---
title: "I Stopped Consuming"
url: https://raskell.io/articles/i-stopped-consuming/
author: Raffael Schneider
date: 2026-12-08
description: "I stopped renting my entertainment and started building it, and it made me the polymath I always wanted to be. It is the most fun I have had."
tags: [attention-economy, games, local-first, augmentation, simulation, life-sciences, future-of-work]
---

# I Stopped Consuming

At some point this year I noticed that I had stopped consuming. Not on principle, and not as a detox. I had simply stopped opening the apps and subscriptions that used to fill my evenings, because the software I wanted to spend time with was software I had built myself.

That can sound like a sermon, and it is not one. I did not give anything up. I swapped one kind of fun for a bigger one, and it is the most fun I have had in years.

## What replaced what

The list is short and a little embarrassing in how literal it is.

| What I paid for or opened every evening | What I open now |
|---|---|
| A video streaming subscription | the things below, mostly |
| A language-learning app | [Manabi](https://manabi.raskell.io), a reading-first app for Chinese, Japanese, Korean, and Hebrew that runs offline in the browser |
| A music streaming subscription | [Driftscape](https://github.com/raskell-io/driftscape), composable ambient soundscapes, still in development |
| Mobile and console games | [Paw & Claw](https://pawandclaw.raskell.io), turn-based tactics with woodland armies, and Waterfront, a crime-management game I am designing in the spirit of [Gangsters: Organized Crime](https://en.wikipedia.org/wiki/Gangsters:_Organized_Crime) |
| Chess and Go apps | [Kishi](https://kishi.raskell.io), chess, shogi, and Go against engines I wrote and trained myself |
| A book-tracking site | the [reading log](https://raskell.io/reading/) and the reviews on this site |

None of these are better products than the ones they replaced, measured the way products are measured. Manabi has one user who matters. Kishi's board is quieter than any chess app, on purpose. My reading log has no social feed and no reading challenge. What they have instead is that every one of them works the way I think it should, and when it does not, I change it.

That is the whole difference, and it turns out to be enormous. A streaming service decides what I watch next. A language app decides how I learn, and it decided long ago that streaks and gems were the way. When the app is mine, the decisions are mine, and making those decisions is the entertaining part.

## Why it is possible now

For most of my life this would have been a fantasy. Building a language app good enough to use every day is a year of evenings. Writing three game engines and the training loop behind them is a winter, at least. Nobody with a day job does five of those for fun.

What changed is that agents write most of the code now. I wrote about the feeling in [Ideas stop waiting](https://raskell.io/notes/ideas-stop-waiting/): Kishi, which would have been a winter of evenings, was done in a month beside a day job. The bottleneck moved from typing the thing to deciding what it should be, and deciding what it should be is precisely the part I enjoy. I called that [power, in the good sense](https://raskell.io/notes/power-in-the-good-sense/): not power over anyone, but the plain capability to make a thing exist.

There is an ownership argument too, the one Nassim Nicholas Taleb makes in *Skin in the Game*, which I wrote about in [The Phrase That Stayed](https://raskell.io/articles/review-skin-in-the-game/). When I rent my entertainment, someone else owns the outcome. When I build it, I do, including the bugs, and I find I like it much better that way.

## Why it is more fun

Consuming is pleasant and passive. You sit, it plays. Building is play with the rules themselves. Instead of learning Japanese the way an app thinks I should, I get to decide how reading should be taught and then find out whether I was right. Instead of playing chess against an engine, I get to [teach an engine to play](https://raskell.io/articles/learning-to-play/) and watch it beat the version of itself from last week. And a strategy game of my own means sitting with the question that every good one answers, [who may act, when, and at what cost](https://raskell.io/notes/board-games-already-solved-agent-coordination/), and answer it my way.

The games I loved most were always the ones with systems you could see into. My [GOAT list](https://raskell.io/notes/goat-games/) is full of them, from Phantasy Star Online to Anno 1602. Building a game is the same pleasure from the other side of the table: you are not discovering how the world works, you are deciding it.

## The polymath part

There is a second half to this. Building is not only how I play now; it is how I learn.

I have always been interested in far more fields than one career leaves room for: life science, neuroscience, game development, simulation, mathematics, machine learning, robotics, space. For most of my life that curiosity ended at the bookshelf. Reading about a field is one thing; doing anything in it took a lab, a team, or years. Now it takes a project, and the project teaches me the field because the thing has to work.

| A field I always wanted to get into | What I built or wrote to get in |
|---|---|
| Life science and microbiology | [Cyanea](https://cyanea.bio), bioinformatics notebooks that re-run in the browser; reviews of [How Life Works](https://raskell.io/articles/review-how-life-works/) and [A Brief History of Intelligence](https://raskell.io/articles/review-a-brief-history-of-intelligence/) |
| Neuroscience | [Hotaru](https://hotaru.raskell.io), simulated neuron cultures playing Pong, after the [DishBrain experiment](https://doi.org/10.1016/j.neuron.2022.09.001) |
| Simulation | [Terrarium](https://terrarium.raskell.io), where LLM agents form societies, and Waterfront |
| Game development | Paw & Claw, [Rainbow River](https://rainbowriver.raskell.io), Waterfront |
| Mathematics, machine learning, game AI | Kishi and the [Learning to Play](https://raskell.io/articles/learning-to-play/) series |
| Robotics | [RoboGym](https://github.com/raskell-io/robogym), robot training in the browser, still very early |
| Smell, of all things | the note [Smell is pattern recognition](https://raskell.io/notes/smell-is-pattern-recognition/) |
| Space and aeronautics | so far only the [reading log](https://raskell.io/reading/) |

The [map](https://raskell.io/map/) of this site shows the same thing from above: one person's interests as a graph, with projects and reading and writing joined wherever they touch. When I [started the map](https://raskell.io/articles/this-site-is-becoming-a-map/) I thought of it as a way to organise writing. It has turned into the most honest description of what I do with my evenings.

The obvious objection is breadth without depth, the dilettante who knows a little about everything. I take it seriously, and my answer is that building is a strong cure for it. Reading lets you believe you understand a field. A simulation of neurons that refuses to learn, or an engine that loses 176 Elo to one constant you forgot to fit, does not let you believe anything. The thing either works or it shows you exactly where your understanding ran out.

## What it is not

It is not a productivity system. I am not optimising my evenings; I am enjoying them. It is not a moral position against Netflix either. People who love films should watch films.

There are real costs, and I would rather name them. Nobody watches my apps with me, so the shared culture of a series everyone is talking about passes me by. Building takes more energy than watching, and on a tired evening the honest choice is to read or sleep, not to open an editor. And there is a risk I watch for: tinkering forever, building tools to build tools, never finishing anything anyone else can use. My guard against that is to ship. Most of what is in the tables above is online and free to use, unfinished in places, and real enough that other people can find its faults.

I used to think of entertainment as something I bought. It turns out it is something I make, and for the first time the making is cheap enough to be the fun part.

## References

- Brett J. Kagan and others, [In vitro neurons learn and exhibit sentience when embodied in a simulated game-world](https://doi.org/10.1016/j.neuron.2022.09.001), Neuron 110(23), 3952 to 3969, 2022. The DishBrain experiment behind Hotaru.
- Nassim Nicholas Taleb, [Skin in the Game: Hidden Asymmetries in Daily Life](https://en.wikipedia.org/wiki/Skin_in_the_Game_(book)), Random House, 2018.
- Hothouse Creations, [Gangsters: Organized Crime](https://en.wikipedia.org/wiki/Gangsters:_Organized_Crime), Eidos Interactive, 1998.
- My own earlier notes this piece builds on: [Ideas stop waiting](https://raskell.io/notes/ideas-stop-waiting/), [Power, in the good sense](https://raskell.io/notes/power-in-the-good-sense/), [My GOAT games](https://raskell.io/notes/goat-games/), and [This Site Is Becoming a Map](https://raskell.io/articles/this-site-is-becoming-a-map/).
