A full research desk that reads the filings, builds the case, and attacks it before it counts — built on AI, run for one owner. You can't hold a position you don't understand.
Institutional research was never sold to private investors — it's written for trading relationships you were never part of. What's left over is a thin substitute. This closes the distance between the two.
Every report comes with the analyst who wrote it. Question them directly — a live interview, just like you're sitting across the table from them. Ask "why" as many times as it takes to actually understand it; every session doubles as your own private lesson.
Professional coverage is decided by banking relationships and trading volume — which is why whole sections of the market go unwatched. This universe is decided by the data instead: only companies whose primary filings are complete enough to build a real evidence base on. Nothing is skipped because it was inconvenient to cover. The full universe runs every night; the shortlist below is reviewed and ranked weekly.
Quality, growth, and valuation are scored together, not in isolation — a cheap company with deteriorating economics doesn't survive the composite any more than a great business at an indefensible price. A separate Sector Scan can be requested at any time, entering the pipeline outside the weekly cycle.
The universe is bounded by data coverage, not by architecture. The pipeline already runs independent data sources side by side, routing each company to the right one — so widening it changes what goes in, and nothing else. Every filter below stays exactly as it is. The standard wouldn't move; only the pool it selects from.
An institutional analyst carries dozens of names off a quarterly update and a maintained model. Here, each company gets an evidence base built from the filings themselves — annual reports, quarterly disclosures, material events, and management's own words on the record, accumulated over a full cycle rather than sampled at a point in time.
Why it matters: that depth is what makes management's record checkable. Promises made in one year can be read against results delivered three years later — the single hardest thing to fake, and the first thing a summary throws away.
Each phase reads the work before it and tries to break it, then writes its own standalone document — handed forward intact, never merged. Disagreements survive to the final report instead of being averaged away in a summary. And the agreements mean something, because they survived a deliberate attempt to dismantle them.
Establishes what the business actually is, from the filings up — history, structure, economics, and the case for owning it.
Puts a number on it. Intrinsic value under bear, base, and bull cases, each probability-weighted, with the assumptions stated.
Attacks the thesis on its own terms. Factual errors are corrected outright; judgment disputes are escalated, not quietly settled.
Reads all three, then writes its own — a fourth independent report, carrying its own recommendation. Free to disagree with any of them, required to say why.
Not a fifth report — this is what happens after. Every decision updates the benchmarks the next one is measured against: valuation comparables, risk patterns, recurring errors.
A rating on its own is worthless — it can't be checked, and it can't be held to account. Every recommendation arrives as a defensible structure: what it's worth, under which assumptions, with how much room for error, and the price at which it becomes actionable.
Most research goes stale the moment it's filed — a desk revisits a position when its analyst gets around to it. Every position here carries a live trigger instead, so the moment price reaches the level the thesis named, you hear about it.
Every company the system has formed a view on, with its current decision and the reasoning behind it. A standing register you can interrogate — not a folder of documents you have to remember to open.
Each open position carries the price the thesis said mattered. That level is checked continuously, and crossing it raises a real alert — not a note you find later.
Exactly where price sits against a position's watch, buy, and add levels — so "how close is this?" is answered at a glance instead of recalculated each time.
What's changed on the companies you actually hold or watch. Deliberately scoped rather than a full-universe feed — a short list worth reading daily, instead of a firehose you learn to ignore.
A timestamped record of every decision to start research on a company. Not just what was concluded — when the system decided to look, and on what basis.
ASX, HK, LSE, and US, each tracked on its own market hours. Positions don't go unwatched because they trade while you're asleep.
The same categories an institution would commission — each one tied to a real stage of the system rather than invented for a brochure. Some you ask for; others simply keep running.
Institutions have research desks — analysts paid to be right about the handful of companies their firm has chosen to cover. We had what everyone else has: notes written for someone else's book, headlines written for clicks, screens that rank without ever explaining themselves. You can read all of it and still not know whether a business is any good. The alternative on offer was to hand the decision to a fund manager whose incentive is to keep us as clients, not to make us better investors.
None of that was good enough to risk our own capital on, so we built the desk instead. It turns out other private investors would rather own one than rent someone else's conviction.
A spreadsheet shows its working. This shows its reasoning — why this business and not the one beside it, what the bear case sounds like when someone argues it properly, which historical parallels hold and which break down, where two analysts disagreed, what each one actually said, and which of them gave ground.
Follow it long enough and the conviction ends up being yours. AI is what makes that possible outside an institution — evening up odds that were never in your favour.