Record AR-INQ-001 · Institutional contact

Specific disagreement is
the most useful message.

This facility improves fastest when someone points at a relationship the engine has modelled the wrong way round, a claim that exceeds its evidence, or a control that does not do what its label says. The form below is structured to make that kind of message easy to write and easy to act on.

Status Preview Form Demonstration only Transmission None Updated 2026-07

01Before you writeAR-INQ-001·01

Eight enquiry types,
and what each one is for.

The routing exists because the eight categories need different things from you. A defect report needs a page and a reproduction; a collaboration proposal needs a dataset and a method; a question about the application needs to say which module. Choosing the right one is most of the work of getting a useful answer.

General enquiryQuestions about what the Foundry is, what it does, how a concept on the site fits together, or whether a particular problem is in scope. If your question is answered on the frequently asked questions record, start there — it is faster and more complete than a reply would be.
The applicationQuestions about the laboratory application itself: what a module does, what is built today against what is specified, where something sits on the roadmap, whether a strategy family or a broker export is in scope, or what a capability on the pricing matrix actually means. The build status record and the roadmap answer most of these faster than a reply would, and are updated more often.
Plans and pricingQuestions about the intended launch tiers, monthly against annual billing, what separates Foundry Pro from Foundry Quant, the founding-researcher terms, or institutional and research access. Note that nothing is on sale yet and no card is being collected, so these are questions about intent rather than about an account.
Research collaborationMethodological objections, replication proposals, datasets you are willing to have analysed under stated conditions, or a study you want run against the validation protocol. Include the method and the sample size; a collaboration proposal without either is difficult to evaluate.
MARS workbookQuestions about the Montex AlphaRail System workbook itself — the spreadsheet governance product the analytical layer of this site grew out of. Note that gate boundaries, pool percentages and tier tables are part of that product and are not published here, so questions asking for those numbers will be answered by saying so.
Press and citationRequests to quote, cite or reference material from this site. Educational quotation with attribution needs no permission; the terms of use set out what that covers. Use this route for interviews, longer reproductions or anything where you need a named source.
Accessibility issueAny part of the site that is unusable with your assistive technology, keyboard, magnification or contrast requirements. Treated with the same severity as a wrong number. State the technology, the page and what happened — see the accessibility section below for what makes a report actionable.
Report a problemDefects: a figure without a disclosure marker, a claim without a stated assumption, a control that does nothing, a number that contradicts another number, a broken link, future capability written in the present tense. Include the page and, where you can, what you did before it went wrong.

A note on what makes a report useful

“The Monte Carlo page is wrong” cannot be acted on. “On the Monte Carlo page the completion probability rises when I increase risk per trade, which contradicts the drawdown reading immediately beside it” can be, and would be investigated the same day it arrived. Specificity is not a courtesy here; it is the entire difference between a message that changes the build and one that does not.

02Enquiry recordAR-INQ

Enquiry form.

This is a demonstration of the intended enquiry record. It validates your entry in your browser and returns a reference, and then it stops — because there is nothing behind it.

This form transmits nothing

There is no server, no endpoint, no email relay and no database attached to it. Nothing you type leaves your browser, nothing is stored between page loads, and the reference number it returns is generated locally and recorded nowhere. It exists to demonstrate the intended intake structure and its validation behaviour. Do not enter anything you would not be content to type into a text editor and close without saving.

Validation runs locally · no request is made

Demonstration intake form — validation only, nothing is transmitted or stored

What the form checks

PresenceName, email, message and the consent confirmation must all be supplied. Empty submission highlights every missing field at once rather than one at a time.
ShapeThe email field is checked against a basic structural pattern. It is not verified, because verification would require a network request this site does not make.
SubstanceThe message requires at least twenty characters. A defect report shorter than that is almost never actionable.
AnnouncementFailed fields are marked invalid for assistive technology, the first is focused, and each error is written into its own text slot beneath the field rather than conveyed by colour alone.

The same validation module runs the early-access register. It is the intended intake behaviour of the application, demonstrated in the only place a static site can demonstrate it: on itself.

If you need to reach a person

A working channel will be published when there is capacity to answer it properly. Publishing an address that is not read is worse than publishing none, so this page does not do it. Watch the facility status board — the channel opening will appear in the build log.

Facility status board

03Scope of correspondenceAR-INQ-001·03

What will be answered,
and what will not.

Stated in advance so nobody spends effort writing a message that was never going to receive the answer they wanted. The second column is not evasion; several of those requests would be unlawful, unwise or dishonest to fulfil.

Will be answered

Defects and contradictionsAnything on this site that is wrong, unsupported, undisclosed or broken. These are treated as build items and their resolution appears in the build log.
Methodological objectionsIf you think a relationship the engine encodes is the wrong way round, or that a conclusion outruns its sample, say so. Objections that hold change the model.
Questions about what a concept meansDefinitional questions about anything documented here — capture efficiency, effective independent bets, gate dwell, variant separation.
Accessibility failuresEvery report, with priority. A site that reasons publicly about honesty cannot be quietly unusable for some of its readers.
Scope questions about the MARS workbookWhat it is, what it governs, how the structure relates to what is published here, and what remains proprietary.

Will not be answered

“What should I trade?”No trade ideas, no directional views, no instrument recommendations, no entries. Not a matter of policy so much as of what this facility contains, which is none of those things.
Requests for personal adviceQuestions of the form “given my account size and situation, should I…” cannot be answered. That is regulated advice, this is not a regulated entity, and a plausible-sounding answer would be worse than silence.
Requests to review your systemIndividual strategy reviews are not offered. The instruments on this site exist so you can perform that examination yourself, which is also the only version of it that is worth anything.
Requests for withheld calibrationGate boundaries, pool percentages, tier tables and workbook Monte Carlo figures belong to the MARS product. The answer is that they are not published, and it will not change with repetition.
Partnership, affiliate and promotional offersNo affiliate arrangements, no sponsored placements, no broker introductions and no paid mentions. There are none on this site and none are planned.

04Research collaborationAR-INQ-001·04

What a collaboration
looks like here.

The Foundry's research output is a set of records, each carrying a proposal, a method, a configuration, a finding and a status that can be failed. A collaboration means adding to that record set under the same conditions — including the condition that a negative result gets published as readily as a positive one.

Three kinds of proposal are genuinely useful. Replication: you have run an analysis whose method can be restated precisely, and you want it re-run under the validation protocol used here, with the disagreements documented. Data: you hold a trade record of adequate size, with risk amounts and costs intact, and you are willing for structural analysis of it to be published in aggregate and anonymised form. Objection: you believe a modelled relationship is wrong, and you can say what evidence would settle it.

What a collaboration is not: a route to a private consultation, a co-marketing arrangement, or an endorsement. Joint work is credited by name where the collaborator wants to be named, its limitations are published alongside it, and the party proposing it does not get editorial control over the finding.

On sending trade data

Do not send trade records through this form — it transmits nothing, so they would be lost, and even if it did work, an unsolicited dataset arriving without an agreed handling basis is a liability rather than a contribution. Describe the dataset instead: instrument coverage, date range, trade count, whether risk amounts and costs are present, and what you would want done with it. The handling basis is agreed before anything is sent.

What a proposal should contain

The claimStated precisely enough to be falsified. “This works” is not a claim; “capture efficiency degrades as trigger timeframe falls, holding friction fixed” is.
The methodWhat was measured, how, over what period, and with which exclusions. Exclusions matter more than inclusions.
The sampleTrade count, instrument coverage and date range. Sample adequacy is one of the eight validation conditions and it is the one most proposals fail.
The disconfirming evidenceWhat result would have changed your mind, and whether you looked for it. A proposal with no stated disconfirmation is a conclusion looking for support.
The publication termsWhether you want to be named, and what you need withheld. Both are accommodated; what is not accommodated is suppressing an unfavourable finding.
Read the research archive first

05Reporting an accessibility problemAR-A11Y-REP

If this site does not work
with your technology, that is a defect.

Not a preference, not an edge case, and not something to be worked around by the reader. The accessibility statement sets out what was implemented and what was informally checked; it also states plainly that no formal audit has been carried out, which means undiscovered failures are likely rather than possible.

01

What to include

The page, the assistive technology and version, the browser and operating system, what you were trying to do, and what happened instead. If a control was reachable but not operable, say which key you pressed.

02

What counts

Anything unreachable by keyboard, any focus that disappears, any status conveyed only by colour, any control below a comfortable touch size, any text that reflows badly at 200 per cent zoom, any motion that persists after you have asked for it to stop.

03

What happens next

Accessibility defects are prioritised above feature work. The fix is recorded in the build log, and where a limitation cannot be fixed quickly it is added to the known-issues list on the status board rather than left undocumented.

Known before you report it

Two limitations are already documented and do not need reporting unless you want to argue about the approach: the animated canvas scenes are marked decorative and cannot be read by assistive technology, with everything they depict stated in the adjacent prose; and the dense figures carry a text summary rather than point-by-point navigation of their data. Both are set out in full, with reasoning, in the accessibility statement.

Read the accessibility statement

06Response expectationsAR-INQ-001·06

No promised timescale.

This page does not commit to a response time, and the reason is simply that it could not be honoured. The Foundry is not a staffed support function. It has no ticket queue, no service agreement and no obligation of the kind those imply, and inventing a “within 48 hours” line would be exactly the sort of frictionless promise the rest of this site argues against.

What can be said honestly is how correspondence is prioritised when it is read. Defects that affect the accuracy of a published claim come first, because a wrong number on a site about honesty is a category error. Accessibility failures come next. Substantive methodological objections after that, because they take real work to evaluate properly and a hurried reply is worse than a considered one. General questions are answered when they are not already answered somewhere on the site.

The most reliable signal that a message was acted on is not a reply. It is the build log: a correction that arrived from a reader appears there as a dated entry describing what was wrong.

Faster than writing

Frequently asked questionsThe eight questions that arrive most often, answered at length rather than in a sentence. Open the FAQ
Facility statusWhether the thing you are asking about exists yet, what its functionality level is, and whether it is on the known-issues list. Open the status board
Risk disclosureWhat the engine models, what it does not, and what no model addresses. Open the disclosure
Learning centreDefinitional questions about expectancy, drawdown, sequence risk, correlation and sample adequacy. Open the learning centre
Terms of useCitation, reproduction and permitted use, which covers most press enquiries without correspondence. Open the terms
AR-INQ-001 · END OF RECORD

Or say nothing,
and go and break something.

The instruments are the argument. Move the friction control until the edge disappears, then work out at which timeframe it disappeared and why. If the model behaves in a way you can show is wrong, that is a message worth sending — and you will have the specifics in hand when you send it.

Demonstration form · nothing transmitted · no cookies · no tracking