The architecture

One source analysis. Three reading surfaces.

The whole product rests on one rule: never generate a surface from raw sources. A surface always projects from an existing analysis. That is what makes it one deep read per company, not three, and it is why every surface stays consistent with the others.

The canonical artefact

A structured due-diligence file, every statement a claim.

The source analysis is generated once per company and holds everything the surfaces draw on. It is organised as the four workstreams, stress-tested from both sides, and framed by the active sector pack. Each factual statement carries its source, page, provenance, and status.

  • FDDFinancial due diligence, figures recomputed from source cells
  • CDDCommercial due diligence, framed from the pack key drivers
  • LDDLegal due diligence, seeded from the regulatory landscape
  • Bear caseA forward downside scenario scaled to a named prior crisis
  • Bull-lever testIts symmetric counterpart, the long thesis levers
  • Sector contextNamed competitors and external, cited market reads

Triage · illustrative

DRAFT

Northfield Bearings Ltd

Industrial components · private · WATCH

  • Corroborated: Revenue growth, FY25 vs FY24+6.2%
  • Sourced: EBITDA margin trajectory18.4%
  • Management asserted: Net leverage covenant headroom0.6x
  • Needs analyst input: Pending litigation exposureOpen
corroborated58%

Illustrative Triage view. Company and figures are examples, not a real target.

Cheap projections over one expensive read.

Triage

≤ 1 page · 5 min · ungated

An untagged subset of the 2-3 pager verdict. A RAG call, the few things that matter, a drop / watch / deep-dive recommendation. Brevity is the authority.

2-3 pager

≤ 5 pages · 30 min · 60% gate

The tear-sheet. Four workstreams, the bear case and its bull-lever counterpart, and a recommendation. Targets the human analyst altitude, not exhaustive transcription.

Investment paper

20 pages · 90 min · 75% gate

The IC paper. Full thesis, modelled scenarios, sector read-across, named kill criteria. It may trigger a deeper analytical pass, still the same engine over a richer analysis.

Portfolio monitor

A two-page per-company surface for held positions. Info, notes, pricing, cap table, and manual financials.

Chat

Single-turn, answers cite the underlying claims, and refuses to go beyond the source.

The differentiator

Industry depth is a generation contract, not a prompt.

A sector is a methodology pack: key drivers, regulatory landscape, named competitors, key metrics, and the recurring analytical moves that are the analyst’s fingerprint. The engine must attempt a claim for every metric, surface every competitor, and perform, not merely name, every move. Where a pack element is not supported by the sources, it becomes an explicit open item, never a silent omission.

Packs are authored in a self-serve editor, versioned like code, and stamped onto every analysis they produce. Breadth comes from the engine and the authoring path, not a pre-built library.

The pack contract

  • A claim attempted for every key metric
  • Every named competitor surfaced in context
  • The LDD prompt seeded from the regulatory landscape
  • The CDD framed from the key drivers
  • Every analytical move performed, with real arithmetic

The rules that hold it together

Generate once, project many

The deep generation happens one time per company. Triage and the 2-3 pager are cheap projections over already-extracted claims.

Depth is data, not a code path

Fully-built and outline-depth are the same engine over a richer or thinner analysis. There is no second engine to drift.

The model never does arithmetic

Every figure is recomputed deterministically in code from pinned source cells. A figure that cannot be recomputed cannot reach corroborated.

One read. Every claim accounted for.