Standing References
The working operating docs Andrew actually reads from, not archive.
Four documents, four different jobs: a design bar, a CRM operating system, a cost map, and the revenue plan. None of these are archived history, they’re what gets checked before the related decision gets made.
AI Design Quality Standard (v1.1)
The rule set fed to any AI tool generating design output, web pages, components, decks, graphics, so it produces premium human-crafted work instead of generic AI slop. Distilled from 400+ shipped Brand Identity Kit Generator component rounds, folded together with Anthropic’s frontend-design skill and the open-source taste-skill system (July 2026).
Structure: Part 0 forces a declaration before any code gets written, three 1-10 dials (Variance, Motion, Density) plus one named Signature element plus a one-screen color/type/layout plan. Part 1 is a six-axis self-critique (Philosophy, Hierarchy, Execution, Specificity, Restraint, Variety) scored 1-5, anything under 3 must be revised before it’s shown. Part 2 is the anti-AI-tell list, no emoji icons, no gradient text, no invented metrics, no left-border accent stripes, no default-AI palettes (editorial beige+brass+oxblood, near-black+acid-accent, broadsheet) or default-AI fonts (Inter, Fraunces, Instrument Serif). Part 3 locks tokens: one accent color per page, one radius scale, off-black/off-white never pure. Parts 4-10 cover hero composition, layout safety CSS, motion rules (transform/opacity only, house ease cubic-bezier(0.16, 1, 0.3, 1)), interactive states, copy voice, and accessibility floors. Part 11 is the verification ritual, an 11-step checklist run before any design is called done, ending in a brand-swap test and a 360px-to-1400px width drag.
Attio Customer Intelligence Operating System (v1, 2026-06-03)
The plan and runbook for Attio as the single customer brain: one record per human that knows who they are, what they bought, what they’re worth, how warm they are, and what to offer next. Baseline at write time: 11,658 contacts, 2,347 companies, 759 paying customers, $735,007.53 tracked revenue, 98 VIPs driving ~79% of it, 439 paying customers sitting in Win-Back.
Five intelligence layers stack on top of the clean spine: Layer 2 (revenue truth from Stripe + PayPal, Subscription Status/MRR/Churn Risk), Layer 3 (Gmail engagement state), Layer 4 (lead scoring + Next Logical Offer, an upsell ladder from AI Writing Starter Kit at $19 up through VIP/Premium), Layer 5 (future: transcript intelligence and social signals). Autonomy model is explicit: the agent runs hygiene, enrichment, tagging, and note-taking automatically; anything touching send, pricing, new offers, $1,000+, or schema changes escalates to Andrew. Cadence: daily revenue sync (~6am), Monday health sweep, Friday pipeline report, monthly reconcile. Section 7 logs hard-won tool gotchas: Attio’s filter API has no is-empty operator, Stripe’s MCP hides customer email (raw API GetCustomers is the workaround), PayPal can’t be attributed per-person at all.
Tool & Cost Map (2026-06-11)
One view of everything that costs money to run, split into predictable fixed subscriptions versus usage-based costs that can spike without warning. Fixed subscription floor: roughly $300-450/mo (Claude Max, ChatGPT, Notion Business, Canva, Gamma, Suno, Attio, Cloudflare base). Usage-based sits on top, variable, and Stripe processing runs ~3% of revenue on every sale.
The watch list, ranked by surprise risk: Cloudflare (R2 + Workers + D1, base $5/mo then usage), scheduled tasks that call paid APIs while nobody’s watching (morning-briefing 7am, stream-prep 9pm, toolbox-extract 4am), Gemini/Nano Banana 2 image gen (~$0.067-0.15 per image, no hard cap by default), Apify (compute units plus per-result fees, credits don’t roll over), Attio (seat $29-69 plus usage credits), Stripe fees (2.9% + $0.30), Notion Custom Agents ($10 per 1,000 credits, no rollover), Zapier (meters by task, 750 on entry Professional tier), and direct API-key usage separate from flat subscriptions (Opus $5/$25 per million tokens, Sonnet $3/$15). Recommends hard spend caps on Gemini/OpenAI, billing alerts on Cloudflare and Apify, and mapping every scheduled task to what it actually calls.
Roadmap to $1M (2026-05-30)
The GTM and monetization plan built on a two-engine split: Engine 1 is cash, agent-run, ~2 hours/week of Andrew’s time (approvals, QA, escalations). Engine 2 is product, Andrew-run, Brand With AI plus the blogging system. Starting point was roughly $8k/mo; the target was $15k/mo profit inside 60 days, then a 12-month ramp to a $1M run-rate by around month 11.
The hero move is the Automated Blog Engine: a done-for-you, brand-aware, AI-produced blog sold at $997/mo (or $1,500 setup + $497-797/mo), targeting 7 low-touch clients to add $7,000/mo at ~82% margin, dogfooding the same blog-system tech DH is already building for itself. Combined with Club growth (+$3,500), Podia evergreen (+$2,500), a packaged BrandBuilderBot (+$2,500), and a monthly paid workshop (+$1,500), the day-60 total lands near $25k/mo revenue, ~$20.5k profit at 82% margin. The escalation boundary is explicit: pricing, new offers, cold-outreach targets, and anything above roughly $1k commitment goes to Andrew; everything else (daily content, nurture, CRM hygiene, reporting, article production) runs on autopilot. The 12-month phase table tracks $18-23k (months 1-2) up to $96-116k (months 10-12) run-rate.