# 101: Build something agents want

Canonical: https://mudpie.ai/101/
Updated: 2026-09-20
Curator: Ali Abouelatta (https://mudpie.ai/authors/ali-abouelatta/)

9 opinionated reading paths for the new agent economy. Mudpie supplies the TL;DR and reading order; every canon links to the original.

- [WebMCP](https://mudpie.ai/101/webmcp/) — The browser contract, the product decisions behind useful page tools, and the compatibility layers you need while the standard is still moving.
- [MCP](https://mudpie.ai/101/mcp/) — The protocol's original motivation, current architecture, secure transport and the maintainer decisions that turned a connector format into shared infrastructure.
- [AEO](https://mudpie.ai/101/aeo/) — What answer engines retrieve and cite, how unstable that evidence is, and which publishing changes deserve to be tested instead of repeated as folklore.
- [SEO](https://mudpie.ai/101/seo/) — The durable search foundations, demand research and technical diagnostics that still matter when result pages, crawlers and interfaces keep changing.
- [Viral marketing](https://mudpie.ai/101/viral-marketing/) — The mechanics of compounding distribution, the product loops that make sharing useful, and the social triggers that make an idea worth carrying.
- [Agent-led growth](https://mudpie.ai/101/agent-led-growth/) — How discovery, evaluation, buying and product use change when software agents become a source of demand instead of merely another acquisition channel.
- [Personal assistants](https://mudpie.ai/101/personal-assistants/) — The half-century path from personal computing and calm technology to assistants with memory, tools, initiative and the ability to act across software.
- [ChatGPT ads](https://mudpie.ai/101/chatgpt-ads/) — The product contract, creative system, targeting inputs and measurement model for advertising inside an answer interface rather than a keyword result page.
- [Managed agents](https://mudpie.ai/101/managed-agents/) — The control planes, durable execution, sandboxes, logs and human approvals that let an agent keep working after the first model call ends.