Most conference programmes are assembled from whoever is available and whoever paid. We think readers are entitled to know how ours is put together, including the parts that are inconvenient to admit.
The selection rule
One rule governs everything: every voice on stage must be someone we would want in the room whether or not there was an audience. In practice that filters out three common categories — the professional circuit speaker with a fixed talk, the vendor whose insight ends where their product begins, and the commentator whose track record consists of confident predictions nobody has audited.
It also means we would rather run a shorter programme than fill a slot. Ten sessions that earn their hour beat twenty that do not.
Why stage time is not for sale
No keynote slot is sold. Partner segments exist, they are clearly labelled as sponsored wherever they appear, and they never occupy an editorial slot. This costs revenue and is the single most common thing we are asked to make an exception on.
The reason is straightforward. Once a keynote can be bought, every keynote is suspected of having been bought, including the ones that were not. The credibility of the whole programme is worth more than the price of one slot.
Speakers may hold positions in assets they discuss — most serious practitioners do — and are asked to disclose material conflicts on stage. Disclosure is the standard, not abstinence.
The five tracks
Macro and monetary order. The opening block: reserves, real rates, and what has actually changed in the monetary system as against what has merely been announced.
Allocation in practice. The late-morning workshop, and the most practical hour of the day: physical against paper, storage, jurisdiction, costs, and rebalancing rules in real portfolios rather than model ones.
Wealth preservation and succession. Deliberately unglamorous. How capital is stored and transferred through a change of regime, and the administrative failures that undo otherwise sound structures.
Reserves and geopolitics. Built on primary reserve data rather than commentary. Historically the session that generates the most follow-up questions.
Supply. Where the next ounce comes from: capex cycles, declining ore grades, permitting timelines, and what producers see that the spot price does not show.
The day closes with a synthesis session and a first look at the following year's programme.
Why the line-up is published late
We announce confirmed speakers individually as contracts and compliance clearances complete, and we do not pre-announce names we cannot yet stand behind. This is worse marketing and better practice: the alternative is the familiar pattern of a headline name appearing in launch material and quietly disappearing from the final agenda.
If a session on the schedule shows a topic without a name attached, that is exactly what it means — the slot is curated and the speaker is not yet contracted.
What a session has to earn
Every proposed session is assessed against three tests before it gets an hour. Does it answer a question a serious allocator would actually ask this year, rather than a question that was interesting in 2019? Can the argument be checked — is there a named data series, a document, a filing or a published methodology behind it? And would a well-informed attendee leave able to do something differently on Monday morning?
Sessions that pass the first two tests and fail the third become written material instead. Some of the best research we commission is not stage material, and pretending otherwise wastes the scarcest resource in the programme, which is attention rather than time.
We also cap the number of panels. A panel is the correct format when practitioners genuinely disagree and the disagreement is substantive; it is the wrong format when it exists to give four people a shorter preparation burden than one person giving a proper talk.
Recordings, questions and the record
Every live session is recorded and available to General Admission ticket holders for ninety days, which is a deliberate commitment: if a speaker makes a call on stage, it stays checkable. Questions are collected in writing during each session and prioritised by how many attendees are asking the same thing, rather than by who typed fastest.
Where a speaker cites a figure we cannot source afterwards, we publish a correction note against the session rather than quietly leaving it in the replay. That policy has cost us one or two claims from an otherwise strong talk, and it is the reason the archive is worth returning to.
Practical details
The Congress runs entirely online on 10 October 2026, in a single day, with sessions timed to be reachable from Europe, the Americas and Asia-Pacific; the schedule page shows the running order in six time zones. General Admission is free and includes every live session plus ninety days of replay access. Pro and VIP places are arranged manually — submit the registration form and the team invoices you directly.
Questions about the programme, including speaking proposals, go through the contact page. We read all of them and reply to most.
How proposals are actually reviewed
Every proposal is read by two people independently before it is discussed. The reviewers see the abstract, the outline and the supporting material; they do not see the submitter's follower count, and job title is treated as context rather than qualification. Where a proposal is strong but the framing is wrong for the audience, we say so and invite a revision rather than declining outright — a meaningful share of the final programme comes from second drafts.
Three things reliably improve a proposal. A single question the session will answer, stated in one sentence. The named evidence the answer rests on. And an explicit statement of what the speaker does not claim to know, which is the fastest way to distinguish a practitioner from a promoter.
Conflicts, disclosure and product talk
We do not exclude people with commercial interests, because that would exclude most of the people who actually handle metal for a living. We do require disclosure of material positions and business relationships relevant to the session, and we require that product references be incidental rather than structural. A talk whose argument collapses if you remove the speaker's product is a sales presentation with an academic accent.
Partner content is labelled as partner content in the schedule, in the replay, and in any written summary. If a reader ever finds a sponsored segment presented as editorial on this site, that is a mistake and we would like to be told about it.
Why the day is built around a single track
Multi-track conferences look generous and behave badly online. Parallel sessions force every attendee to miss most of the programme, split the audience for the questions that matter, and remove the shared reference point that makes a discussion cumulative. Because the Congress runs in one day and entirely online, we run one editorial track: everyone hears the same macro framing in the morning, so the allocation and preservation sessions can build on it rather than re-explaining it.
The cost of that decision is breadth. Topics that would have filled a second room become written material, Academy lessons or newsletter editions instead, which is a better home for them than a half-attended parallel slot. The benefit is that the closing synthesis session can reference the whole day and assume the room was in it.
What we ask speakers to prepare
Every confirmed speaker receives the same brief, and it is deliberately narrow. One question the session answers. The named data series, filings or documents the answer rests on, sent to us in advance so we can check them. An explicit statement of what the speaker does not claim to know. And no slide that cannot be understood in ten seconds by someone watching on a phone, because a meaningful share of the audience will be.
We also ask for the falsification condition: what observation, in the next twelve months, would tell the speaker their argument was wrong. Almost every strong talk has an answer ready. The reluctance to answer that question is, in our experience, the single most reliable predictor of a session that will not survive the replay.
Corrections policy
Sessions and editions carry dated correction notes when a figure turns out to be wrong or unsourceable, appended rather than silently edited. Small typographical fixes are made in place; anything that changes a number or a conclusion is disclosed. The reason is simple: an archive nobody can trust is worse than no archive, and the only way to earn that trust is to publish the times we got something wrong.
What we will not do
No price targets presented as forecasts. No unaudited track records read out as credentials. No paid keynotes. No pre-announced names we cannot yet stand behind. No gating of the newsletter archive or the reading list. These constraints cost us reach and revenue, and they are the reason the programme is worth building at all.