01
the roster
eight rings, eight domains. hover for the one-line version, open one to read the thesis. every field on a record is real or an em dash.
livebuilding researchconcept open, take it
02
seven hours, & none of them spent drafting
count the hours you actually decided something last week. not the hours you worked, the hours where the outcome depended on your judgement rather than your typing. it is a small number. the rest was carriage.
so the agents take the carriage. what is left is the part that was the job all along: deciding what is worth building, & saying when it is right. the compression is real. the signature is still yours, & so is the liability.
03
the handset is the last chapter, not the first
the end state is a device where the agents are the surface & there is no app grid at all. Set0 in the hand, SetX on the desk. now the honest thing, immediately: hardware has killed everyone who tried this recently. the phone already in your pocket is faster & already logged into everything you own. so the agent layer runs on phones people already have, & earns a device only if it ever becomes the thing people open.
04
the written work
each one separates what it proves from what it merely constructs. one ends with a ledger of every mistake made while writing it, kept because a record showing only successes is not a record.
The Closure Condition
One theorem, followed from a drifting table of eleven to rāga, room, and the comma that never comes home.
A cycle closes if and only if the ratio of its two rates is rational. Section 9 is a ledger of every error made writing it.
Adaptive Cryptographic Intelligence
An AI-driven adaptive security layer around existing cryptographic standards.
Six hypotheses, none tested, all stated as hypotheses. No security proof, and the paper says it does not expect one.
Kyklos
n² − 7n + 2, the circle, and the digit seven.
Take an exact structure, perturb one axis, watch what breaks. The breaking is never noise.
Does a mirror force the centre?
Five plates. The last plate is the Riemann Hypothesis.
Symmetry alone proves nothing, which is why 160 years of the functional equation has not settled it.
An Imperfect Harmony
On choirs that sink, pianos that lie, and the small wrongness that lets music move.
We have never had perfect harmony. Only a long argument about which imperfection to accept.
05
good artists copy.
- usually put on picasso, though the trail is thin.
great artists steal.
every concept here is written out in full, the thesis, what it does, & where we think it breaks. if you can build it better or faster than us, it is yours. no attribution, no permission, nothing to sign.
open any agent to declare a steal, or to say you want to build it together. both are one field & an email.
06
open bench
agents that streamline real work across a domain boundary, the places where an expert has to become a second-rate version of somebody else to get through the day.
- An agent that makes a domain expert faster without pretending to be one
- The multi-agent orchestration you keep hand-rolling and wish existed
- A workflow where the handoff between agents is the whole product
- Something that should exist but nobody has the standing to build alone
dump an idea
the agent you keep wishing existed. one field & an email.
07
four doors
pick the one that matches why you are here. all four end up in the same inbox & all four get a reply from a person.
see who is building
the board, ideas, comments, & the votes they drew.
open the board → 02take an idea
open any agent & declare the steal. no permission needed.
go to the roster → 03put money behind it
request the deck. we send it & leave you alone until you reply.
request the deck → 04just be around it
join square8. people building agents & arguing about them.
join →interested?
investing, or want the deck. we’ll send it & leave you alone until you reply.
join square8
the community. people building agents, arguing about them, & stealing from each other in the open.