An implementation of tasteweb concepts, to be used for mapping respect within a particular research field.

A well designed timer, for phones.

Comes with three games for getting good with time.

A language, framework, and suite of exceptional software and components designed to make composable user-owned software frictionless to use and produce.

It could be thought of as a new kind of radically extensible forum/chat/wiki/twitter/reddit, which users can add features to without having to wait for permission from the hosts. But it's many other things as well.

A cohabitive game is a multiplayer game that's not purely competitive, nor purely cooperative. A little anarchic dojo where players can develop the art, craft and social rhythms of thriving in pluralistic conditions. I hope that cohabitive games can help laypeople and theorists both to develop their intuitions for cooperative bargaining, because it is an infrastructural social literacy which is today nowhere taught. It seems that the best way to learn it is through play.

Cohabitive games are likely to end up being better games in general, supporting a richer variety of relationships and dynamics between player characters, and loosening the constraints on balancing, by decoupling power level from available strategic depth.

the cohabitive prototype: Optimal Weave 0.1

design principles for cohabitive games

The use of overlapping Webs of Trust with well defined meanings to appoint the curators of content tags in a way that is easy to grow, easy to collectively moderate, and when it needs to be, subjective. This is the only moderation system suited to web-wide annotation systems ("reply-anywhere"), and it makes tagging systems much more practical than they previously have been.

Encompasses an engineering effort to develop graph database indexes that permit new uses for webs of trust; fast distance queries and sparse segment queries.

As well as the development of "radically extensible" social application frameworks.

A writeup on the unexplored breadth of the applications of webs of trust.

Applying these concepts requires a discourse platform that would have them. Right now, I don't see any great projects in that space. So tasteweb is waiting on/part of modular web.

github

A package that significantly improves the way mice do scrolling. I believe that if you try this, you will decide to keep using it for the rest of your life.

Currently only for linux. A windows version would make sense, but I'm probably not going to make it (help wanted). Mac users probably wont care much about it though since their touch-based scrolling modalities are already very good. (I think it's still better than those, but only a little bit)

worldbuild.ai/to-light/

A vivid, detailed, plausible rendition of a future-history in which our world meets with the worst contingencies in the development of AI, and survives. The first-winning entry in the Future of Life Institute's 2022 positive worldbuilding contest.

Told through short stories, a timeline of events, and a Q&A.

It discusses the scenario where capabilities arrive abruptly but alignment is hard, in which the world needs to manage unprecedented levels of global coordination. I propose that humanity will not survive this outcome unless it's prepared to produce very clear demonstrations of the presence of danger immediately after it manifests.

So I outline a technical programme for relatively safely demonstrating the presence of danger without unleashing it, and an international response that follows it. I have doubts about these solutions, there are always doubts, we'll just have to sit down and do the work if we want to see whether they're resolvable.

There's are also some depictions of success at the end.

For non-fictional (... is a prophesy fictional?) discussion of this area, I wrote this little page to explain why some of us must prepare for super-human AI before it arrives.

You may also want to read the appendix, a non-fictional exploration of some of the core ideas from the worldbuild, with much more thorough discussion of the technical program.

In the years since, third party auditing protocols have been implemented in the real world, and there is at least discussion of international cooperation.

My recommendation to all people is to live to see the future. There's a decent chance that the world succeeds in creating better agents of human wishes.

A short story/thought experiment/speculative scifi about why we need a concept of indexical measure (or observer-measure) that's related to but conceptually distinct from the agency behaviors of consciousness. It wasn't totally clear to me that there existed a hard problem of consciousness until I found this thing. I think it clarifies a lot.

the story

tweets

A short story set in an era of long reflection, where things have mostly turned out alright. Written as a part of Issue 1 of the sleep-advocacy project; Zs (it got in Zs because it depicts a time and place where it's very much alright to relax, extols the virtues of napping) https://somnotica.com/issue-1/being-patient-with-challenging-reading.html

A vaguely interactive abstract frustration about memetic evolution and agency. (android app)

bokik.ttf

A Glance Font. A font reduced to the simplest and most pronounced shapes. I made bokik to test the hypothesis that a fluent reader of a glance font would be able to read faster or with fewer errors (or assumptions) than can be had with traditional typefaces. As far as I'm aware, no one has ever become a fluent reader. Please tell me if this changes.

It looks like this. I expect that becoming fluent in bokik wont be easy, people don't read letter by letter, they read word by word, so it doesn't just require you to learn 26 things, it may end up requiring you to learn around 50,000.

I've made a trainer program for learning to read bokik, but I haven't given it a serious attempt, myself. While making it, it occurred to me that there are a few changes I'd like to make to bokik, before I learn it. Next time I stumble across a decent font editor. Until then, I recommend the trainer program only to the adventurous.

The trainer program itself tests another hypothesis: Do humans learn efficiently as a result of seeing inputs, attempting a prediction, then being shown the output? This is how artificial neural networks learn, and reading bokik is the kind of simple sensory processing task that ANNs do, so I'd expect this to work for humans too? I'm excited to find out.

It's canon that Link met the goddess of wisdom, prior to Link's Awakening. What might have happened if he had lingered longer in her company and become an unbearable consequentialist goth with strong feelings about simulation ethics, before arriving at grim Koholint? Do not read this, it's awful. There is very little design in it. Some people seem to enjoy it. Epub version

A puzzle game about drawing glyphs and developing correct theories through experimentation. A story about two futures.

A media scene.