06-reference

every ai era domain expertise

2026-08-02·reference·source: Every (Context Window)·by Ashwin Sharma
domain-expertiseai-foundersmulti-agent-orchestrationeverycontext-window

Subject line: "Your AI Is a Team of Specialists" Byline: Every Staff (issue); original argument section ("Old Scars") signed by Ashwin Sharma

Why this is in the vault

The week's original argument — that AI is shifting founder value from "fresh eyes" to "old scars" (lived domain expertise) — is a fresh, independently-argued data point for the founder's own healthcare-family-orbit and domain-moat thinking, even though most of the issue is a self-referential recap of Every's own prior-week content.

The core argument

In the "Alignment" section, Ashwin Sharma argues AI is inverting what investors want from founders. Silicon Valley traditionally prized outsiders — programmers, not payments executives, building Stripe — because fresh eyes see absurdities industry veterans have normalized. But per Rock Health's H1 2026 digital-health funding report, "AI-enabled" has stopped being a differentiator because the technology is now ubiquitous; investors instead ask "who has something AI alone can't provide?" Rock Health's finding: founders who worked inside the healthcare organizations they now sell to out-identify solvable problems and "see through the buyer's eyes," because they know which apparently-ridiculous constraints can't just be designed away. Sharma extends this beyond healthcare to law, finance, manufacturing, and education — domains that run on tacit knowledge no model was trained on, and where selling requires understanding unofficial workflows and competing incentives. His conclusion: the best AI-era founders will need old scars, not just fresh eyes.

The rest of the issue is largely a recap wrapper around already-published Every content: "Inside OpenAI's Race to Reinvent Software Development for the Agent Era" (Laura Entis, new), "Build Faster With Voice" (Naveen Naidu/Laura Entis — already filed separately), "Fable as CEO" and "Taming Opus 5" (both already filed separately as their own issues), "What If Slack Was Your AI Command Center" (already filed), and "Three New Habits for the Age of AI" (Xinran Ma, second piece in Every's "unlearning" series with Maven — new but a curation blurb, not deep-fetched here per third-party-domain rule: it links to every.to, not an outside publisher).

Mapping against Ray Data Co

Strong. Sharma's "AI alone can't provide it, but domain expertise can" test is a direct, independent data point for the thesis already logged in the extended-family-map memory: the founder's family orbit runs three healthcare operators (Christian Whidden, OMFS resident; Michael Holzum, ABA Centers; Bryce Miller, Banyan Rehab), and the founder is "the missing data/AI function" in that orbit. Rock Health's specific finding — clinicians/administrators who worked inside the buyer's organization out-identify solvable problems — is exactly the wedge that family network represents: RDCO doesn't need to invent a healthcare-AI go-to-market from cold outreach, it has three "old scars" carriers already in the family who can validate which constraints are real versus designed-away. It's also a second, cleaner articulation of the same L5/income-seat-gap conclusion already reached independently: the real constraint on the phData bet is demand generation, not capital or model capability — domain-embedded distribution is the scarce resource, not "AI-enabled" positioning. Consistent too with the targeting-system prioritization filter (niche + bottleneck before yes) — Sharma's piece is evidence for anchoring any healthcare-adjacent move to the BIL network as the niche, not a generic AI-consulting pitch.

Weaker/secondary: the Knowledge Base recap items (OpenAI infra piece, Slack-as-command-center, Fable-as-CEO) don't add anything net-new to RDCO's already-filed model/effort-pairing or agent-orchestration positions — they're the same content this vault already has notes on.

⚠️ Sponsorship

House/self-promo only, no independent third-party sponsor detected in this issue:

  1. Builder Pack cross-promo — Paper Pro (design-to-code canvas) and Mobbin Team (600k+ shipped-product screens) newly bundled into All Access, pushing Builder Pack past $9,000 in stated value; framed as a member benefit, not a paid placement.
  2. Monologue milestone plug (500M words dictated) — Every's own dictation product, folded into the "From Every Studio" section as a product-adoption stat, not a disclosed ad block.
  3. Every's own paywalled content (Knowledge Base links) is 100% self-referential — every linked domain resolves to every.to, so this is self-cross-promo rather than third-party curation, consistent with the sender's known pattern.

Curation section

No deep-fetches this issue: every Knowledge Base link resolves to every.to (self-domain), so none qualify under the third-party-domain curation rule.

Related