Pareto Improving Life Policy

23 Apr, 2026

My life policy is to always improve at everything that matters.

Once I establish that something in my life matters, I am not allowed to get worse at it.

If I begin to get worse at something, I think long and hard about whether it actually matters. Usually it turns out to not have mattered much.

Why?

Let $f_{i}$ be my score on axis $i$ , say lean muscle mass, conflict resolution, writing skill, charisma, etc. Let $Δ f_{i}$ be the change in $f_{i}$ across a candidate life policy.

I see two primary ways to decide what changes to accept.

Weighted sum: accept a change only if $\sum_{i} w_{i} Δ f_{i} > 0$ , for some chosen weights $w_{i} \geq 0$ .
Pareto dominance: accept a change only if $Δ f_{i} \geq 0$ for every $i$ .

Future $w$ is unpredictable. My preferences will change over time in ways I cannot forecast.

$Δ f_{i} \geq 0$ for every $i$ implies $\sum_{i} w_{i} Δ f_{i} \geq 0$ for every $w_{i} \geq 0$ , so acting according to Pareto dominance guarantees progress in the weighted sum. That is to say, Pareto optimal is robust to future reweighting.

However, $\sum_{i} w_{i} Δ f_{i} \geq 0$ for a single $w$ implies neither $Δ f_{j} \geq 0$ for any individual $j$ , nor $\sum_{i} w_{i}^{'} Δ f_{i} \geq 0$ for any other $w^{'}$ . A weighted-sum move is a bet on the specific $w$ I chose today, and may lead to negative overall progress under a different weight regime.

What Matters?

The set of items that matters to me is small, named explicitly, and largely stable over multi-year periods. There are very few items permanently in the set: physical health, relationship with spouse/direct family/close friends, and ability to exert fluid intelligence. Right now additional items in this set include C++, tennis, and (obviously) understanding of LLMs.

I perform assessment of whether I am beginning to get worse at something over a trailing period to avoid momentary panic over temporary regressions. When I notice a sustained trend of regression on a particular axis, usually I realize I just didn't care so much and drop it 100%. This occurred, for example, with competitive video games, photography, and swimming. I would rather not do something at all than have it encroach on my constraint space.

I do remain careful to assess cross-item interactions. I do not in particular care for the activity of reading, but exerting fluid intelligence is dependent in large part on developing a corpus of experience which is partially absorbed through text. Therefore reading speed is transitively included.

I am prone to prune enjoyable but non-progressing activities. I accept this as a tradeoff for a consistent policy that is brain-dead easy to follow.

Always?

Pretty much always. I find that if I hold the principle of non-regression sacred, I am able to progress a number of items that surprises myself.

Exogenously or semi-exogenously imposed binding windows do recalibrate an axis temporarily. Recovering from a surgery renders improvement in physical health impossible. Taking care of a newborn would tank everything else for a period. Mental health episodes throw a wrench in the whole system.

Outside those windows, most apparent tradeoffs between items that matter are not actually Pareto-binding. They are joint directions I have not found yet. For example, suppose I have a high-focus task left at the end of a long day and have not yet done my reading or workout. I hit the gym first, read between sets, and handle the focus task after with post-workout clarity.

When a tradeoff does appear truly binding, I shadow-price it. Let $g (x) \leq c$ be the binding constraint, say hours in a day or units of energy. Let $λ$ , the shadow price, be what one more unit of $c$ would be worth to me in extra progress. Any action that expands $c$ (delegation, automation, better sleep, rearranging my schedule, better tools) has some cost in time, effort, or money. If the cost per unit is below $λ$ , I take the action instead of accepting the binding constraint. Given every item that matters is one I refuse to regress on, $λ$ is large, and most "I can't do both" situations fail this test.

The binding policy creates subconscious search pressure, both for joint directions to find and for constraints to relax. The overall Pareto improving policy sustains itself, and the longer it runs the more my life reshapes around it.