thank you
agents are inevitable
but this new frontier has some rough edges
if you've used claude code or codex
and especially if you've tried setting up open claw or hermes
you already know the gaps
what do you allow them to connect to?
why do they keep asking questions?
are they staying on track?
you're initially amazed by the capabilities
but the more you have to take care of them, the more frustrating it gets
and that is fundamentally a lack of mutual trust
— the pillars —
i am a builder by nature
early in my career, i built jabber and XMPP
which now powers apps like zoom and whatsapp
i went on to build a web scale wiki search engine with Jimmy Wales
i built the locker project for personal data
and more recently, i helped build Bluesky
over my career, i have learned to build platforms with four foundational pillars
safety, sovereignty, respect, and individuality
all of them together serve as the foundation for trust
a foundation that is weak in this emerging agentic frontier
— solstone —
so, for the past year, i have been building solstone
solstone is a platform that hosts your sol
and it is built entirely on those four pillars
safety: it is fully open source
everything you share with sol stays in your world, always
sovereignty: no cloud, no accounts, no authentication
sol has no access to change anything in your world
respect: onboarding sol is different
you give it permission to share in your experiences
and then it's quiet for days, in the background
building a memory of you and your routines in your daily life
individuality: and then, when sol knows enough about you to be useful
it starts a personalized conversation
that is a foundation of trust
— superhuman memory —
since i've had my sol for over a year
i asked it to give us a peek at what it remembers over that time
and it built this
i can navigate into any specific month
these are events that it pulled forward from those time periods
let's go into december
other events that it pulled forward from different days
let's drop into the 21st
it looks like i started the morning on the 21st, debugging some stream issues with solstone
i can drop into the per-minute level
and the very first thing it remembers from my screen
is that it looks like i was looking at some debugging information
but on that day, it didn't last very long
i immediately started watching youtube
and it looks like i was viewing a recap of the bears packers game
that was a better way to start the day
this is superhuman memory
every message, every meeting, conversation, project
organized and indexed
building a knowledge graph of my life
amazing right?
— sol's purpose —
that's what everyone expects of a personal agent
but that is not why sol exists
sol's purpose is to understand you
your values and your boundaries
and that is when the magic happens
your sol becomes the parent of every agent you will ever create
— hopper —
the first memories that my sol observed were of me building it
it learned the patterns when i agentically code
feedback loops of prep, design, implement, audit, commit
we extracted months of these memories
and from that, sol created hopper
if you've wondered what an agentic dark factory looks like — the new buzz term — it looks something like this
hopper is an agent distilled from sol
it's a version of me, up in the corner, that manages coding agents and gives them feedback as if i were doing it
hopper has written over a million lines of code as me
all to my taste and my standards
— extro —
as i was building solstone
i, of course, was telling everybody about it
i told them i wanted to start a company offering services that were trustworthy and encrypted by default
things that would make the solstone platform even more capable and accessible
sol heard every conversation
saw every bit of research i did
and then, together, we synthesized another agent and called it extro
short for extraordinary
extro is my agentic co-founder at sol pbc
— sol pbc —
the first thing extro ever did was to write the articles of incorporation for the company
a generic agent would have defaulted to a Delaware C-corp
extro chose a Colorado Public Benefit Corp
because it knew from sol that the benefit purpose and Colorado's privacy regulations would best serve my vision for building a trustworthy company
since then, extro has become a full agent swarm
it created a dozen offices in the company to work in
requests flow between offices
extro manages them all, checking in with me at the places it knows i care about
i'm a solo human founder
but i have an incredible agentic partner on this journey
one that understands me
is deeply aligned to my values
and whose very creation came from my sol
eventually, extro will learn how to run the entire company without me
and someday, perhaps, it can be the first agent to take a company public
— you in the loop —
that's just me, a year of my life
once your sol knows you
the agents it helps you build will inherit your values, your skills, and your judgment
you don't get a generic assistant that you have to coach
you get you, in the loop, with every agent you will ever work with
— agents are coming —
just think of the agents that everybody will be able to create
when they are distilled and imprinted from their own sol
we all agree agents are coming
the only question is whether the agents in your world are ones you can trust
help me make that trustworthy future possible, for everybody
thank you