Skip to content

SYS-06 · System

Research Workflow Systems — the system behind the research itself.

How the lab decides what to study, captures findings, and turns research into linked artifacts that feed the publication.

System code SYS-06Version v0.3Phase DesignOwner Solo
§ 01Opening

Research effort is lost at three points: the question is never recorded, the experiment is never completed, or the finding is never distilled into a publishable artifact. Of the three, the first is the most common and the least visible. A question arises during a build session, sits in a notes document for a week, loses its context, and disappears. The experiment that would have resolved it is never run because the question was never formalized. The publication layer never receives the finding because the finding was never produced.

The research workflow system is the operational infrastructure that closes this loop. Its job is to make questions capturable at the moment they arise, experiments traceable from hypothesis to outcome, and findings distillable into artifacts that the publication layer can reference. Without this infrastructure, research quality is a function of memory and discipline rather than system design.

At v0.3, most of the system is in design rather than operation. The experiment ledger is the only stable component. The intake surface, hypothesis tracker, run log, distillation pipeline, and artifact link registry are specified but not fully implemented. This page documents the design as it currently stands and is honest about the gap between design and current operational reality. The framework that governs how information is processed once it enters the system is FRM/05. This system is its operational implementation.

§ 02Architecture

Three layers in sequence: intake captures and qualifies incoming questions; the experiment layer records hypotheses, runs, and outcomes against active questions; the distillation layer converts completed experiments into publishable artifacts. Each layer has a ledger component — a persistent record that makes the work at that stage findable and citable by downstream layers and by the publication system.

RESEARCH FUNNELLEDGER SPINEQUESTION POOLINT-01INTAKE FILTERINT-02ACTIVE EXPT.EXP-01FINDINGDST-01ARTIFACTDST-02Question logTriage recordLedger entryDistillation draftArtifact registryNARROWS BY QUALIFICATIONPUBLICATION

Fig. SYS-06 — Research Funnel with Ledger Spine · Question pool → intake → experiment → distillation → artifact · ledger records at each stage

§ 03Question Intake

The intake surface has one design constraint above all others: low friction at entry. A question that requires any formatting, classification, or context-setting to record will be deferred. Deferred questions disappear. The intake surface must accept freeform text — a sentence, a half-formed observation, a link with a note — and hold it without requiring it to be researchable yet. Refinement happens at triage, not at capture.

INT-02 is the triage step — a periodic review of the intake queue that converts raw questions into researchable hypotheses. Triage does two things: it qualifies questions that can be operationalized as experiments and moves them to the active backlog, and it archives questions that cannot yet be operationalized with a note on the dependency that would make them researchable. An archived question is not a discarded question. It is a question that is waiting for the right precondition.

The relationship between the intake layer and the broader research progression model is documented in FRM/01. The intake surface is where the researcher's path begins — with a question that does not yet have a structure around it.

CodeNameStatusDescription
INT-01Question Intake SurfaceIn designA standing, always-available capture surface for open architecture questions. Accepts questions in freeform — refinement happens at triage, not at capture. Low friction at entry is the design constraint: a question that requires formatting to record will not be recorded.
INT-02Intake TriageIn designPeriodic review of the question intake queue. Classifies questions by domain, assigns a research priority, and moves qualified questions into the active experiment backlog. Questions that cannot be operationalized as experiments are archived with a note on why — they may become researchable as adjacent work matures.
§ 04Experiment Ledger

The experiment ledger is the canonical record of all research activity. A ledger entry exists for every active and completed experiment. The entry structure: the question being investigated, the initial hypothesis, the method used, the outcome, and a link to any artifact the experiment produced. An experiment that is not in the ledger is not findable by the publication layer. An artifact that cannot be traced back to a ledger entry has no research provenance.

EXP-01 is the only stable component in the current system because it is the simplest: a versioned document with a defined entry format. No tooling is required beyond a structured document. The hypothesis tracker and run log require more infrastructure — versioning of hypothesis state across iterations, structured run records that can be cross-referenced. Both are in design.

The most important property of the ledger is completeness over quality. A rough ledger entry that captures the question and outcome is more valuable than a polished entry that was never written because the experiment did not seem significant enough to document. The ledger's value is cumulative — it is the foundation of the artifact link registry and the primary evidence surface for the publication layer's research pages.

CodeNameStatusDescription
EXP-01Experiment LedgerStableVersioned record of all active and completed experiments. Each entry: the question being investigated, the hypothesis, the method, the outcome, and a link to any published artifact the experiment produced. The ledger is the canonical reference — research not in the ledger is not findable by the publication layer.
EXP-02Hypothesis TrackerIn designExplicit versioning of hypothesis state across iterations of a single experiment. A hypothesis that was wrong at iteration one and corrected at iteration two should both be preserved — the correction is part of the finding. Currently handled informally in ledger notes. Formal versioning is in design.
EXP-03Experiment Run LogIn designStructured record of individual experiment runs: inputs, parameters, raw outputs, and observations. Feeds the ledger at outcome time. Currently unstructured — raw notes in associated documents. The structure required to make runs cross-referenceable has not been specified.
§ 05Finding Distillation

Distillation is the stage where a completed experiment becomes a finding that the publication system can use. The raw ledger entry and run log contain the evidence. Distillation extracts the transferable insight — the claim that holds beyond the specific experiment conditions, that connects to existing frameworks, and that meets the publication's standard for load-bearing content.

The distillation bottleneck is the system's most significant current gap. Experiments complete faster than findings are distilled. The gap accumulates as a backlog of completed experiments that have not yet produced published artifacts. Each item in that backlog represents research effort that has not yet contributed to the publication's compounding value. The research pages across this publication represent the distilled findings that cleared the backlog. The findings still in it are invisible until DST-01 closes the gap.

DST-02 tracks the relationship between ledger entries and published artifacts. A research page may draw on multiple experiments. A single experiment may contribute to multiple pages across different research domains. The registry makes this dependency explicit — which is a requirement for the publication's load-bearing link architecture. A link from a framework page to a research page is load-bearing only if the research page's claims can be traced to experiments that produced them. The artifact link registry is what makes that traceability possible. The second brain at SYS-01 receives distilled findings as processed knowledge — the handoff between DST-01 and SYS-01 is currently manual and unspecified.

CodeNameStatusDescription
DST-01Finding DistillationIn designThe stage between a completed experiment and a publishable artifact. Takes the ledger entry and run log, extracts the transferable insight, and produces a draft that meets the publication standard. The distillation step is where most findings are currently lost — the experiment completes but the finding never reaches the format required for a research page.
DST-02Artifact Link RegistryIn designTracks the relationship between ledger entries and published artifacts. A single experiment may produce multiple artifacts across research pages, framework updates, and build log entries. The registry ensures the dependency is explicit — a published page can be traced back to the experiment that produced it, and an experiment can be traced forward to every artifact it contributed to.
§ 06Operational Notes

Current operational state

EXP-01 (the experiment ledger) is the only stable component. Question intake is informal — a running notes document without structure. Distillation is ad-hoc — findings are converted to artifacts when there is time, not on a defined cadence. The gap between what the system is designed to do and what it does today is wide. v0.3 reflects design clarity, not operational completeness.

The intake problem

Questions captured informally disappear within days if they are not operationalized. The intake surface needs to be frictionless enough that a question is recorded at the moment it arises — not later, when the context that made it a question has evaporated. The current informal surface does not meet this bar. INT-01 design is the most pressing gap.

Ledger coverage

EXP-01 covers experiments that were deliberate enough to merit a ledger entry. Informal explorations — the research equivalent of spike work — are not recorded. This creates a gap in the artifact link registry: some published findings cannot be traced back to a ledger entry because the experiment that produced them predates the ledger or was never recorded. Retroactive reconstruction is impractical.

Distillation bottleneck

The finding distillation step is the current rate-limiting stage. Experiments complete faster than findings are distilled. The backlog of completed experiments with no corresponding published artifact is the most direct measure of where the system is losing output. DST-01 automation is intended to lower the distillation cost — the design has not been finalized.

Relationship to SYS-01

Distilled findings feed directly into the AI Second Brain as processed knowledge. The handoff between DST-01 and SYS-01 is currently manual — a distilled finding is added to the second brain's knowledge store by hand. The interface between the two systems has not been specified.