Agent-Era Engineering with Hector Ramos of Wallfacer:Your Agents Are Ready.Your Workflow Isn’t

This week I sat down with Hector Ramos, CEO and Founder of Wallfacer, who just recently announced a $1.2m round led by Maria Palma at Freestyle and Vermilion Cliffs Ventures.

Hector and I go way back. We worked together at Parse, the mobile backend-as-a-service that developers loved and Facebook eventually acquired. We spent a year traveling the world doing meetups, events, and conference talks together, riding the mobile wave when it was the hot new thing. It feels more than a little fitting that we’re back working together now, at a moment when AI is doing to software development what mobile once did: changing the conditions of the game entirely.

Hector spent a decade at Meta, growing with the engineering org from 3,000 to tens of thousands of engineers, and later worked on the React Native team. That background gave him a rare lens on what infrastructure actually has to exist for parallel, high-scale engineering to work. Wallfacer is his bet that that same infrastructure logic applies now. Not to thousands of humans, but to each engineer running multiple agents at once.

Listen in not just for the technical depth, but for Hector’s clear-eyed take on where today’s developer tools are already falling short, what teams that are winning in this new era are doing differently, and what it means for the craft of engineering when AI stops being an assistant and starts being a collaborator.

Full interview here with highlights below.

The tools got 10x faster. The workflow didn’t.

Most teams discovered AI coding tools the same way: one engineer brings it up, the team tries it, and suddenly features that were scoped for next quarter ship in a week. The productivity gain is real. But the existing workflow like the pull requests, the code reviews, the branching structure was never designed for this volume.

“A year ago when Claude Code had just come out…and we actually were shipping features in a week that we had budgeted for next quarter. So that’s when I first saw how these tools could really empower engineering teams to move faster. But quickly, I started seeing the issues. When every team member on your team is now 10x in their output, code reviews quickly became the bottleneck. When I joined Meta, it was like 3,000 engineers. By the time I left, it was tens of thousands of engineers. And I made the connection that we have all this tooling in place to handle having thousands of engineers committing code on the same monorepo, without stepping on each other. And that’s not too similar to the problems we’re seeing now with having each person on your team having multiple agents.”

The teams winning right now have figured out where humans belong

The question isn’t whether to use agents. It’s knowing where to keep humans in the loop and where to get out of the way. The teams that are moving fastest have already started making those decisions deliberately rather than defaulting to the old model.

“The ones that are better prepared for this new way of working are the ones that are just really taking a good look at where are humans still in the loop and where can agents just do that job better. I mentioned code reviews earlier, but it’s not just code reviews. It’s just do we need to wait for a human to deploy, to merge a PR? If you can set up your system where you’re not waiting on humans and you have the confidence that these agents are just doing their job, these are the companies that are just gonna be able to move way faster in this new era.”

This is a process question as much as a tooling question. Infrastructure can remove the bottleneck, but someone has to decide where the bottleneck should be.

For the first time, the question isn’t how long it will take.

There’s something Hector said near the end of our conversation that stuck with me. It wasn’t about infrastructure or architecture. It was about what it feels like to build right now and why this moment is different.

“I’m so excited. I don’t think I’ve had this much fun just working. Because now the question is not how long is it going to take. It’s just, OK, do we need to build it? And it just gets built.”

That’s the shift. Not just faster output, but a fundamentally different relationship between intention and execution. The constraint is no longer time or cycles. It’s clarity about what you actually want to build.

Final Thoughts

The through-line here is that the bottleneck in AI-assisted engineering has already moved. The code generation problem is already solved. It’s everything that surrounds the code: the environments, the previews, the reviews, the deploys, and the decisions about where human judgment still belongs.

Hector has seen this movie before, at Meta, at Parse, through every wave of infrastructure change that made distributed work tractable. Wallfacer is the bet that the next version of that infrastructure needs to exist now before the gap between what agents can produce and what teams can actually ship becomes the defining constraint of the next decade.

For founders and engineering leaders adopting AI, the tools are ready. The question is whether the workflow around them is.