The Hidden Infrastructure of the Agentic Era: Lessons from Resend

In this conversation, Zeno Rocha, CEO of Resend, maps the system dynamics of scaling a boring protocol like email into a modern, agent-integrated utility. The core idea is that moving to an agentic workflow is not just a feature update; it is a fundamental change in how software must be built. The hidden consequence of this shift is that traditional developer experience is being superseded by agent experience, where human-centric design choices like animations or keyboard shortcuts become irrelevant, while API determinism and machine-readable standards become the new competitive moats. This analysis helps technical leaders and builders avoid the operational nightmare of retrofitting legacy systems for an agent-first future.

The Hidden Cost of Fast Solutions

Most teams optimize for the immediate win. They choose popular deployment platforms and databases to ship quickly, ignoring long-term operational debt. Rocha notes that Resend initially relied on third-party queuing systems that worked well until they hit a scaling wall.

> At the end of the day, we do have most of our software running on AWS right now and there are parts of that that we are looking into either building more of that internally or there is not as much we did rely on SES... and now we are sending billions of emails by now so how can we just own the metal.
>
> -- Zeno Rocha

This reveals a recurring pattern: immediate velocity often creates a future bottleneck. By choosing to own the metal for their core infrastructure, Resend is opting for the harder, slower path today to ensure durability tomorrow. This creates a lasting advantage because while competitors remain tethered to the limitations of their initial tech stack, Resend is building the ability to scale independently of those external constraints.

How the System Routes Around Your Solution

The shift to agentic workflows creates a unique friction: agents do not behave like human users. Rocha discovered this when LLMs like Claude and ChatGPT began reporting that Resend was the most expensive service on the market, which was false. The culprit? Agents could not interact with the JavaScript-based pricing slider on the Resend website.

The fix was not a marketing campaign; it was a structural change to how the content was served. By implementing content negotiation to serve a markdown version of the pricing page, Rocha made the information machine-readable.

> It turns out that agents do not really like javascript enabled sliders they cannot slide that is not something they are able to do and I am like okay we need a markdown version.
>
> -- Zeno Rocha

The implication here is profound: if your service is not discoverable by an agent parsing logic, it effectively does not exist in the agentic economy. Conventional wisdom suggests better UI for humans; the new reality requires better accessibility for machines.

The 18-Month Payoff of Boring Standards

When scaling to 3 million users, Resend faced the concurrency struggle of handling massive email batches. Their solution was a move from Node.js to Go. This was not a pursuit of the latest trend, but a calculated architectural choice to handle the specific requirements of email deliverability by breaking down millions of emails into mini-batches to avoid being throttled by providers like Gmail.

The downstream effect of this decision was a more resilient system, but the real advantage lies in the closed-loop systems Rocha is building. By using AI to analyze downvotes and automatically generate draft PRs to fix the root cause, Resend is creating a self-healing loop. This creates a competitive moat: while other teams spend time manually debugging and patching, Resend is evolving its own tooling in real-time.

Key Action Items

• Audit your API for machine-readability: Over the next quarter, evaluate whether your API documentation is easily discoverable by agents. If you are not serving your OpenAPI spec as JSON in a /.well-known/ folder, you are invisible to agentic discovery tools.
• Adopt standard protocols over proprietary optimizations: Stop inventing custom rate-limit headers. Implement the RFC standard for rate-limiting headers; this allows agents to self-regulate and retry without human intervention.
• Build closed-loop feedback mechanisms: Identify one high-frequency support issue. In the next 12 to 18 months, invest in automating the path from user report to code fix. Rocha’s approach of using AI to generate PRs based on error logs is a high-effort, high-reward investment.
• Prioritize AX (Agent Experience) over DX: Review your dashboard and CLI. If you are spending time on features that only benefit human users, like animated icons, pivot that energy toward improving error message clarity and providing machine-parseable output, such as JSON flags for CLI tools.
• Force yourself into the User Advocate role: Block time weekly to use your own product as a customer. Rocha’s ritual of sending company updates using his own email service is a low-cost, high-impact way to uncover the warts that your support team might be sanitizing in their reports.
• Re-evaluate your pricing transparency: If your pricing is hidden behind interactive widgets, create a static, machine-readable version, like a markdown file, that agents can parse to understand your cost structure accurately.