AR-STA-001·PL-01 — Facility state board. Stylised brand illustration of the laboratory, not a photograph of a physical facility. The charts and figures inside the plate are simulated.

Record AR-FAC-009 · Facility status board

What is built, what is not,
and what is broken.

A preview that describes twenty-five modules and ships six of them owes the reader a precise account of which is which. This page states the facility's operational state, the functionality level of every planned module, a dated log of what was constructed through 2026, and the defects that are currently known and unfixed.

Facility state Preview Operational Modules 25 planned Log entries — Read at —

01Current facility stateAR-FAC-009·01

Preview Operational.

Preview Operational is the facility's declared state and it means something specific. The public research environment is complete and functioning: every documented instrument on this site runs, every control responds, every figure computes, and the whole thing works with the network disconnected. That is the part that is finished.

What is not finished is the application. The Foundry as described in the roadmap is a twenty-five-module environment that ingests your trade history, evaluates your frameworks against it, and governs deployment on the result. None of that data path exists yet. There is no upload, no account, no persistence, and no evaluation of any real record. The preview demonstrates the analytical layer by modelling it; it does not yet apply it to anything of yours.

The distinction matters because the two are easy to conflate when a preview is polished. A working interface is not evidence of a working pipeline behind it. On this site there is deliberately no pipeline behind it, and the module table below is the honest accounting of where each piece stands.

Counts derived from the planned module register · five stages × five modules · functionality levels are declared in the data layer, not estimated on this page.

Facility state definitions

Preview OperationalCurrent. The public research environment runs in full; the data application does not exist. All outputs are model results from stated assumptions.
AlphaNot reached. Would require a working upload, schema mapping and validation path with a small invited group operating on their own records.
BetaNot reached. Would require persistence, saved laboratories, and the diagnose stage running against uploaded data rather than against the model.
ProductionNot reached. Would require the governance stage, exports, and an operational commitment this project does not currently make.

No uptime figures are published

This is a static site with no backend, no session state and no incident surface. Publishing an availability percentage or an incident feed would be theatre: there is no service to be unavailable beyond the file host, and no monitoring behind the number. What can honestly be reported is what is built, and that is what this page reports.

02Module registerAR-APP-STAGES

Twenty-five modules,
three honest statuses.

Every module in the planned application carries a declared functionality level in the site's data layer. This table reads those levels directly and converts each into a plain status, so the register and the roadmap cannot drift apart. Nothing here is a projection of a delivery date, because none exists.

L1 Operational

A working demonstration of this module exists on the public site today. It models the relationships rather than reading your data, and it is labelled as such wherever it appears.

L2 Preview

Partially represented. The concepts are documented and some behaviour is demonstrated, but the interactive module described in the roadmap is planned rather than present.

L3 Planned

Specified and described only. No implementation exists in any form. Everything written about these modules on this site is in future tense, and should be read that way.

Every planned application module, its stage, declared functionality level, plain status and where it can be seen today
StageModuleLevelStatusWhere it is demonstrated
Figure 1 · Module count by operational status Derived

Counted from the module register.

Reading the distribution

The shape of that chart is the honest summary of the project. Roughly a quarter of the planned application is demonstrable today, and it is the analytical quarter — Monte Carlo, execution diagnostics, regime analysis, friction modelling and the two governance instruments. The construction stage and the persistence layer, which are what turn a research environment into a product, are almost entirely unbuilt.

That ordering is deliberate. The analytical layer is the part with the intellectual risk: if the relationships it encodes are wrong, no amount of build quality elsewhere rescues it. It is cheaper to be wrong about it now, in public, than after a storage layer has been written around it.

Roadmap detail Planned tiers

03Build log · 2026AR-FAC-009·03

What was built,
in the order it was built.

A changelog rather than an announcement feed. Entries record construction, correction and removal alike; two of the items below are things that were taken out, and one is a claim that was withdrawn after it failed its own validation conditions.

Log conventions

BuiltNew capability that a reader can use. Dated to the point it became publicly reachable, not to when work started.
RevisedExisting behaviour changed materially enough that a previously published reading would now differ.
CorrectedA defect fixed. Corrections are logged with what was wrong, not merely that something changed.
WithdrawnSomething removed — a claim that overreached, a figure whose method did not hold, a feature that could not be made honest.

Dates are the calendar dates of publication. No entry is backdated; where a correction followed a defect, both appear.

04Known issues · limitationsAR-FAC-009·04

Seven defects and limitations,
currently unresolved.

These are not hypothetical risks; they are the things that are wrong or absent right now. Some are deliberate consequences of the preview's design and will remain until the application phase. Others are simply not done yet.

This is the largest limitation on the site and it is structural rather than accidental. The engine reads no price data. It encodes directional relationships that are argued for in the research records — that friction scales against the working stop, that partials trade tail participation for capture efficiency, that correlation converts concurrency into concentration — and it produces internally consistent outputs from them.

What that means for a reader: every number on this site is an illustration of how components interact, not a measurement of what any market did. A configuration that grades well here has demonstrated nothing about its behaviour in live conditions. The engine is a model of a system, not a model of a market.

Status: permanent for the preview phase. Historical evaluation requires the data pipeline, which is unbuilt.

The Wizard framework library currently contains twelve abstracted decision frameworks against a target of eighty to one hundred. The twelve that exist are deliberately spread across style families — trend, breakout, mean reversion, discretionary macro, systematic short-term — so the comparison dimensions are exercised, but the sample is far too small to support the cross-framework analysis the library is intended for.

Anyone using the archive to reason about how doctrines cluster should treat the current contents as a demonstration of the record format rather than as a corpus.

Status: in progress. Each record requires source reading, abstraction and a limitations note, which is slow by design.

The learning centre lists fourteen articles with categories, levels and reading times. The individual article pages behind those entries do not exist yet; the listing is a table of contents for material that is written into the instrument pages rather than into standalone essays.

The practical consequence is that a reader following a learning path is redirected into the relevant laboratory page, which covers the same ground at greater density and less pedagogically. That is a worse experience than the listing implies, and it is the most commonly noticed gap on the site.

Status: queued behind the module work. The listing will not be removed, because the material it points at does exist — just not in article form.

The data laboratory documents a ten-step intake path — upload, inspect, detect schema, map, validate, normalise, preview, resolve, save mapping, prepare — and demonstrates the states a row can be rejected into. None of it accepts a file. There is no upload control anywhere on this site and no storage behind one.

Equally, nothing you do here is remembered. Parameter states, Monte Carlo configurations and comparisons are lost on reload, because the only thing written to your browser's storage is the motion preference. Saved laboratories are a planned capability.

Status: deliberate for the preview. Accepting trade data implies a server, an account and a duty of care that this phase does not take on.

The animated scenes — the drawdown terrain, the regime tunnel, the branch network, the correlation field — are canvas renderings marked aria-hidden. They convey no information that is not also stated in the prose beside them, which is why they are hidden rather than described: a partial description would imply there was something more to miss.

This is a considered position rather than an unqualified success. A sighted reader gets an intuition from watching a correlation field collapse under stress that a text summary does not fully replace, and closing that gap properly would require rebuilding those scenes as accessible figures.

Status: mitigated, not solved. Detailed in the accessibility statement.

Every chart carries an accessible title and a plain-language summary, so a screen-reader user learns what the figure shows and what conclusion it supports. What no chart currently offers is point-by-point navigation of its underlying series, or an equivalent data table.

For a two-series comparison the summary is adequate. For the correlation matrix, the path field and the regime timeline it is a genuine loss of detail — those figures reward inspection, and the summary compresses that into a sentence.

Status: known gap. Data-table equivalents for the dense figures are the intended fix.

The site is built to WCAG 2.2 AA as an aim and has been checked informally — keyboard-only traversal, a screen reader, reduced-motion behaviour, 360 px viewport, browser zoom to 200 per cent. It has not been audited by anyone qualified to certify the result, and the statement on the accessibility page says so rather than implying otherwise.

There is also no high-contrast or light theme. The design is a single dark palette. Readers who need higher contrast are currently dependent on operating-system-level inversion, which this site does not specifically support or test against.

Status: open. Reporting a specific failure is the fastest route to it being fixed.

Issues that are not listed here

This list covers defects and limitations that are known. It cannot cover the ones that are not — and on a site of this density, unknown defects certainly exist. If you find a number that contradicts another number, a control that does nothing, or a claim that is not supported by the method beneath it, reporting it is genuinely useful.

Report a problem →

05Under constructionAR-FAC-009·05

What is being worked on,
without a delivery date.

Ordered by what is actually in progress, not by what would be most impressive to announce. No dates are attached, because a date attached to a single-operator project is a guess dressed as a commitment.

In progress now

Wizard archive expansionContinuous. Each record needs source reading, framework abstraction and an explicit statement of what the abstraction discards. Target is a corpus large enough to compare doctrines across regimes rather than one large enough to browse.
Learning-centre article pagesFourteen listed, none published. Written as standalone explanations rather than as extracts from the instrument pages, so they can be read without the laboratory context.
Data-table equivalents for dense figuresAccessible tabular alternatives for the correlation matrix, path field and regime timeline, addressing known issue 06.
Research record detail pagesThe archive lists studies with abstracts and findings; the full method, configuration and validation outcome for each belongs on its own record page.

Next, in order

Client-side CSV parsingThe first step of the intake path, run entirely in the browser with no upload. A broker export read locally and mapped against the analytical schema — no server, no storage, no account.
Local persistence for laboratory stateSaving a parameter configuration and a Monte Carlo setup to your own browser so a session survives a reload. Local only, exportable, deletable, and explicitly not synchronised anywhere.
Configuration comparisonTwo or more saved configurations evaluated side by side with the differences and their downstream consequences made explicit rather than left to be eyeballed.
Framework record expansionThe twelve framework records carry structure and regime fit; they should also carry a worked calibration walk-through and the failure modes specific to each family.

Explicitly not being worked on

Live market data, broker connectivity, order routing, automated execution, copy trading, a mobile application, and any form of signal distribution. None of those are on the roadmap at any stage. If the Foundry ever ships a product it will be an analytical one, and the boundary between analysis and execution is a design commitment rather than a sequencing decision.

06Methodology · how this page worksAR-FAC-009·06

How a status is decided.

Each module's functionality level is declared once, in the site's data layer, alongside the stage it belongs to. Every page that mentions that module — the roadmap, the pricing tiers, this status board — reads the same declaration. There is no separate status database that could disagree with the roadmap, and no manual list on this page to fall out of date.

The three levels map to statuses by a fixed rule with no discretion in it:

  • Level 1 → Operational. A working demonstration is reachable on this site today, and a reader can exercise it.
  • Level 2 → Preview. The concept is documented and some behaviour is shown, but the module described in the roadmap is not present.
  • Level 3 → Planned. Specification only. No implementation of any kind.

A module is only promoted when the demonstration exists and is honest — meaning it carries its disclosure, its assumptions are stated, and it does not imply a capability behind it that is absent. Promotion is logged in the build log with the date it became reachable.

What this page does not do

No uptime percentageThere is no monitored service. A published availability figure with nothing measuring it would be a fabrication.
No incident feedStatic files served from a host. Outages, when they happen, are the host's and are not visible to this page.
No live telemetryNothing on this site reports back. The reading timestamp above is generated in your browser from your own clock.
No delivery datesRoadmap ordering is published; schedules are not. An unmet date erodes more trust than an absent one.
No silent editsMaterial changes to published claims appear in the build log as revisions or withdrawals rather than being quietly overwritten.

Cross-checks a reader can run

Open your browser's network panel and confirm no third-party request is made. Disconnect the network and confirm the instruments still compute. Reload a figure and confirm the numbers are identical. Each of those is a claim made elsewhere on this site, and each is checkable in under a minute.

AR-FAC-009 · END OF RECORD

Six modules work today.
Start with those.

The operational modules are the analytical ones, and they are the reason to be here at this stage. Move a parameter and watch eight outputs disagree. Run a thousand resequenced paths and watch the median stop being a forecast. Turn the governance layer off and watch the drawdown envelope open.

Open the parameter laboratory Monte Carlo core About the Foundry Report a problem

Status read at — · generated in your browser · no telemetry is collected

Reconciliation · AR-FAC-002

Where a number has more than one true value.

A demonstration ceiling and a production ceiling are both real, and so is a preset that differs from a default. What is not acceptable is two figures on two pages with nothing saying which is which. Each row names the canonical value, then names every other value that is also correct and says where it applies.