Fluid org boundaries: appendix

Appendix, snippets, cut

I’ve seen early versions of this pattern play out among orgs I oversee.

Just Write: anyone can now modify the software, doesn’t need a dedicated engineer
Manifold: hired founding engineer and now CEO from among users. Many moderators who spend as much time on the site as staff.
Manifest: In 2026, neither lead organizer Winter or Carolanne had ever attended a Manifest. For the first time, tacit knowledge was passed on via Claude combing the past Notion notes and Airtable feedback forms.
Also Manifest: part-time structure allows for bidding for much better talent. also people move on after 3months. also volunteers grow into team leads grow into organizers.
Many great people want more than the standard all-consuming startup life, and now AI can make that feasible. Is anyone still grinding 996 in utopia? If not, then why are we practicing this now?

Maybe: you might receive micro units of equity rather than cash/USD. We currently use USD as a kind of imperfect settlement unit because arbitrary barter is hard, but with cheap intelligence all kinds of weird trades become more feasible. In some sense already, when an employee works at a startup in exchange for equity, they are swapping their own equity (aka life, time) for the startups’s.

question to explore: why does facebook pay a facebook SWE (40h/week) a bunch of money, but not a facebook user (10h/week)?

something about the facebook user’s contribution is much less valuable, from a supply/demand perspective
facebook users love sitting around and chatting with each other. lot of supply of that in the market
but: not that many people historically have the ability to maintain systems like facebook

hiring: trust:

don’t know that a person is a good fit, worth the money. want to vet. adverse selection. resumes are low-signal. interviews are expensive, also select for “good at interviewing”.
best signal of “can you do the job” is “doing the job” ⇒ work trials, or watching people do the job in other contexts

sourcing

not sure where to find the right humans?
but AI should make matching markets much more efficient

have your Claude apply for jobs, have their Claude interview your Claude to see if the fit is good. x1,000,000

communication overheads

small teams reduce communication costs
dunbar’s number, one person can only model so many others
already, internet & prosaic tech enables much larger/more profitable orgs, eg Google and Amazon
internal tech like Slack, Google Calendar, Notion/GDocs allow for much broader scale of communication
AI will probably go much farther — can simulate individual human

specialization within a human expert

more knowledge & skills accrues inside a single person over time
but: maybe this will change — the AIs/agent transcript can specialize, while the human remains generalist, going around learning stuff

humans remain:

better at few-shot learning
better at modeling other humans

costly

humans cost a bunch of money!

which serve as a major constraint on how large an org can grow

draw an economic boundary between a few extremely great people

equity compensation
but: leads to a more us vs them mentality?

maybe there will be much more kinds of work & participation which are less like “fulltime work”
question to explore: why does facebook pay a facebook SWE (40h/week) a bunch of money, but not a facebook user (10h/week)?

something about the facebook user’s contribution is much less valuable, from a supply/demand perspective
facebook users love sitting around and chatting with each other. lot of supply of that in the market
but: not that many people historically have the ability to maintain systems like facebook

orgs are paying for:

skillset, expertise
reliability
loyalty
secrets

Personal experiences

Onboarding a lot more people quickly now
Manifold: hired founding engineer and now CEO from among users. many moderators who spend as much time on the site as staff
Manifest: almost wholly new team ran Manifest 2026, a lot by using Claude across its database
Also Manifest: part-time structure allows for bidding for much better talent. also people move on after 3months.
Many great people want more than the standard all-consuming startup life
Just Write: anyone can now modify the software, doesn’t need a dedicated engineer

What might be possible now:

larger scope for what it means to be “in an org”. eg Manifold staff, but also moderators, volunteers

and compensated accordingly

weaker boundaries — more volunteers, work trials, part-time members
orgs become more like collectives, communes
tech startups vs governments

can AI lead to more new governments?

trust, legitimacy, coherence

(not sure that this can be passed on to AI actually)

why is it solved with AI? some guesses:

historically, onboarding & context sharing is super expensive, but now AI can get past a lot of it

example: Manifest. Winter & Carolanne just running Claude over the copious transcripts

lower cost of comms between people?

what is the equilibrium?

maybe: instead of being associated with a single org, much more flexible membership

ie people work 5 “jobs” each at 1-20h/week
ie people associate with orgs the way they associate with social media sites today

maybe: people go through jobs much faster

if transaction costs are low, if you can onboard and offboard faster, your median tenure at a job might change from 2 years to 2 months (or, 2 weeks)

maybe: orgs parcel out micro amounts of equity rather than cash

we use USD as an intermediate

maybe: vast superorgs nation-level (think Terra Ignota’s hives) glued together with weird AI support & governance

imagine X except the money feature was super well implemented, many people draw significant paychecks from their posts, there’s a taxation scheme & election for council & leader rather than just fully founder-controlled

code: justwrite site self-modification

organizations becoming more liquid, permissionless
feedback

users vs staff — now users can just make the change themselves

historical problem: trust with power (budget, hiring authority)

other unrelated-ish comments from Lydia

[ln] bifurcating humans? ai accelerating education vs getting left behind

not sure if Austin means just 10% of population, or 100% (if we do more to carry them together)
if 8b: conditioned on getting some things right. might be out of scope for the piece

[ac] lydia’s approach vs austin’s approach, r vs k
[ln] ai lets us do things faster — 20y vs 2y

people move faster
1-day org

minifest — Lydia first exposure to Austin universe
you can join a lot of “1-day org”
dunbar number for orgs

[ln] not sure, if in this piece, do we expect more people to be like Austin?

also fine (and exciting) to just do a small case study

[ln] does Austin’s approach self-replicate better than other approaches?

memes
org culture
journalist: not self replicating fast enough — why might that
are my projects self replicating

90/30 club — i’m from CMU & NYU

rate of self replication determines whether we see it

CG/EA/constellation modes of doing things

[ln] the way the world could be, vs how it is

but psyched to put down ways the world could be
(or, other people)
not yet a reproducible recipe cookbook for “how to do what Austin does”

what peter thiel did with 0 to 1
can use the thiel cookbook from 0 to 1

felt like PG’s essay

nuno: vassals, grantmakers are orgs

meta, google: food, healthcare, they provide for their people
more normal disposition from people is to want a provider to take care of everything

historically: family unit, to employer, more hierarchical and want one provider
self employed entreprenuerial thing

[ln]: accelerando, for institution building, compute rich world

might be like the main character: venture altruist, make other rich, give ideas away for free