MOSAIC: What You Get, and How to Check It
What this is
MOSAIC is named for the method. In professional research it is called mosaic theory: no single piece of public information tells you much, but thousands of small fragments, assembled with discipline, form a picture visible in none of them alone. A capacity commitment here. A supply agreement there. A lead-time remark on an earnings call. A customer named in a grant announcement. Individually trivial. Together, a regime change taking shape.
Most readers meet these fragments one at a time, in passing, and lose them. MOSAIC catches the fragments continuously, holds them in one frame, and states what picture they are forming, early, as forward calls that are graded in public. Every fragment is public. The edge is the assembly.
Where the picture is sharpest right now
The clearest picture in the current evidence is constraint migration underneath the technology buildout. When demand shocks hit physical supply chains, the binding constraint moves. It moved from GPUs to memory capacity. From memory to advanced packaging. From packaging to optical interconnect, and beneath that to the substrate materials optics are built on. From silicon to grid power, and from power to cooling. Each migration has re-rated the companies sitting at the new choke point, and each time, most coverage was still watching the last one.
Today the coverage universe spans compute, power, optics, advanced materials, and biology. That is the current map, not the borders. The method goes wherever the fragments accumulate; when the picture moves, so do we.
What you get
A running answer to one question: what picture is the evidence forming, and which companies sit inside it. Today that question takes a specific form: where is the next binding constraint forming, and who sits on it.
Concretely:
Forward calls, made in public, with deadlines. Each names the constraint, the companies positioned on it, and the exact conditions under which the call will be graded pass or fail.
Live conviction on every open call. Our confidence moves as evidence lands, and every move is published the same day with the evidence named.
A weekly synthesis. What resolved, what moved, what catalysts are approaching, and one clearly labeled directional read on named companies.
A public record proving whether any of this works.
You could assemble this view yourself by reading every earnings call, capacity announcement, supply agreement, and trade filing across five industries, daily. The subscription is for people who want the output of that reading without doing the reading.
How to check it
Most research asks you to trust the author's memory. Calls are made, misses are forgotten, hits are retold. This publication is built so that neither you nor we get that option. Every call hangs off the Ledger, a public scorecard where calls are registered before the fact and graded after it by rules written in advance.
A Registration is published before the event it concerns. It states the claim, the deadline, and the checkable assertions inside the claim, called legs. One leg is marked core: the assertion the call lives or dies on. A verdict rule, written at registration, maps every combination of leg outcomes to exactly one verdict. Numeric legs get their baselines snapshotted and frozen before the event, with the snapshot date published. Once posted, a Registration is never edited. Genuine errors get a separate Correction post, plainly worded. At grading time there are zero degrees of freedom left: the goalposts are public and dated.
Every live call carries a tier: HIGH, MEDIUM, or LOW. The call is frozen; the tier is live conviction, and it moves with evidence. Read tier changes as information, not indecision. A call demoted on named evidence tells you more than one held out of stubbornness, and every call's full tier history is visible on the Ledger with the reason attached at each step.
At the deadline, each leg is scored against the public record, pass or fail, evidence cited. The registered rule converts the leg results into the verdict: HIT, PARTIAL, or MISS. No judgment is applied at grading time. PARTIAL is not a hedge; its meaning is fixed per call, written into the verdict rule at registration. Read any call's rule and you know exactly which leg outcomes produce PARTIAL before the event happens; at grading time it means precisely what was pre-committed and nothing else. The Ledger shows two scores: the call-level record, which is the headline, and a leg-level number for granularity, with a standing caveat that legs inside one call are correlated.
Misses
Misses get more words than hits. Every miss publishes with a full accounting: what we believed, which leg failed, and what changed in our thinking as a result. Hits get the verdict line and nothing else. If you are deciding whether to trust this publication, start by reading the misses. They are the most informative documents here, and they sit at the same link depth as the hits.
Silence
Some days nothing publishes. That is deliberate. Posts are triggered by the record moving: a call registered, a tier changed on real evidence, a grade issued. When nothing met that bar, nothing ships. A quiet day is the bar holding, and it is what makes the loud days mean something. You will also periodically see a rejection log: evidence that looked like confirmation and was declined, with the reason. Restraint is part of the record too.
Free and paid
Free, permanently: the calls and the grades. Every Registration, every graded result, every tier change, every rejection log, the quarterly calibration review, and the Ledger itself. If it is evidence about the record, you never pay to see it, and it is never truncated or teased. The daily desk notes are free too: name-level reads of the day's evidence, published in full. A single fragment is not the picture, and the fragments are how you decide whether the picture is worth paying for.
Paid: the assembled picture. The Weekly Mosaic, extended reads when one event moves multiple theses at once, and occasional deep reports that follow one claim or one name across months.
Paid access is $49 a month, or $490 a year. There is a founding tier at $1,100 a year, annual only, and it adds exactly one thing: a written Q&A published alongside the quarterly calibration. Nothing else, ever. No tier buys access to a position, a target, or a private read.
Nothing is gated yet. Everything publishes free until a full quarter of public resolutions exists, and the calibration post that closes that quarter is where the paywall arrives. The price is stated here now, in advance, for the same reason the calls are.
You will never find sizing, entries, or price targets here. What you get instead is what has to sit underneath them: the constraint, the companies positioned on it, the evidence as it lands, and a public grade on whether the read was right. Positioning depends on capital, horizon, and risk tolerance no publication can see, so it stays yours. This is general research, published in full and graded in public, not investment advice, and nothing here is tailored to any individual.
Cadence promise: never more than three emails in a week, and the cap is structural, not aspirational. One weekly, guaranteed. Desk notes, at most two emails per week, and fewer in any week that carries the quarterly calibration email, so the cap holds. Extended reads publish to the site and are linked from the weekly rather than mailed. Graded results are not emails at all: they publish to X and the Ledger the day they grade, and the weekly recaps them.
One sentence on how this works
MOSAIC is a research system: it evaluates evidence and maintains the Ledger continuously, and the Desk owns every judgment published here. That is why the cadence never slips and the grading never softens.
The standing invitation
Do not take any of this on faith. That is the one thing this publication will never ask of you. The Ledger is live, every claim above is checkable against it, and it only gets longer.
Ledger: https://themosaicledger.com/scorecard
MOSAIC Desk