The work is changing

My pointer fingers rested on the smooth, small bumps of the F and J keys: home position. I tuned out the whispered chatter of the other sixth graders. Computer Lab was almost over, Social Studies was imminent, and I had serious business to attend to.

I had already memorized ♥ and set the HTML’s bgcolor to hotpink. I had sent messages flying across the screen with <marquee>. I had drawn butterflies in MSPaint and animated flapping wings: my first known encounter with gifs. Today, though, I had discovered something altogether new and stunningly powerful: the script tag.

With this, I’d make my own Mad Libs game. An input would appear on the page and you’d type the answer to each prompt. At the end, you’d be presented with a flashy and utterly ridiculous story that you helped co-author.

A portal opened for me that day. I allowed myself to be transported to a new world. This technology unlocked new possibilities that weren’t available to me before.

This is an experience that we have in common. Maybe yours wasn’t rudimentary JavaScript. It’s likely that your experience wasn’t so hot pink. But as builders, we’ve shared the same feeling. We’ve both discovered a new tool, taken it apart, and put it back together again, walking away armed with new knowledge and the satisfaction of having made something. I’ll call it “the Builder’s High.” You know how runners report a feeling of euphoria after the endorphins start kicking in during a long run? Don’t you get that when you’re building, too?

Today, it feels like the Builder’s High is in peril. It’s far less satisfying to prompt than it is to step into a problem space, engineer a solution, crack your knuckles, and write every line of code from scratch. Agents are outpacing us, first in speed, now in quality. What used to take you months to write by hand can be churned out by a robot that spent two minutes “doodling.”

The changes are destabilizing, disheartening, and scary. They’re happening at the speed of light. It’s impossible to keep up with every new development. Most of us haven’t even stopped to grieve what we’ve lost.

Let’s pause, just for a moment, to survey how software engineering is changing at a real, live startup. You can get back to Jev, or whatever today’s news is, after the break.

I work at Resend. We make an email API for developers. It helps you seamlessly add transactional and marketing email to your apps. Like everyone else, we’ve been swept up in the river of change that’s come for the software industry. Software engineering at Resend is changing in three fundamental ways:

1. Everyone can ship.
2. Experimentation is more important than ever before.
3. The balance of work is shifting.

Everyone can ship

Once upon a time, the Builder’s High was reserved for those who had access to room-sized computers. Later, it required fluency in Assembly, Visual Basic, C++, C#, React. Now, it’s English.

Last week, during Resend’s weekly Demo Day meeting, I showed off a side project that I had built. Right after the demo, I got a message from our video editor saying that she had been working on something similar using base44. “It was my first attempt at building something,” she confided, sharing a screen recording of her site.

Julia isn’t the only new wave builder. At Resend, everyone has unlimited AI credits, and we’re encouraged to use them, even on personal projects. The best part is, people do! Of course, engineers take advantage of this, building sites like Forecast Vibes or Propgate. We also have less-technical folks starting to build. The customer success team is especially productive. They built dns.email, a tool to help users check their site’s DNS records. They have also spearheaded several projects to build AI tools that actually made it to production.

Neobuilders are making fun side projects, new product features, and, now, the middle ground, custom software built for an audience of one: yourself.

Custom software

The build-versus-buy debate has always been around, but the math is changing now. When the cost of software becomes the price of a few tokens, it can make sense to build ultra-custom tools to streamline your work. For example, Chris recently built a tool called re-view, which lets him share videos internally. There’s version history, comments, and more, and it all comes backed with an API and an MCP server. He can fine-tune the software to fit our exact needs, iterating over time, rather than paying a monthly subscription for something off-the-shelf that can’t be customized. Building your own custom software lets your software adapt to you, not the other way around.

Of course, there are tools that we’re happy to continue to pay for. Not all subscription software is dead. At least not yet.

Everyone can ship… so choose your “everyone” carefully

AI makes it easier than ever before to produce new content: websites, apps, text, images. You’ve probably heard the terms “slop grenade” and “meat proxy.” You can produce more, faster. That doesn’t mean that the final product is good.

It’s more important than ever to hire the right people. You need to have the discipline to focus, and you need to be able to trust your team to build the right things, too. Hiring has always been hard. To hire well in these shifting times, you should prioritize people who are curious self-starters, who have taste, and who care.

• Curious self-starters are marked by their perseverance. They’re interested in solving a problem, and they’ll use every tool at their disposal to do so. They don’t give up, and they have a penchant for tinkering.
• Taste has become a buzzword, but it’s still important. Your teammates should be able to make sound decisions about product scope, design, and accessibility. You should be able to trust that if someone on your team worked on it, it will be done well.
• Care is at the root of it all. A product is only great when its makers care about its users, the problem space, and their team. Everyone at Resend is marked by these three personality traits. These are the kinds of people we need to survive shifting tides, because these were always the marks of a successful builder.

Experimentation is more important than ever before

When OpenClaw was nascent technology, I found myself sleepless in a San Francisco hotel room at 2 a.m., messaging a robot in an EC2 rack in Ohio. My personal bot, Woz, was setting up its own email inbox with Resend. It was posting about its process on Moltbook, and writing a skill so other bots wouldn’t trip on themselves as much as Woz did. It was thrilling to tinker with this new technology. It was a Builder’s High just like Computer Lab in sixth grade.

The next day, I excitedly shared my findings with Zeno, my CEO, in the back of an Uber. He soon decided that it was time that we hired an OpenClaw bot.

At Resend, our bots are named after Greek gods. The DX team has one named Hermes. It’s been armed with some tools to help it succeed:

• An email address: the identity layer that allowed it to get accounts to tools like Slack and GitHub.
• A 1Password vault.
• A list of “trusted managers” and a healthy measure of suspicion for anyone else.
• Access to some data sources.

Hermes is helpful in some ways, and massively struggles in others. After quite some time invested in trying to help it understand our brand voice and writing style, we have altogether abandoned the idea of using it for any writing at all.

The clearest benefit of hiring an OpenClaw bot has been its impact on Agent Experience at Resend. This experiment that the Developer Experience team obsessively spearheaded taught us a lot about the holes in our product. While Resend has always been focused on building delightful developer experiences, we hadn’t been building for agents. Each time we watched Hermes stumble through our product or misunderstand our docs, stubbing its metaphorical toes at every sharp corner, we learned how to improve the product for every agent that came after.

OpenClaw was a relatively expensive technology with rough edges. It wasn’t strictly a good investment when you look at token spend. Soon enough, other tools like Hermes took over. I’ll defend our decision to hire an OpenClaw bot, though, because it taught us lessons that we wouldn’t have learned otherwise. That, and: it was fun!

Curious, adaptable people love exploring new technologies. We all got into this field because we like to tinker. Let people tinker! Being on the bleeding edge of AI, even while “playing,” is productive. If you’re AI native in play, you’ll learn. Ideas will naturally flow, and you’ll soon be using cutting-edge features and concepts in production.

The balance of work is shifting.

Pre-AI, companies couldn’t hire enough engineers.

Now, engineers are empowered to output code at much faster speeds. A good engineer is still extremely valuable, and replacing your engineers with AI is ill-advised. However, speeding up one area—code output—doesn’t necessarily speed up others. Organizations must adjust to regain their balance.

Good design takes time. If you want to stand out from the sites filled with purple gradients, Instrument Serif, and eyebrow text, you need real humans with taste. Good copywriting takes time. The Resend website could say: “it’s not just an email API—it’s an experience. And that’s what matters.” But it doesn’t. That’s because those of us who decide what goes on the website care.

Good events, product strategy, go-to-market, and recruiting all take time, and even though many of these areas are being sped up with AI, nothing approaches the pace of innovation in AI-assisted engineering.

These shifts can be destabilizing. They’ve thrown companies out of balance, causing overhiring, layoffs blamed on AI, and more. To regain balance:

• Hire proportionally.
• Focus relentlessly.
• Resist the urge to rush work that should be slow.

Hire proportionally

At Resend, we have an incredibly strong brand and high standards for how it’s represented. As engineering output grew, we noticed that the Design team was becoming a bottleneck. To widen the bottleneck in this situation, you can choose between two options: lower your standards, or grow your team.

Lowering our standards wasn’t a serious option for us, so instead, we invested in growing our Design team, hiring several Design Engineers, a motion graphics expert, and a Brand Designer. With this expanded team, we’re able to meet the needs of the other departments without slowing them down.

Your team’s needs may vary. Maybe it’s QA or Infra or Marketing that’s slowing you down, instead. Assess the areas where your developers are experiencing the most friction and staff teams accordingly, so they can be a help to each other rather than a drain.

Focus relentlessly

Focus has always been notoriously difficult for companies. It’s becoming harder now that code output is faster.

Every launch cycle, Resend’s leadership team shares a list of features we will develop next, as well as a list of what was cut. Each time we decide not to build something—or, at least not yet—it’s torture. It’s also a crucial discipline.

What’s actually valuable to your users? Do you even know? No, I’m not talking about the laundry list of feature requests that they ask for on a daily basis. I mean: what can you build into your product that will solve their biggest problem in your space? Identify it, plan for it, execute, and repeat until no problems remain.

Resist the urge to get caught up in things that don’t matter: copying your competitors, accomplishing vanity metrics, or trying to do everything at once. The only way to do things well is to focus.

Resist the urge to rush work that should be slow

There’s work that can and should be automated. We’re AI-native at Resend, and love the productivity gains we’ve earned. There’s also work that needs to stay slow.

We take time to think through product direction. We can build anything. It’s important to decide what to build next, and stick to that.

We take time to create meaningful connection with our users. We have many avenues for this: hosting thoughtful events; listening to customers on social media; producing timely videos, tutorials, and blog posts; a caring and responsive Customer Success team; and more.

We take time on platform stability and quality so that our users can trust us with their most sensitive email needs.

We take time simplifying product design so that APIs are predictable, friction is removed, and every experience feels like a fresh breath of air.

We take time to connect with each other in person at least twice a year.

At AI speeds, these things can feel like a waste of time. There are ways that you can invest less time into these areas and get an output that’s okay. The thing is, we aren’t okay with okay. We slow down when it matters most to ensure that we’re putting our best thought into the task at hand.

Long live the Builder’s High

Engineering is changing, and no matter how much we speculate, none of us can really claim to know how things will turn out. The best we can do in these shifting times is exactly what we’ve always done: show up with our whole selves, and tinker until we figure it out.

You remember the Builder’s High: that feeling of euphoria from practicing a new skill and creating something new. It’s not beyond your grasp, even now. In fact, you have all kinds of new tools and avenues to explore. It’s uncharted territory.

The best part is: we don’t have to go alone. As engineers, we’re part of a curious, intelligent, and deeply generous community. We’re all trying to figure out where this is going. Let’s explore this brave new world together. Let’s shape it together.

There’s more building yet to do. Join me—and the rest of the community—in building into the future.