1. Complex web interfaces are dead in the long term. They are solving human perception limitations that will no longer be relevant. No one will want to learn how to use complex UIs in the future. If this is your app’s differentiator, move fast.

  2. Personal agents will win over dedicated app agents. People are getting used to agentic experiences, so building a dedicated agent for your app may seem like a good idea, but this is just a “faster horses” solution. No one will browse to 4 different websites to use 4 different agents. They’ll use their own single agent and won’t care how your websites even look.

  3. The new frontend will be UI snippets that will be returned from MCP servers to be rendered in your personal agent. The agent will control the experience, and will display only the specific visualization or interactivity piece that you need from every app.

I envision a future where my LLM client is more akin to a web browser, observing most DOM events (discarding most), allowing me to have tabs that are general chats, tabs that are browsing/resource interaction sessions (with chat about them), etc.

If you’re reading this and thinking “there is no way I’m giving up on dashboards” - you’re still thinking like a project manager and not like the US president. The president doesn’t look at dashboards, he has helpers who look at them and extract only what he needs to know. All visualization is a way of answering questions and make sure you’re not missing anything. Eventually it’s either “this is going well” or “this requires attention”.

And you’ll be interrogating things through the LLM and asking for proof along the way (this is where the UI embeds come in).


Keyboard Shortcuts

Key Action
o Source
e Edit
i Insight
r Random
h Home
s or / Search
www.joshbeckman.org/notes/909050201