Overview of 1015: Browsers and UIs are dead. Everything is chat
Recorded live at JS Nation in Amsterdam, this episode is a conversation between Wes Bos and Scott Tolinski about the increasingly loud claim that chat interfaces, agents, and AI-generated UI will replace traditional websites and apps. Wes is preparing a talk on agentic interfaces, generative UI, and WebMCP, and the discussion centers on a nuanced view: chat is powerful, but it will not replace every UI. Instead, the future likely includes a mix of clicking, prompting, voice, automation, and AI-generated interface elements, depending on the task.
Core argument: chat is not the universal interface
The hosts push back on the idea that “everything will be chat.” Their main point is that different tasks need different interaction modes:
- No UI at all for obvious, repetitive automation
- Voice UI for hands-free or situational interactions
- Chat/prompting for batch tasks and busywork
- Traditional UI with buttons, lists, and visual context for browsing, exploring, and decision-making
Wes’s view is not anti-AI; it’s anti-“chat everywhere.” He argues that the best interface depends on the job, and a lot of web use cases still benefit from a real UI.
WebMCP, agentic interfaces, and generative UI
A major topic is WebMCP, which Wes describes as a way to expose the actions of a website as tools that agents can use. The idea is less about the spec itself and more about the experience:
- Let an agent interact with a website’s capabilities
- Let humans use the normal UI
- Let both access the same underlying actions
- Avoid forcing a chat bar into every site
They also discuss MCP apps and generative UI, where an AI can either:
- assemble a UI from known components, or
- generate HTML/UI on the fly
Their conclusion: the likely future is somewhere in between. The AI may help compose or surface interfaces, but well-designed product UX will still matter.
Examples used to explain where AI fits
Good fits for AI/chat
These are tasks the hosts feel are ideal for prompting or automation:
- Calendar management: paste email details and let AI add events
- Expense categorization: tedious bookkeeping in tools like FreshBooks
- Smart home automation: turning on lights, reacting to context, building dashboards
- Customer support/admin actions: refunding, looking up stats, performing repetitive actions
- Data aggregation across systems: calendar, groceries, fridge camera, etc.
Poor fits for “everything is chat”
They push back on using chat for experiences that benefit from browsing and discovery:
- Shopping: people often want to browse, compare, filter, and discover
- Grocery shopping: online ordering can be efficient, but in-store/market browsing is often more enjoyable
- Social media: people don’t want to type queries for every content decision
- Driving/car controls: chat is unsafe and awkward for many in-car tasks
They use the Soylent analogy to criticize the “optimize everything into a bland utility” mindset. Just because a task can be reduced to prompting doesn’t mean that’s the best human experience.
Concerns and tradeoffs
1. Browsers aren’t dead
The episode strongly suggests the browser is not going away. Instead, chat and agents are new surface areas for web applications—similar to how mobile and voice added new interaction layers rather than replacing everything.
2. Platform gatekeeping is a real risk
The hosts worry that a few major AI companies may become the new gatekeepers:
- Websites may need to surface through a handful of chat platforms
- Free tiers could manipulate results
- Companies may preserve dark patterns inside agent flows
- Some brands won’t want to become “just an API” to a chat app
3. AI-generated UI can be hit or miss
They note that vibe-coded or AI-generated interfaces can be surprisingly good for niche uses, but often lack strong UX. The best outcome may be:
- AI chooses among approved components
- The app owner retains control over product quality
- The user gets a more personalized, context-aware interface
Notable takeaways
- “Clicks and clankers” becomes a shorthand idea: sometimes you want to click, sometimes you want to prompt.
- Chat is great for busywork, not for every product experience.
- The web will likely become more agentic, but not fully chat-only.
- Context matters: AI is most useful when it has access to rich data and can act on it.
- Good UX still wins: the companies that know their product and data best will often make the best interfaces.
Closing question posed to listeners
The episode ends by asking the audience whether:
- the web is dead
- everything will move into chat apps
- or whether chat is simply one more interface layer in a broader, multi-modal future
The overall tone is skeptical of hype, but optimistic about practical AI tooling where it genuinely reduces friction.