Building an anti-capture environment
We're building our own tools, anti-capture down to the fibres. Cryptographic systems that give each builder control of their own experience. Unix-like, composable, mostly harmless.
We say "composable" and "anti-capture" like we've got it all figured out. We don't. Half this toolbelt is duct tape held together by good intentions, but good intentions are the root of good cryptographic properties.
A calendar server that never learns who spends time with whom.
A group chat that holds no list of who's in it.
A transaction the machine running it can't read until the moment it runs.
An agent that looks things up without accidentally revealing who's searching.
A group that can split in two without anyone having to be in the wrong.
Groups that form on mutual preference, without anyone finding out who passed on them.
We built a calendar. Then we parked it. The maths underneath turned out to answer a question Ink & Switch had left open for five years, which we found out afterwards, and by accident.
You can't plan that. You just need a shelf and people who listen to your craziness.
Shelves are underrated. A shelf costs nothing to run, holds everything at roughly the height of a person, has no opinion about what goes on it, and never once asks what your plans are.
Put something on a shelf and forgot about it for years. One day, you'll eat your breakfast, wondering about something, and notice the answer was there on your shelf, waiting for you.
One cryptography concept puts you in a room with whoever visited the same webpages in the same order. Nobody builds the room and nobody else can find it. The websites that made it possible never find out it happened. It is the crinkliest possible answer to a question nobody asked.
Unfortunately, rooms nobody can find are easy enough to find if you are willing to be in all of them at once. A scraper walks every page in every order, arrives everywhere, and waits, which it is very good at.
So you reach for a hash that costs memory. Argon2 makes every guess reserve a large block of RAM that can't be shared with the next guess. This works right up until you notice nothing stops an attacker running thousands of guesses side by side.
So you reach for a delay that can't be hurried. A verifiable delay function takes a fixed number of steps that have to happen one after another, and no amount of hardware shortens it. But it uses almost no memory, so a single chip can run thousands of them at once.
So you use both, since they're both on your shelf anyway. Each guess now costs seconds that can't be parallelised and memory that can't be shared, and having plenty of one gets you nothing off the other.
Running the current design on a phone needs eighty-four seconds of work per minute. A minute has sixty. So this design breaks here, unless you can convince a clock that time is merely a construct.
The Chia blockchain solved this puzzle in 2018, but for data storage instead of chance encounters with strangers. Their version pays in disk space instead of electricity, on the grounds that the world was already full of disk space nobody was using.
What the world did was go out and buy every hard drive in it, because now they could make profit on empty hard disks. And the economic model collapsed in spite of the cryptography being right.
Humans built the wheel, New York, wars, and venture capital. We are very proud of this.
Most of what we build here is a wheel someone else already invented. We build it again anyway, partly so nobody else can repossess it, and partly because reading someone else's code is worse than writing your own, badly.
No investors, answering to each other instead, which is principled except when your friends aren't free on a Tuesday to be answered to.
Dolphins built none of it. They just swim around and occasionally mess with puffer fish and lost sailors.
Either they never worked out capitalism, or they worked it out first.
Guessing what we're building next is basically brute-forcing a hash: technically possible and comically slow.
You can explain hash functions to a beginner using only long division. You can then use the Monster Group to explain that mathematics is the language of the universe. Nobody here will think you've been playing with puffer fish.
You might think that Truth, Value, Fairness and Debt are unrelated. But they can all be explained with Euler's identity, eiπ + 1 = 0, if you squint hard enough.
The library is open. Cryptographers are poets, infinite players redrawing the boundaries of conflict, communication and secrecy. Someone in this group wrote that sentence sincerely. We know who. We're not saying.
How we work
Transit Tuesday.
Everyone screenshares whatever they're actually working on. Your live environment, mid-thought. At least one person is contractually required to say "this was working this morning."
Screensharing is mandatory unless you have a doctor's note.
The chatroom is always there when you need it.
For some reason it's also there when you don't.
Half of us have it muted, and a stubborn remainder are still in the old Telegram chat they weren't checking when we started the Matrix one.
Every crypto community eventually produces someone with a manifesto and a token, quietly convinced they've solved civilization. It is very funny until you count your own.
None of them have tokens, which we consider a moral distinction and everyone else considers a technicality.
Walk into an Ethereum dev event and you can read the room in about four seconds. Half of them run Arch, or NixOS, or never quite left CCC. Then there's everyone in it for the merch and free money. We are, unfortunately, extremely the first half.
There used to be an official system for keeping up. Directions, Discoveries, Difficulties. Bi-weekly for full-time projects, monthly for part-time ones. It had a name and everything. Nobody follows it. We just tell each other what's going on.
We still try to dignify the phrase "testing in production."
Quantum photons are observed.
Nassim Taleb gets a mention.
Chomsky is involved.
Chomsky's point is that you can't run a controlled experiment on a country, since there's no second, identical Earth to test against. Sadly, everything we try has to be a patch on this one.
Quantum outcomes are only predictable probabilistically. You might think that's a metaphor for a roadmap. It isn't. Roadmaps themselves are, in reality, quantum in nature.
Taleb argues that people trust predictive models far more than the models deserve. We don't see any limits in that assertion.
"Test in prod" really means the code is broken but we're sick of it so we're shipping it anyway.
Your horizon can be infinite while you're standing in a finite field. (That's true of a game of snake, and it's true of most things.)
Come play
No application. No evaluators. Four questions from the heart.
This isn't a test. It's an invitation to learn something with us.
Category is: come as you are.
Why do these things have application forms? Applying to meet people you're about to be friends with is a strange thing to do. We're strange, but not that strange.
Think of the answer to these questions:
- What do you keep opening tabs about?
What do you keep opening tabs about?
Some people answered:
Tabs: alternative social media designs, comedians and their processes, permaculture.
If you opened my browser tabs right now, you'd get a chaotic mix of automated workflow error logs, multi-platform analytics dashboards, obscure game mechanics breakdowns, music DAW projects, platform appeal emails, and late-night road trip route maps.
- Anything you're building, learning, or tinkering with? Links welcome.
Anything you're building, learning, or tinkering with?
Some people answered:
I'm writing a mental poker implementation, mostly for learning some more advanced cryptography.
some signage and event tools for CCC
- Anything you've been stuck on, or powering through, right now?
- When do you feel most like an imposter?
When do you feel most like an imposter?
Some people answered:
I feel like I'm an imposter, when I'm around people who actually know how to code.
When I propose a solution that's vaguely valid in my head and then hit the reality of having to implement it.
Imposter: here, since there are so many talented coders (seriously).
Every time metrics plummet or a custom workflow breaks overnight after a platform update, I catch myself wondering if I've just been building a house of cards on pure luck.
Nobody grades your answers. You'll probably feel like you're oversharing. That's the correct amount.
This whole thing has technically been an early draft for two years. You're reading the parts that survived the last edit.
We have a theory called waves of centralisation. Decentralise, consolidate, decentralise again, historically inevitable. We have citations, but thought this ripple effect would make the point better.
No need to answer all four. Just write what comes to mind and throw it out there in the group chat:
Nobody's sure which room is the main chat at this point, but someone will read you and say hello back.
Remember to turn your matrix notifications on.
There's a call most Tuesdays, 2pm Berlin time. The link usually gets made up the same day and shared by someone in Matrix.