This post is largely an output from ChatGPT because I wanted to save myself a lot of the effort of writing it all out, which I find too tedious.

I’m going to do something a little spicy: compress a big chunk of what people call “ADHD traits” and “autism traits” into one decision equation.

This is not a clinical definition, and it’s not saying these conditions are “just math.” It’s a useful model—a way to describe the logic of why certain situations derail you, and why certain interventions work.

If you’re the kind of person who keeps thinking “I know what to do, but I can’t reliably make myself do it,” this may be the most honest framing:
your brain is choosing the action that wins according to its internal scoring function.

So let’s write the scoring function.

The monolithic expression

At any moment, you choose an action a (work, rest, plan, scroll, leave the room, text someone, etc.) that maximizes a subjective value:

Read it like this:

Pick the action with the most “good stuff” minus the most “bad stuff,” adjusted for time, uncertainty, sensory load, and randomness.

That’s it.

What each term means in real life

1) Discounted rewards: why “later benefits” lose to “now relief”

• r_t(a) is the reward or benefit you expect from an action over time.
• D_k(t) is the discount function: how much future rewards “count” right now.
• A common shape is hyperbolic discounting:

Plain English:
If “future benefits don’t feel real enough,” then actions whose payoff comes later (studying, long projects, life admin) lose to actions whose payoff is immediate (scrolling, snacks, novelty, games, arguments, research rabbit holes).

A larger k = stronger present-bias.

2) Execution costs: why starting feels impossible

This splits “cost” into different kinds of pain:

• C_init: the cost of starting (activation energy / identity threat / dread)
• C_switch: the cost of switching contexts
• C_effort: the cost of doing the work

Plain English:
Sometimes you’re not avoiding effort—you’re avoiding initiation. The first 30 seconds can cost more than the next 20 minutes.

3) Sensory + prediction-error cost: why environments can melt you

• δ(a) = mismatch between expectation and input (noise, interruptions, unpredictability, “things aren’t as they should be”)
• π_s = how heavily sensory evidence/mismatch is weighted
• α = how punishing mismatch feels

Plain English:
If your system weights sensory mismatch strongly, then chaotic environments carry a real penalty. You’re not “being dramatic”; your cost term is literally larger. That makes leaving the environment or shutting it out a rational move in the model.

4) Uncertainty/ambiguity cost: why you plan forever

• H(a) = ambiguity/entropy: “how unclear is the next step?”
• λ = how expensive uncertainty feels right now

Plain English:
If ambiguity is expensive, you’ll try to reduce ambiguity before acting. That can look like over-researching, over-planning, or “I can’t commit until I’m sure.” Your brain is minimizing H, not maximizing progress.

5) Noise/variability: why consistency is hard even when you care

This term is a placeholder for volatility: sleep, stress, stimulation, random fluctuations, hormonal cycles, social context—anything that makes “the same you” behave differently day to day.

Plain English:
Some days the same task is easy. Some days it’s impossible. That’s not hypocrisy; it’s a noisy control system.

Where ADHD and autism live inside this equation

This is the key idea:

ADHD-like and autism-like traits can be described as different parameter settings in the same general decision function.

ADHD-like settings often involve

• higher k: future rewards feel weaker right now
• higher C_init and/or C_switch: starting/switching costs more
• higher ε: more variability in control
• plus a tendency for immediate rewards to dominate when stressed

Translation:
ADHD looks like an agent that struggles to “buy” long-term outcomes with present effort unless you make rewards immediate, reduce initiation cost, and stabilize attention.

Autism-like settings often involve

• higher π_s and/or α: sensory mismatch costs more
• higher λ: ambiguity is more expensive
• social contexts often have higher H: lots of unspoken rules and uncertainty

Translation:
Autism looks like an agent for whom sensory chaos and ambiguity are objectively costly, so predictability, explicit structure, and environment control are rational and stabilizing.

The combo (ADHD + autism traits together)

This is where life gets spicy:

• you get high initiation cost + high uncertainty cost + high sensory cost, and
• immediate relief options are always available.

So the model repeatedly chooses relief unless you engineer the environment and reward structure.

Turning the equation into a practical guide

This is the part that matters: you don’t need to “become a different person.”
You need to shift the terms so the action you want wins.

Lever A: Make future rewards feel immediate (lower the effective discounting)

Target: the discounted reward term.

Practical moves:

• micro-deadlines (today’s payoff is “I did it,” not “someday it helps”)
• visible proof artifacts (a checkbox, a commit, a one-paragraph trail note)
• earned rewards right after effort

In equation language: make the immediate payoff of work larger.

Lever B: Reduce initiation cost (the “2-minute start” is not a meme)

Target: C_init

Practical moves:

• 2-minute start rule (start, stop on purpose)
• pre-staging (open the tab, lay out tools, remove choice)
• templates (same starting ritual every time)

Lever C: Reduce ambiguity (stop paying the uncertainty tax)

Target: λH

Practical moves:

• define a single next action
• define done in one sentence
• timebox planning to 2–10 minutes
• break projects into atomic steps

Make the next step so obvious it feels almost trivial.

Lever D: Control sensory mismatch (environment isn’t “extra,” it’s part of the math)

Target: the sensory mismatch term in the equation:

Practical moves:

• headphones / predictable audio
• consistent workspace
• leaving the environment when it spikes (walk / car / library)
• reducing interruptions (boundary phrases, do-not-disturb)

Lever E: Treat variability as real (design for bad days, not ideal days)

Target: ε

Practical moves:

• have Tier 1 tasks that succeed even on trash days
• rely on minimum viable progress, not ideal output
• measure consistency by “did I run something?”

A simple “debug protocol” when you derail

When you catch yourself doomscrolling or freezing, ask:

Which term spiked?

• If it’s starting: initiation cost spiked → do a 2-minute start.
• If it’s uncertainty: ambiguity spiked → define the next action, timebox planning.
• If it’s environment: mismatch spiked → change location / headphones.
• If it’s immediate relief winning: raise friction on escape, add earned reward.

The punchline

You don’t have to beat yourself up for “not doing what you know.”

In this model, you are doing exactly what any agent does:
choosing the action with the best score.

Your job is to redesign the situation so the action you want has the best score.

That’s not self-help. That’s control theory for a human nervous system.