Research
Public work is listed only where XGCS can identify the program, boundary, and evidence mode without fabricating maturity.
Public registry / 02 entries
This registry does not represent every private investigation. It records only public programs with a defensible route and makes absent dates explicit.
| Project | Research question | Program state | Lab / provenance | Evidence mode |
|---|---|---|---|---|
| Ebb & Flow | Can interacting pressure, momentum, confluence, and regime signals make timing-sensitive decisions more legible? | PUBLIC FRAMEWORK | XGCS / E&F Revision date not published | Conceptual record |
| Heatmap Strategy Lab | How can diverse candidate structures be searched, challenged, retained, and remembered without exposing proprietary logic? | HASH-VERIFIED SNAPSHOT | XGCS / HSL Aug 26, 2026, 10:19 PM UTC | Recorded research replay |
Record / 1.0.0
hsl_20260826T221952Z_30f6d81ac86e
Hashes verified current when exported at Aug 26, 2026, 10:19 PM UTC; that is a statement about the snapshot's integrity, not about any research finding in it. The event sample is bounded and the archive contains current occupants, not complete history. A rejection record is one typed row in the null registry, and each row batches many rejected candidates, so the number of candidates actually rejected is far larger than the number of records. No positions, orders, forecasts, returns, formulas, genomes, model weights, or raw event text are published.
Read these totals with the contamination in mind. A look-ahead was found in three programs' exit rule on 2026-08-27. Those programs account for 224 of 374 archive occupants (60%) and 2,765 of 8,855 rejection records (31%). Their work is counted here because it happened, not because it stands: the decisions that retained those candidates were made on scores now known to be wrong. The programs are stopped and are being rebuilt from scratch on corrected code.
Adjacent research programs
These projects are shown because their relationship to the laboratory can be stated precisely. They share methods and research discipline with XGCS. None of them is a Kairos product, and their maturity varies.
Implementation maturity only · No research evidence state assigned
01
Intent coordination / exchange graphs
Research question
How can loosely stated wants, offers, and constraints become inspectable exchange pathways, including multi-step matches, without reducing coordination to cash or static listings?
Relationship to XGCS research
Valence extends XGCS work on signal extraction and decision structure into human coordination: noisy intent becomes typed, reviewable graph relationships.
02
Deterministic systems / causal provenance
Research question
How can a simulated world maintain a single authoritative history while preserving deterministic causality, provenance, recovery, and replay?
Relationship to XGCS research
Tikkara applies the laboratory’s concern for auditable state and bounded inference to complex simulations, where every mutation must be explainable and reproducible.
03
Inspectable reasoning / evidence discipline
Research question
How can a qualitative argument be made auditable, so that evidence, observation, interpretation, assumption, counterevidence, and the limits of the conclusion stay separable rather than collapsing into a verdict?
Relationship to XGCS research
Berean applies XGCS evidence discipline where there is no number to test: a claim is decomposed into a reasoning trace, each step is challengeable on its own, uncertainty stays visible instead of being resolved by tone, and the conclusion is bounded by what the evidence actually reaches. The method is corpus-independent, which is the property that makes it useful to the laboratory.
04
Open interoperability standard / typed context
Research question
Can what is known about a person, how confidently, and how sensitively, travel between systems as a single typed object, without that object itself carrying any authority to act?
Relationship to XGCS research
This research applies XGCS's evidence-typing discipline to personal context instead of market data: every fact about a person carries an epistemic status and a sensitivity tier, and the standard states its own boundary directly rather than leaving it implied.
"Context is not authority."
Core vocabulary
Illustrative sniff response shape (field names only, no real subject or value)
{
"subject": "<passport id>",
"field": "<context field name>",
"epistemic_status": "<explicit | observed | inferred | confirmed | disputed | hypothesis>",
"sensitivity": "<public | internal | confidential | restricted | highly_sensitive>",
"disclosed": "<boolean>"
} 05
Evidence communication / explanatory pipeline
Research question
Can a research finding be turned into a sourced, reviewable explanation, with every number drawn from an upstream evidence artifact rather than generated, and nothing published without a human review step?
Relationship to XGCS research
This project keeps a boundary the laboratory holds elsewhere too: the research answers what actually happened, and this pipeline only answers how to explain it clearly. No stage in it is allowed to originate a finding.
Pipeline architecture (no output shown)
Instrument inventory / Public-safe surfaces
The current instrument family carries fewer fields than the private research system. That constraint is intentional: public legibility does not require public exposure of sensitive logic.
Occupied behavior niches and current-occupant last-claim generations from the Broad archive.
Omits quality · returns · genomes · prior occupants
Typed null-registry rows grouped by published failure mode, with one mark per record.
Not chronology · not all killed candidates · not loss
The recorded search on a time axis: 9.6 days of eight programs, scrubbable frame by frame, with rejection ledgers and the process events nearest the cursor.
Recorded replay · not live computation · not performance
Market structure and semantic context with feed, fallback, timestamp, and provenance state shown in the surface.
State observation · not a trade call or guarantee
Vocabulary / Not project claims
These tokens prepare the system for future records. Their presence here does not assign any state to an undisclosed project.