stage: vision
product: prism-site
status: APPROVED — P6 2026-07-16
skills_invoked:
- creative-brief
- marketing-strategy
owner: שרה (brief) + דן (strategy), reviewed by גל
date: 2026-07-16
APPROVED at P6, 2026-07-16. Primary audience: prospective clients. Public name: PRISM.
Company homepage: one page. Portfolio: separate property. Brand: Glass Box. The product also
includes a multilingual AI publication and organic-search property engine. Receipt:
decisions/p6-20260716.yaml.The public face of PRISM: the company site for a solo-founder AI agent company that runs a
live portfolio of products (ben-portfolio, foodflow-home, plakton, and whatever SEED-002 spins
up next). It is not a landing page for one product — it's the factory's storefront: proof
that a governed AI organization (P6 human gates, capability registry, blueprint model, live
portfolio) actually runs, with links to the running proof, not screenshots of it.
Research (research/research.md §4) surfaced four plausible primary audiences that pull the site
in different directions. This vision draft is written to hold up under any of the four, but the
copy, structure, and CTA hierarchy change materially depending on which one wins:
project")
blueprints, capability registry, "how this is actually built")
the engine behind a visible body of work, not a product being sold)
P6 selection: prospective clients are the primary audience. Partners, talent and Ben's public
audience remain secondary readers and do not control homepage copy or CTA priority.
PRISM is also a serious multilingual publication about applied AI. The publication compounds
organic demand by turning verified research, implementation experience, governance patterns,
tools and case evidence into topic clusters with stable URLs. It is not a generic high-volume AI
blog and does not publish model-written claims without sources and expert accountability.
The company homepage remains one page. The publication necessarily uses separate canonical topic
and article URLs so search engines and readers can discover, cite and revisit individual assets.
Hebrew and English are the initial locale pair. Additional languages enter through an evidence
gate based on search demand, editorial capacity and localization quality; they are not blind
machine translations.
Every content property must support canonical URLs, hreflang, locale-aware slugs, structured data,
author/expert ownership, provenance, update dates, internal links, conversion paths and measurable
search outcomes. PRISM remains the master brand; topic hubs may become separate properties only
after evidence of demand, use and revenue.
One sentence the site must prove, regardless of audience: **"This is a real, governed AI
organization that ships live products — here is the portfolio to prove it, and here is how the
governance actually works."** Every other page/section serves that one claim.
The primary business success criterion is at least one qualified inbound inquiry within 30 days of
launch. The publication adds a compounding measurement layer: indexed pages, qualified non-brand
impressions, topic-cluster coverage, assisted conversions and content refresh health.
touch to an investor/partner conversation without needing a deck alongside it.
days of a single share (e.g. LinkedIn/X post linking the site).
2 weeks, replacing whatever is there now.
Baseline measurable regardless of audience choice: Lighthouse 95+ (matches existing PRISM web
standard — ben-portfolio is 99–100), live portfolio links resolve (no dead cards), zero invented
metrics (constitution honesty rule — unconfirmed claims marked unconfirmed and gated).
Anchor move — see research §2 — and it contradicts the ownership-not-lock-in positioning).
separate property. The multilingual publication has its own topic and article routes because
stable content URLs are required for organic discovery.
constitution — this isn't a careers page).
constitution honesty rule applies here as strictly as any product.