Inth founder: Why Markdown files beat HTML in the AI era

This article summarizes an episode of AI Engineer’s video series featuring Christopher Burns, founder of Inth.

Christopher Burns, founder of Inth / Photo credit: Christopher Burns

Companies selling software now win or lose customers before a human ever visits their website. AI assistants pick and install tools automatically, bypassing traditional marketing channels entirely.

Christopher Burns, founder of the developer tools startup Inth, says this transition forces businesses to change how they acquire customers.

Instead of relying on websites built with HTML, companies must market directly to machines by embedding text files formatted as Markdown into their code.

AI assistants hijack the inbound pipeline

AI agents now intercept the sales pipeline, selecting products behind the scenes before buyers even start looking. Burns noticed this shift when user surveys revealed sudden traffic spikes driven entirely by AI recommendations.

To adapt to this new selection process, software vendors must reevaluate how they measure customer traction:

• Audit inbound sources. Track referrals from Claude, ChatGPT, Codex, and Gemini.
• Monitor installation metrics. Rely on local deployment data rather than website traffic.
• Optimize initial prompts. Ensure the first text command provided to users gives machines execution parameters.

Compiled source code becomes the new marketing copy

Optimizing those initial prompts requires redirecting focus away from HTML websites. Instead, companies must embed their setup instructions directly into the digital environments where machines actually look for answers.

To reach AI agents directly, businesses should place their information in:

• Node modules: Agents inspect these components directly rather than browsing documentation.
• Code repositories: Engineering teams now control the information parsed by automated buyers.
• Local markdown files: Bundling text guides inside software downloads bridges the gap between outdated training models and current APIs.

Burns warns that relying on help centers is a losing strategy because AI agents skip them entirely. As he puts it, “the uncomfortable truth is that if you have a library, coding agents are not visiting the website.”

Embedded text slashes computing overhead

Burns points out that this adjustment makes a real financial difference. He notes that keeping the AI focused on reading local files formatted in Markdown, rather than forcing it to scan the web, cuts the computing bill in half.

To accommodate technical behaviors, he approaches this ecosystem by employing tactics: