Thread by @alexandrosM

Could corporations be code?

A perma-🧵 to gather thoughts on what's needed to build a robot in the domain of legal personhood. A legal android, if you will.

https://twitter.com/zackkanter/status/1312425126428770304?s=20

But first, what is a corporation, anyway? Put simply, it's a legal abstraction allowing many people, with a defined relationship between them, to appear as one, in the eyes of the law (and other humans). Perhaps a better way to think of it is as a "composite person".

However, as always, the devil is in the details. The character of that composition is what ultimately gives rise to corporatism, the reduction to the lowest-common-denominator.

https://twitter.com/alexandrosM/status/1451918929782198272?s=20

It very well might be that the solution to eradicating this tendency is to make the working of companies like software: open source, transparent (at least to their participants), and automatic: make them predictable by removing the element of human "toil" from their operation.

The word "toil" is taken from the discipline of Site Reliability Enginering (SRE) and DevOps - disciplines that have, by definition, deep exposure to the real world.

https://twitter.com/zackkanter/status/1258946040926855168?s=20

More on Toil:

https://twitter.com/zackkanter/status/1258949553664876545?s=20

That way, stakeholders can always come together and discuss how to modify behavior for the future, but that consultation is the exception, not the rule. Humans should do what they do best - handle exception, and machines should do what they're good at - handle repetition.

In that world. everyone should be spending the vast majority of their time contributing directly, and the day-to-day operations of their collaborative engine should be automated.

This does, however, ask a question: what does an automated, code-first corporation look like? How do we reach a steady state that can operate as a foundation for further evolution? My initial reply to Zack lays down some hints, but let's break it down:

https://twitter.com/alexandrosM/status/1312437919672459265?s=20

Corporations are governed by some core documents. UK Limited companies (the kind I'm most familiar with) usually have Articles of Association and often an additional Shareholders Agreement (which governs how the shareholders will approach the underlying Articles of Association)

As such, problem zero is obviously to translate those documents from lawyerese into code. And that code doesn't have to be just formulas and symbols. It can be the basis for an interactive application that enables stakeholders to interact with the company and each other.

I've actually played around with converting our Shareholders' Agreement at Balena into code, and it is shockingly tractable. Think of a data structure that is accompanied with procedures of what is needed to change the data or even the schema of that data structure.

For instance, it may define how many directors are part of the Board, what the process is for changing that composition, as well as what the process is for altering the number of directors, and which directors have special powers that give them veto power over certain decisions.

That process, pretty much always, involves the very board that is defined in the data structure, as well as the shareholders, making it pleasingly self-referential.

With core governance matters settled, we have to figure out how our corporation will handle its day-to-day tasks. My preferred analogy is to think of this as a framework, in the same way as Ruby on Rails composes multiple abstractions into an overarching paradigm for automation.

So what are the kinds of activities our corporation with have to tackle? Here's some basics:

- Paying people (Including employees and contractors)
- Paying suppliers (Including one-off and recuring)
- Collecting money (Including one-off and recuring)
- Accounting & Taxes

In addition, there's more complex matters that have to be handled on the next level up:

- Investment rounds (the company selling ownership in itself)
- Hiring and firing
- Benefits to employees (healthcare, pensions, ...)
- Insurance for all involved
- Compliance w/ regulations

It will also have to manage its own property, (at the very least things like trademarks, domain names, keys, bank accounts, etc), and also manage who has access to that property.

Very much like ruby on rails composes various abstractions, it feels like, to me, this new kind of framework will have to take the best existing frameworks (Stripe Atlas, Gusto, ...?) and wire them together with something like Terraform, to create a structure that can continue...

...to operate in steady-state in perpetuity, without bothering the humans that populate it. As with everything I build from now on, I'd want it to focus on enabling its participants to benefit from participating, not necessarily controlling their behavior.

This is all I have *for now*. Feel free to respond with:

- abstractions you know may help solve part of the problem
- essential tasks I've forgotten
- pre-existing work in this field

I intend to work on this in some capacity, let me know if you have interest in contributing.

This isn't about fixing legislation but about a prior, and possibly necessary step. It talks about removal or separation of bureaucrats from bureaucracy. In addition, by making legislation formal, consistency can be assured. We don't have that now.

https://twitter.com/fiddledbits/status/1451936850571153408?t=GNITV8MwKEe2H22s-1plRg&s=19

Once you do have algorithmic governance however, alongside consistency checking and simulation tools for giving stakeholders visibility into the impacts of proposed changes, I can imagine a directly democratic approach may be far more tractable than our current options.