Adopt Autonomous Coding Now, Or Find a New VP of Engineering

A month ago, Dan and I weren’t giving this advice. The technology wasn’t quite there. This month, it is — and now we’re telling every single client the same thing, regardless of size: adopt autonomous coding immediately.

We consult with a wide range of organizations. One-person startups. Fortune 500s. Everything in between. As of right now, the recommendation doesn’t change with headcount. There’s no org too small or too large to do this. There are only orgs that are behind.

Here’s the uncomfortable version of it: if you’re a CEO paying your developers to write code, you’re wasting money. Not your money — your investors’ money. That’s a hard thing to say out loud, and we wouldn’t have said it a month ago. But the tools crossed a threshold. They’re producing real, high-quality, relevant code. It’s not AI slop anymore.

Why This Is a CEO’s Call

You’ll notice we keep saying “CEO.” That’s deliberate.

Developers are not going to walk into a room and propose adopting the thing that takes away what they believe they’re paid to do. That’s human nature, not a character flaw. A lot of developers think they’re getting paid to write code and stare at a terminal. As of July 2026, that’s over — and asking someone to volunteer for that realization is asking a lot.

So the push has to come from the top. Ask whoever runs your engineering org a simple question: “Have you looked into Devin? Have you looked into Shipwright? And if not — why have you chosen not to 10X your team’s performance, for free?”

Because that’s the actual decision on the table. Shipwright is free and open-source, running autonomous coding in your own infrastructure, as long as you’re already on Claude Code. Devin is the commercial option — a $25 billion valuation, raising a billion dollars. Both produce pull requests without a human writing the code. Pick one and start. But start.

And if your VP of engineering isn’t the one pounding on your door about this? That’s the flag. Not the tool — the leadership.

Developers Get Paid to Plan, Not to Type

Let’s be clear about what doesn’t change, because this gets lost in the panic.

Developers were never really paid to write code. They were paid to plan — to make judgment calls, to decide what to build and why, to review what ships and catch what’s wrong. That part doesn’t go away. It gets more valuable. The typing was always the least important part of the job; we just couldn’t automate it until now.

If you’re an engineer reading this with a knot in your stomach: the skill that made you valuable isn’t disappearing. The keyboard part is. Lean into the judgment.

The Part Nobody Warns You About: It’s a Human Problem

Here’s where most teams get it wrong. They think adoption means installing a tool. It doesn’t.

We’ve rolled out Claude Code across entire organizations and watched some engineers use zero tokens while others burned through millions a week. Same tool, same access, wildly different outcomes. The gap was never technical. It was behavioral.

So installing Devin or Shipwright and wiring it into GitHub isn’t adoption. You have to actually push your developers to change how they work — and depending on your org, that push is significant. The nimble five-person startups adopt out of the box. But if only 10% of your engineers are excited about this, you’re going to need real organizational force behind it. And it starts at the top: CEO, to planning, to engineers. Everyone has to change how they think.

You’ll also probably need to flatten your org. Top-heavy, management-heavy structures don’t survive autonomous velocity — empowered developers moving several times faster don’t fit a slow, multi-layer approval chain. That’s another change, and it’s a big one.

The Economics

The numbers are almost unfair. Autonomous coding runs roughly $10–20 per pull request in tokens — call it $10–20 per feature, once you’ve optimized the flow. On Shipwright, that’s your whole cost; it’s open-source and runs on your own infrastructure. On Devin, it’s tokens plus their seat cost.

Compare that to a fully-loaded engineering day and it stops being a cost conversation. Ten dollars of tokens versus a developer-day isn’t a debate — it’s a rounding error.

The real question was never “can we afford to adopt this.” It’s “can we afford to let a competitor adopt it first.” Because the bigger the change, the longer humans take to absorb it — and this is a massive change. Telling developers they’ll plan and review instead of write is going to take your org months to internalize. Which is exactly why you treat it as a level-one priority now, not in six months. If your competitors move first, you don’t catch up.

Where to Start

You don’t need a perfect plan. You need to pick one tool and begin, because the adoption curve is measured in months and the clock is already running.

If you’re a Claude Code shop, our answer is always Shipwright — it’s free, open-source, and it’s an extension of what you’ve already built, not a migration. If you’d rather have a fully managed, enterprise-checkboxed solution, Devin is a real option.

Either way: choose one, master it, push your team, and give it three months. That’s how you get to 10X.

Ship it.

This post is adapted from The Velocity Lab — Dave O’Dell and Dan McAulay on shipping faster with AI. If you want help adopting autonomous coding, book time with us.