The short version. Aldous's structure is professionally sound and his positioning instinct — don't enter the crowded "AI due diligence tool" category; sell the investment lifecycle — is not just right, it is our story: the deal engagement holds the diligence weeks, the fund record holds the decade, the instrumented stream holds the business itself. He arrived at our architecture from the marketing side without knowing it.
The problem. Roughly a third of the skeleton's specific claims describe a product the seed commits us not to build (an automation pipeline ending in "Decision Ready"), or a capability we have barred ourselves from demonstrating (ask-anything recall), or outcomes we cannot substantiate (time-compression percentages with no pilots). Building the site as written would contradict the record we keep — and for a product whose entire pitch is that claims trace to evidence, the marketing site is itself an exhibit.
What this note does. Keeps Aldous's skeleton, replaces the dishonest organs. Section 2 is the claim-by-claim table. Section 3 states the positioning we own. Section 4 is the proposed site map. Section 5 handles the domain question. Section 6 asks five decisions.
Four things, worth keeping whole:
The category dodge. "AI due diligence tool" is a crowded category converging on identical claims. His counter — position across the investment lifecycle with room for portfolio monitoring — maps one-to-one onto the three-element package this project already built on the record: diligence weeks, fund decade, operational visibility. We do not need to grow into his positioning; we already are it.
Outcomes, not features. Correct discipline, and one of his three example outcome lines is our native truth stated better than we have stated it: "Generate investment memos with every conclusion linked to supporting evidence." Provenance reachable from every claim in one step is the seed's own sentence. That line survives to the hero section almost verbatim.
Jobs-grouped features and the security section up front. Both right; the security questions he predicts (where is data stored, is it used to train models, who can access what) are questions our substrate answers unusually well — and "is customer data used to train models" has a flat no behind it that most of the category cannot give cleanly.
Show, don't ask visitors to imagine. Right — and we have exactly one thing that is genuinely demonstrable today: the door-3 creation experience, a firm's own diligence playbook becoming a governed engagement in one sitting. That is the product tour.
| Aldous's element | The conflict | What we say instead |
|---|---|---|
| Workflow diagram ending "…Risks Detected → Memo Generated → Decision Ready" | An automation pipeline. The seed: no automated state transitions; the system surfaces and signals, the Operator approves; the walkthrough: not a scoring model, not an automated recommendation. "Decision Ready" as a machine-produced state is the exact category error the architecture exists to refuse. | The workflow ends at the partner: "…Evidence organised, provenance intact → You decide." The refusal is the differentiator, not a limitation — every claim in the memo traces to its source; the decision is yours and defensibly so. |
| "Founder claim validation" (automated) | The pain-points mapping ruled this claim out explicitly: the record makes divergence durable and visible; it does not hunt automatically. | "When a founder's claim and the evidence diverge, the divergence cannot be smoothed out of the record afterwards." Stronger, and true. |
| "Ask questions across every deal your firm has ever reviewed" | The recall claim, barred: the no-live-recall-demonstration rule stands until §18.2 lands, and a marketing promise is a standing demonstration request. | The record claim, not the recall claim: "Every deal your firm has ever reviewed, in one governed record — with provenance." Search-shaped promises wait for the retrieval work. |
| "Reduce diligence from weeks to days", "Reduced diligence time by 60%", "memos in under 10 minutes" | Fabricated outcomes. No pilots exist; the customer-stories section has no honest content yet. | The pilot-invitation posture Aldous himself suggests for early stage — and our continuous-diligence truth: the record does not close at the decision; it carries through the hold to exit. Numbers arrive when pilots produce them. |
| "AI-powered" as the lead adjective; "AI Operating System" | The product-identity standing note: the Companion is the entity; the AI is invisible — no model names, no AI-forward identity. Leading with "AI-powered" also files us into the category Aldous told us to dodge, in the same breath. | Lead with the record and the guarantees. Where machine work appears, say the true and differentiating thing: machine-assisted contributions are marked, permanently, and the marking cannot be turned off — the one AI sentence in the market that builds trust instead of spending it. |
| Integrations grid (Slack, Notion, CRM, virtual data rooms…) | None of these adapters exist. The upload pathway is real; live-system feeds are the adapter roadmap. | "Your documents, in one place from day one" (true), and a roadmap-honest line on system feeds. No logo wall for connections we have not built. |
| Primary CTA "Book a Demo" | What is demonstrable is narrower than the skeleton assumes: door-3 creation yes; recall no; external-contributor isolation not until CR-2026-157 lands. | Primary CTA: "Book a working session" — the D-3 conversation posture from the FORAY investigation: we walk your own playbook into an engagement live (door 3, real), and we talk through the rest against the record. Honest, and more senior than a canned demo. |
| Three-tier pricing (Starter/Pro/Enterprise) | The settled pricing posture is annual license anchored to the fund's life, not seat tiers. | A pricing page that says exactly that, plainly: licensed to the fund, priced against the vehicle, one conversation. Transparency about the model satisfies Aldous's underlying point without inventing tiers. |
One more, structural rather than claim-level: Aldous's "Fund Operations" feature group (deal workspace, collaboration, knowledge management, audit trail) reads as an afterthought bucket in his skeleton. In ours it is the second product — the fund record vertical, the decade, the across-tenure test. It gets promoted from feature group to headline.
Not "AI operating system for investment teams" — the honest superlative is stronger and no one else can claim it:
The governed record of everything your fund knows — from first look to exit, with provenance on every claim.
Under it, the three elements as the lifecycle story: the deal engagement (diligence weeks — evidence-grade, corrections preserved, unknowns visible at the decision); the fund record (the decade — thesis trajectory, pass reasoning, institutional memory that survives partner turnover); operational visibility (the FORAY direction — the business's own attested account of itself, positioned as where this goes, clearly marked as direction). The differentiators section writes itself from the substrate guarantees, each one a sentence that is checkably true: provenance on every claim, not every document; corrections preserved, not smoothed; machine origin marked, unsuppressibly; attestation at the moment of the act; the partner approves — always.
Aldous proposes doneinseven.com — the root. Current state: the existing marketing site sits at loomworks.doneinseven.com (Cloudflare Pages, loomworks-marketing repo), and it markets Loomworks-the-environment, not the venture pair. Three shapes:
Recommendation: A, built lean — the FORAY and OVA protocol pages can be single paragraphs at launch; what must be good is the venture path. Whatever is chosen, Claude Code should first report what currently answers at the doneinseven.com root before anything is pointed at it.
D-1 — Site scope. A, B, or C above. Recommendation: A, lean.
D-2 — Positioning line. Confirm Section 3's line (or edit it here, in conversation) as the hero and the standing one-liner. Recommendation: confirm and let the Ogilvy audit pass at copy stage sharpen it.
D-3 — CTA posture. "Book a working session" over "Book a Demo," per the demonstrability boundary. Recommendation: confirm.
D-4 — Claim discipline as a standing rule for this site. Every capability claim on the site must be either substrate-true today or explicitly staged as direction — the Section 2 table as precedent, enforced at every future copy change. Recommendation: adopt; it is the marketing form of verification honesty.
D-5 — Commission the copy draft. On D-1 through D-4, I draft the full site copy (per-section, Ogilvy-audited) as the next artifact; the build itself is then a loomworks-marketing-repo arc for Claude Code with its own handoff. Recommendation: commission.
A closing note on Aldous: the right reply to him is warm and true — the structure is adopted, the positioning instinct confirmed our own architecture, and the places we diverged are the places our product's honesty is the product. If useful, I will draft that reply.