From Chatbots to Generated Front Ends

▶ Watch (00:01)

In 2023 and 2024, businesses raced to bolt chatbots onto their front ends. That era is ending. Generating an entire app, not just a chat widget, is now fast enough to be the default move. The goal shifts from writing prompts for a single AI agent to generating complete front ends that carry business context, call backend APIs, and handle both internal productivity tools and customer-facing interfaces. Dykstra calls these “bicoded” apps: AI-generated, not just AI-assisted.

What Industry-Specific Context Actually Requires

▶ Watch (04:32)

A supply chain studio needs to visualize items moving through a network, not just answer questions. A finance terminal must know the firm’s risk guardrails before it lets a trader place a limit order. A legal studio needs catalogued case documents loaded before it can simulate a court situation. None of that works with a generic chatbot. Space and Time wrote roughly 10 pages of instructions covering their database schema, read/write rules, and visualization options before Dreamspace could generate useful apps on top of it.

MCP and the Move Toward Standardized Tool Context

▶ Watch (08:35)

MCP (Model Context Protocol) addresses the same problem from a different angle. It creates a single registry of two-way connections between a business’s REST APIs and the AI agent calling them, with human-readable instructions for each tool. An agent using MCP knows how to send a Twilio SMS, post with the X API, or schedule a Google Calendar event without custom wiring. The standard signals a broader shift: the industry is converging on formalized ways to give AI agents structured knowledge about business tools, moving past one-off prompt files.

Dreamspace: Generated UI on a Domain-Specific Backend

▶ Watch (10:36)

Dreamspace sits on top of Space and Time’s blockchain database and was built with Microsoft AutoGen, OpenAI APIs on Azure, Kubernetes, Key Vault, and Fabric OneLake. It generates the app UI, handles industry-specific auth and paywalls, and connects the agent to the database with pre-loaded schema context. In the Dreamspace dog-walker app live demo (11:29), Dykstra opens a chatbot that already knows the app’s table structure, directs a form submission to the grooming sessions table, then drags in a new gradient theme in under a minute.

Notable Quotes

Dreamspace is an AI studio that lets you generate your app’s UI, Scott Dykstra · ▶ 10:36

I intend to place a limit order on this market. Scott Dykstra · ▶ 5:21

To be honest, it’s pages and pages of prompts Scott Dykstra · ▶ 7:35

Key Takeaways

  • Industry-specific AI studios need 10+ pages of written context about a business’s APIs and database schema.
  • MCP standardizes how AI agents receive human-readable instructions for calling a business’s REST APIs.
  • Dreamspace generates full app UIs on top of Space and Time’s blockchain database using Microsoft AutoGen and Azure.