Zed founder Nathan Sobo's first IDE, Atom, used web technologies (creating Electron) for maximum extensibility. This drove rapid adoption but hit a performance wall that required a complete rewrite. Performance cannot be added later; it's baked into the initial architecture choice.

Unlike screen-reading bots, web agents can leverage HTML's declarative nature. Tags like `<button>` explicitly state the purpose of UI elements, allowing agents to understand and interact with pages more reliably and efficiently. This structural property is a key advantage that has yet to be fully realized.

Despite offering modern browser interfaces, the company found that expert data entry clerks were significantly faster on old text-based "green screen" terminals. They could type without looking at the screen, using muscle memory for tabs and function keys, making the modern UI a downgrade in efficiency.

To appeal to the "layperson" rather than tech early adopters, Comet's designers made the core browser experience familiar, like Google Chrome. This reduces cognitive load, allowing users to focus their limited learning bandwidth on the novel AI features, even if it disappoints power users expecting a radical redesign.

Standard file formats like .docx and .pptx are filled with complex code that LLMs struggle to parse. To build effective AI workflows, companies must create deliverables in formats that are both human-readable and AI-friendly. HTML is a prime example, as it is visually appealing for people and easily ingested by AI.

The current chatbot interface is not the final form for AI. Drawing a parallel to the personal computer's evolution from text prompts to GUIs and web browsers, Marc Andreessen argues that radically different and superior user experiences for AI are yet to be invented.

For decades, the goal was a 'semantic web' with structured data for machines. Modern AI models achieve the same outcome by being so effective at understanding human-centric, unstructured web pages that they can extract meaning without needing special formatting. This is a major unlock for web automation.

The slow development of consumer-facing crypto applications isn't a sign of failure, but a constraint of "block space"—the capacity for on-chain computation and storage. Just as low bandwidth throttled the early web to text-only sites, limited block space gates crypto apps to simpler financial transactions for now.

Markdown, originally designed for blogging, has emerged as the de facto standard for interaction between LLMs and tools. This happened not by design, but because it's human-readable, highly token-efficient compared to alternatives like HTML, and familiar to the early adopters who trained the models.