Twitter acquired Threader! Learn more

Sarah Jamie Lewis
+ Your AuthorsArchive @SarahJamieLewis Executive Director @OpenPriv. Cryptography and Privacy Researcher. @cwtch_im icyt7rvdsdci42h6si2ibtwucdmjrlcb2ezkecuagtquiiflbkxf2cqd Sep. 04, 2021 5 min read

Need a break from research, ask me any technically oriented question and I will give you my honest, unfiltered, probably sardonic, answer.

Computation is a fundamental property of nature that humans are starting to learn to exploit but, unlike nature, we haven't learned to be good or efficient at it yet.

I subscribe to the theory that computation is mostly what the universe does i.e. the "fabric" of the universe can be described by a (probabilistic) state machine. Although I think that description says very little about emergent systems above it.

Federation is the worst of all worlds (see: )

All the good things can apply to pure peer to peer systems.

Have backups.

In the past I'd have recommend Gödel, Escher, Bach but I recently re-read it and it destroyed the nostalgia I had for it - but I don't think there is a better place to start.

Using math to prove things without revealing any more information than is necessary is just very cool and it is how society should operate in general if there is any hope of avoiding surveillance states.

C started in 1972 which was about 50 year

Although I'm hoping more Rust combined with automated theorem proving about higher level properties combined with excel/interactive notebooks.

Capabilities are a better model for peer to peer systems and shared resources in small systems.

ACLs are a better model for managing access in large organizations where you don't want manager approval flows for every single document/resource.

Yes. But good luck implementing a secure digital election system.

Turing machines.

Ah fuck that's just type theory isn't it.

No. Humans have been forged in the fires of millions of years of evolution are are still pretty shit. It's hubris to think we can replace ourselves so easily.

Yeah shared block lists, actual consent over the content we can see, powerful on-device contact management tools, community reporting tools, moderated shared spaces.

I don't think any of that needs centralized actors.

I'm concerned about what inconsiderate humans will do with what they think is "AI"

Since you said 90s the answer is all of Hackers.

No. But we've gotten very good at generating powerfully predictive statistical models from mediocre datasets.

I don't have a universal answer but I think it starts with pharmacare. There is no point in wasting patients/doctors time every 12/6/4/1 months for long term regimes many are perfectly capable of self managing. Most drugs should be more accessible.


I'm pretty sure it was by mostly being a sardonic smart arse on Twitter that does good work.

More specifically many of our larger donations came after our work on vulnerabilities impacting the Swiss Post / Scytl e-voting System 

More question please. Otherwise I'm just going to start making up questions to answer.


I think Wolfram if probably pointing in better direction than many.

On probability: I believe the universe permits both true randomness & free will.

Nope. And "lol"

Let children have (some) privacy. It's not healthy for anyone to surveil every moment of anyone else's life (especially someone who is trying to work out who they are for the first time).

This is a good question.

Bitcoin, since it is the only one I have ever really used to actually buy things in (directly).

Humans have to learn that perpetual growth is antithetical to the actual goal of development.

We need to learn to build systems that are stable and sustainable if we have any hope of making it out of the next hundred years.

Light bulbs now have wifi built-in, which is why I mostly write by candlelight.

I'm not a physicist but my understanding is that hawking radiation permits no kind of information transfer in any of the models we have that allows for it.

I've used Linux for over 20 years so no one should take any kind of computer setup advice from me.

The French presented a good model for this in 1789

Sad, but it makes up some ground because it's fairly easy to exchange for cash.

Nice question.

Anything that involves how marginalized communities, after decades of slow progress, should approach a future that is set to be worse. I struggle with those questions a lot.

I wouldn't

I've never understood the appeal or technical soundness of stablecoins. Let markets compute value.

I think multiple distribution approaches are necessary for a decentralized twitter, although distribution is the easy part.

The hard problem of a decentralized twitter is "discoverability" which can probably only be solved through "curation".

Yes. But only because I think the chaos would be good for the system as a whole.

Yes. All those approached are fundamentally limited by the available anonymity set (and the efficacy of the scheduler/mixer). And those security arguments always leave much to hope rather than math.

I don't believe in trickle down economics. And I don't necessarily think that any cryptocurrency presets a "better" economic model. However, I do think that competition to state mediated financial systems is better than the alternative.


If there was ever a demand. I have a book half sketched out which I might return to in time.

As for fundamental courses, roughly: Cryptography. Statistics. Probability. Evolutionary Algorithms. Linear Algebra. Operating Systems. Logic and Philosophy

Ideally the anonymity set would be "everyone" like in modern ZK based approaches.

In mixing approaches the anonymity set is "current transactions" and mixer efficacy is "How well can the current transaction set be randomized while preserving latency"

1. Yes.
2. The only good approach to such risks is decentralized, anti-fragile communities. Centralizing power and authority is a recipe for further disaster.
3. I'm helping hold the line. I hope.

Rust now actively stops me from doing bad things and forces me to do good things. Mostly things like that I think. And ridiculous optimizations.

Land is the only true asset and everything else is a derivative, but all economic systems are divorced from that separation permits competition of arbitrary complexity.

But because land is the only true asset: protest, occupation & revolution are viable.

There are enough known partial attacks on the sha2 family that I wouldn't be brave enough to bet against a major break in the next 2 decades.


Thanks all, late here so signing off for the night. I'll pick off any left over interesting questions tomorrow! If there are any subjects you would like longer form content of let me know and I'll consider them for my blog ( ).

You can follow @SarahJamieLewis.


Tip: mention @threader on a Twitter thread with the keyword “compile” to get a link to it.

Follow Threader