The Unreasonable Inefficiency of HTML

Yes, this is an argument against HTML rendered in HTML.

The irony is acknowledged and building this doc has only moved me more strongly into the anti-HTML camp (for now). Let's get into why.

Anthropic thinks HTML is superior to markdown files for agents to communicate with humans. I agree. But using HTML as the core unit of collaboration (between humans and agents) today presents a meaningful product experience regression vs. something like Notion or Google Docs, which hundreds of millions of people use to share, comment on, and edit rich documents today.

Notion

Mature ecosystem. Mobile native, persistent auth, threaded comments, indexed by Claude and every enterprise agent that matters.

HTML artifacts

Expressive and agent-native. But isolated from the ecosystem until a Moda-class platform reaches Notion-grade polish.

The five pillars

Mobile

Notion

Native iOS and Android apps. Offline mode. Opens in one tap from any link.

HTML

Browser-only. Rarely responsive. PWAs exist but lag native feel.

Collaboration

Notion

Inline comments, threads, @mentions, page discussions, resolutions. Years of refinement.

HTML

Read-only by default. Moda ships commenting but it's new and unproven at scale.

Auth

Notion

Persistent login across all devices. SSO, SCIM, enterprise-grade. Click link, see content.

HTML

Separate auth surface per host. No SSO maturity. Every link is a login prompt.

AI discoverability

Notion

Indexed by Claude, Notion AI, every enterprise agent. Native MCP. Years of corpus already there.

HTML

MCP integration exists but the platform must be the canonical store. Most aren't.

Editability

Notion

WYSIWYG. Click and type. Fix a typo in two seconds without thinking about markup.

HTML

Find the right tag inside style attributes and class soup. Every edit is a search.

The pro-HTML case (steelmanned)

HTML is the format agents naturally excel at for human consumption. Information density, interactivity, SVGs, live previews, copy-as-prompt buttons. Markdown is fine for agent loops; HTML wins when humans are the audience.

The collaboration, auth, and discoverability gaps are solvable platform problems. Moda is the proof-of-concept. The bet is that purpose-built HTML platforms compound faster than Notion can bolt rich visuals into its block system.

Why the timing argument wins right now

scale

Notion: 100M users, 4M+ paying, >50% of Fortune 500. Moda: ~3K users at launch. The gap is not bridgeable in the timeframe most teams plan against.

moat

The existing Notion corpus is the moat. Fragmenting it into HTML islands erodes the AI advantage faster than HTML platforms can rebuild it.

velocity

Notion is shipping custom agents, MCP, rich blocks, presentation mode. The expressive gap is closing inside the ecosystem, not outside it.

switching

For any working team, the switching cost from "good enough integrated" to "polished but isolated" is the killer. Local polish doesn't beat systemic friction.

Bottom line

The pro-HTML side is right about the direction. They're wrong about the unit of exchange today. Until a Moda-class platform reaches Notion-grade polish across all five pillars, HTML artifacts are local wins that create systemic costs. Bet on Notion to close the expressiveness gap before HTML platforms close the ecosystem gap.

Subscribe to Jareau's Newsletter

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe