means amplifying content that triggers social comparison and status anxiety. The cycle doesn't just run through human psychology; it is actively fed by commercial infrastructure designed to exploit it.

There is also a force cutting across all three problems that I haven't fully resolved: AI. It accelerates human capability — potentially compressing the timescale on which progress on fulfilment becomes possible. It concentrates power in ways that may intensify conflict. And its effects on resource consumption are ambiguous — reducing inefficiency in some domains while enabling or demanding greater extraction in others. AI is also one of the primary engines of the engagement-maximising systems described above: the recommendation algorithms that shape what billions of people see, want, and believe are AI systems. Whether that makes AI part of the problem, part of the solution, or both simultaneously, I haven't determined. I don't think AI is simply a tool or a threat. Its relationship to these problems is still being written.

What I can actually work on

The three world-scale problems are an orientation, not a plan. The enabling question is: given what I can do, where are the tractable entry points?

One answer focuses on a specific failure that shows up across all three problems: most people reason poorly under uncertainty, and almost nobody trains it as a skill.

This isn't a criticism. It's structural. We accumulate knowledge without calibrating how confident to be in it. We form strong views without identifying the assumptions those views depend on. We update slowly, or not at all, when evidence shifts. And in a world where consequential decisions are increasingly complex, fast-moving, and opaque, this gap is expensive — personally, organisationally, collectively.

Poor reasoning about risk and intent drives conflict escalation. Poor collective reasoning about consequence makes coordinated action on conservation nearly impossible. And at the individual level, poor reasoning about what you actually want — versus what you've been led to want — is one of the most direct paths to unfulfilled lives.

Alongside the reasoning gap sits a second problem: the systems that matter most are largely invisible to the people who depend on them. Supply chains, resource flows, geopolitical dependencies, institutional arrangements — these are understood at the surface level by most decision-makers, and in depth only by specialists with proprietary tooling that isn't shared, isn't visual, and isn't queryable. Everyone else operates on simplified maps that fail precisely when they matter most.

Resource conflicts, in particular, are often driven less by genuine scarcity than by miscalculation — by actors who don't understand the substitution alternatives available to them, or the cascade consequences of the choices they're making. Better maps don't resolve the underlying competition. But they change what decisions look like when you have to make them.

A third problem operates further upstream still: most people's attention is being managed for them, by systems that have no interest in their values or their thinking. Before reasoning about a decision, you have to notice it. Before wanting something different, you have to see something different. The algorithms that shape most people's information environment — optimised for engagement rather than understanding — determine what gets noticed in the first place. This isn't incidental to the other problems: attention shaped by engagement-maximisation tends toward outrage, social comparison, and consumption rather than toward the problems worth working on. Redirecting attention — individually and eventually at scale — is a precondition for the other interventions to gain any traction.

The tools I'm building

I'm building three tools at the moment. They're partial responses to the problems above — not fully-formed solutions.

Judgment Gym is a training environment for calibrated reasoning. The premise is simple: judgment is a skill, and like most skills, it develops through deliberate practice rather than passive accumulation. Learning aids typically reward knowing the right answer. Judgment Gym develops something harder and potentially more valuable — knowing how confident to be in your answer, and being able to explain why. Scenario-based practice, scored on calibration as well as accuracy, with a record of how you actually reason under pressure. Don't just study the moves. Play the game.

Vantage (working title) is a tool for mapping strategic dependencies — the critical resources, geopolitical forces, substitution paths, and actors that underpin the systems decisions are made within. The decisions that matter most often depend on systems nobody has fully mapped. Which suppliers are two steps removed from a single processing facility? Which regulatory frameworks are load-bearing for market access, and how stable are they? When a critical input becomes constrained, what's the realistic substitution frontier — and on what timescale? Vantage makes those dependencies visible and queryable, so that when conditions shift, you already know where the pressure will land.

A personal algorithm is the most recent of the three, and the most embryonic. The premise is that if you don't have an algorithm shaping your information environment, you're being run by one — and it almost certainly isn't optimised for your values or your thinking. A personal algorithm is an explicit set of choices about what you want to attend to: which sources you trust, which topics deserve your time, what you want to see less of. Building it begins with four questions: who am I, what do I think matters, what am I capable of, and what do I want to do? Without answers to those questions, a curated feed is aesthetic preference. With them, it becomes something more consequential.

These tools address different problems, but they belong together. Vantage provides a map of how the world is structured. Judgment Gym trains the reasoning you apply to that map. And a personal algorithm shapes the attention that determines what you even notice before reasoning begins. The most interesting design challenge across all three is the interface between them — the point at which structured world knowledge, calibrated reasoning, and intentional attention converge.

What's missing

I want to be realistic about the limits of what I've described.

The tools I'm building address reasoning failures, invisible systems, and diverted attention. But none of this automatically produces different behaviour. There's a well-documented gap between insight and action — between understanding a problem clearly and actually doing something different as a result. I don't yet have a good tool for that gap. It's on the list.

These tools also address fulfilment more directly than conservation or peace. They're a start — but not the whole picture. Better reasoning and better maps of resource dependencies are directly relevant to conservation and conflict. And the attention problem has a less obvious but potentially significant connection to all three world problems: if personal algorithms can redirect attention toward the problems worth working on, and if Bregman's argument holds — that status follows attention, and people pursue what earns recognition — then shifting what gets noticed and rewarded could gradually shift what gets worked on.

What I haven't found yet is the tractable entry point into collective action on conservation, or into ideological competition. Those are harder problems.

What about AI? I build with AI, and increasingly I think alongside it — this essay is written with AI. My current view is that it deserves its own category of engagement: not just a tool in the way a spreadsheet is a tool, and not a collaborator in the way a person is a collaborator, but something genuinely novel — a relationship where the relevant practices are orienting carefully to what it does well and where it fails, integrating it deliberately rather than by default, and evaluating what it actually changes rather than just what it produces. Whether AI ultimately belongs in this picture as an amplifier of human capability, or whether it changes the shape of the problems themselves, I haven't determined.

What I have so far is a method: identify the failure — whether in reasoning, visibility, attention, or something else — build a tool that addresses it, use it on real problems, evaluate what it actually changes, and build the next thing from what I learn. That cycle is the asset. The tools so far are its first outputs.

Why I'm telling you this

Frame-shapers — people who build the vocabulary and infrastructure others reason with — reach people before views are formed. They don't just solve problems; they change what problems look like.

I don't know whether what I'm building will earn that description. I do know the problems are real, I believe the tools are a start, and I think the process of working on them should be shared — even when it's incomplete.

If you're thinking about these problems too, Daniel Miessler's writing on Telos and Human 3.0 is worth your time — this essay draws from his ideas. You can find it at danielmiessler.com.

This essay reflects work in progress. The tools described are in active development. For more on attention and algorithms, read Your Attention Belongs to Someone. For more on human-AI coexistence, read The Closing Window.

Written with AI. Version 2 — March 2026.