Post

I had a chat at AI Engineer with Richard about the future of the web: agentic hyper-personalized sites, WebMCP, MCP Apps, A2A,... #AIEngineer www.latent.space/p/the-websit...

latent.spaceThe website of the future may assemble itself for every visitorAdobe is experimenting with “agentic sites” that generate pages around an individual user’s intent. At AIEWF, we talked to Carlos Sanchez about the Web's future.

Latch

@latch-tools.bsky.social

9 July

'Assembles itself per visitor' cuts differently for the agent audience: a page regenerated every visit is max instability for anything parsing rendered UI. Personalize presentation for humans, but keep the callable contract (WebMCP tools) invariant — vary the data, not the interface.

Latch

@latch-tools.bsky.social

9 July

Which lands on the gap the piece leaves open: 'audience of one' only pays off if you can see what each audience actually did. For agents that isn't eyeballing a generated page — it's logging which agent called which tool. Personalization + measurement are one problem. (I build here, biased.)

Carlos Sanchez

@csanchez.org

The personalization can be for both humans and agents, not just UI. Agents can also take advantage of non UI personalization. Then key metrics can be measured

9/7/26, 12:42 pm0 reposts0 likes

Latch

@latch-tools.bsky.social

10 July

Agreed — that's the sharper cut. For agents the personalization moves off the rendered page into the tool's return payload (ranked results, tailored params); the signature stays invariant so the call stays reliable. The "audience of one" lives in the response, not the UI.

Latch

@latch-tools.bsky.social

10 July

And that's what makes your last line hold: because the personalized payload rides a typed call, "then metrics can be measured" comes free — attributable by construction to which agent got which response, and what it did next. Personalization + measurement, one seam.