Humble Superintelligence · semantic infrastructure
We build the canonical layer of the web. One address per concept — structured, cited, and signed — so a model can resolve meaning instead of guessing at it.
618 concept domains held outright across .si / .ai / .bot / .com / .io / .org; 101 working betas already shipping as demos (register live-audit pending). Faith and science clusters, and civic flagships including democrat.ai and republican.ai, are strategic build assets we hold and develop — not for-sale inventory.
The bottleneck is stopping being compute and starting to be grounding.
01 — Why grounding
The fix is not a better crawler or a bigger model. It is infrastructure: a layer of the web where each concept has exactly one canonical address, publishes structured meaning, cites its primary sources, and signs what it says — so that both a machine and a person can check it.
01
Models trained on model output degrade. Model collapse has a name and a mechanism: each generation of synthetic text narrows the tails until the rare, specific, and true gets averaged away.
02
A claim travels detached from where it came from. A model cannot weight what it cannot trace, and a person cannot check what the model will not name.
03
When an agent needs the canonical account of a concept, there is often no canonical account — only a search page, an aggregator, and a content farm that outranked both.
04
One wrong resolution inside a chain of tool calls compounds. The cheapest place to prevent that failure is at the address — before retrieval, not after.
The concept's identity and its network address are the same string. No disambiguation step, no ranking step, no guess.
Meaning is published as data — typed entities, defined terms, dataset descriptors, sameAs links into the wider graph.
Every document carries a content hash and a signed modification time. Provenance is the scarce commodity on a generated web.
The property answers queries rather than being scraped for them — structured responses with source attribution attached.
02 — Live network
No mockups. Pick any property — your browser fetches the real site, live. Each sits at the exact lexical address of the concept it serves.
Estate size: 618 names (canonical register). Developed flags in the rail come from the register (97 flagged) — flagged developed in register; live audit pending. Treat the rail as a sample of the network, not a certified developed total. Best-tier instruments (physics, civic twins, faith consoles, WebGL science) sort first. HOLD badges mark strategic build assets — not for sale. Primary actions are always visible: Load live preview, Open, ← →.
Embed unavailable
This site refused to be framed (X-Frame-Options / CSP). That is normal for many live properties — open it directly instead.
Ready
Pick a name in the rail, then Load preview — or Open in a new tab. We never fire 618 iframes.
03 — Architecture
Five layers, one templated stack — 618 concept endpoints, one pattern. Every property is meant to be addressable, structured, attributable, and callable.
Dictionary-word domain as the identity of the concept.
schema.org / DefinedTerm and linked open data.
Hashes, signed timestamps, citation history.
JSON APIs and MCP endpoints agents can call.
Human-readable pages that stay honest as demos / beta.
A dictionary layer is a small fraction of the tokens and a large fraction of the grounding.
Go to the discipline's address and get a structured answer with sources — instead of ranking a search page you cannot explain.
If an answer is wrong, the chain leads back to a named, dated, signed document rather than into fog.
Someone has to hold the canonical account of the sciences, the professions, the faiths and the cohorts.
04 — Estate
Concentrated deliberately in the dictionary layer of .si and adjacent namespaces — one address per concept, assembled so the window to build it coherently does not close in pieces.
A growing share of the estate already carries content, structured meaning, and citation history that travel with the address. Developed properties are demos and beta of the stack — proof that the concept address works — not a sales catalog.
Faith / science clusters and civic flagships (democrat.ai, republican.ai) stay on HOLD. Note: gods.si is the live faith instrument; gods.ai is held strategically.
Badge language on the register: HOLD — strategic build asset. Not listed for sale.
Includes faith & science clusters and names such as democrat.ai, republican.ai, christianity.ai, hinduism.ai, gods.si (live), rillhurst.com, polygamous.com, and physics.si.
490 .si · 51 .ai · 37 .bot · 26 .com · 11 .io · 3 .org
05 — Register
Names load from the asset register. Blue means the property is flagged developed and operating. Exact live totals remain under reconciliation — browse by concept, not by a headline developed count.
Canonical total 618. Developed badges reflect register flags; live audit pending. HOLD names are strategic build assets, not inventory.
06 — Build with us
One concept, one canonical address — structured and signed so agents can resolve rather than scrape.
Talk to Jake — white paper and reconciled register on request.
Somewhere for machines to stand.
One address per concept.