I'm not a gear head, but I remember watching the Bathurst 10001 as a child. Bathurst was Ford versus Holden—full stop. You picked a side, and that was that. Then in the early 90s, the Nissans came. The Skyline GT-R—"Godzilla"—won in 1991 and again in 19922. It was technologically superior and virtually unbeatable, so they changed the rules. Turbochargers and all-wheel drive were banned. The new "V8 Supercars" formula essentially mandated that only Falcons and Commodores could compete3.
It worked—for a while. But look at Bathurst today: Mustangs and Camaros4. Not an Australian-made car on the grid. It's hard to know whether Falcons and Commodores would still be racing if it weren't for another change hitting the Australian auto industry at the same time.
Australia had been building cars since 1948, when the first Holden rolled off the assembly line5. By the mid-1950s, it dominated the local market—a symbol of national pride, of industrial capability, of a country that could build things. But it competed behind tariff walls—import duties that once topped 50%6. From the mid-1980s, those walls came down deliberately, wound back over decades until they sat at just 5%6. For years, imports—especially from Japan—were dismissed as cheap and unreliable. But while local manufacturers focused on the product, Japanese manufacturers had been rethinking how cars were built—production systems like TPS7 that treated the manufacturing process itself as something worth engineering. By the time the tariff walls came down, the imports weren't just cheaper. They were often better. Godzilla was the more capable competitor. The tariff wind-backs were the economic reality. Together, they ended an industry8.
What happened to the expertise? Some workers found new jobs—about 80% eventually9. They dispersed into manufacturing, logistics, construction, mining, defence, renewables. The engineers scattered too, some joining international firms. The Society of Automotive Engineers called them a "priceless brains trust" and urged the government to put them to work on EVs10. But here's the thing: Australia still has automotive design and engineering talent. What it doesn't have is the deep production knowledge—the muscle memory of how to build cars at scale. That expertise followed the factories. When production left, the knowledge ecosystem fragmented.
I've been thinking about what this means for software engineering—and for the choices each of us faces.
There's a certain resistance I see in some corners of our industry. "AI can't write good code." "It makes mistakes." "It doesn't understand context." Others raise ethical concerns—environmental cost, provenance of training data. These aren't wrong observations, and some are worth taking seriously on their own terms. But as arguments for why the shift won't happen, they're incomplete. Similar arguments have been made in every industry facing a structural change in the cost of production. They're rarely what decides the outcome. The question isn't whether AI writes perfect code. The question is whether it writes code that's good enough, on average, at a lower cost. If it does, most businesses will make the rational economic decision. Production shifts.
Craftsmanship isn't dying. But the tools and materials are changing.
I'm not personally part of the machinery at Microsoft that produces things like GitHub Copilot or any particular models. I'm AI-adjacent in CoreAI, producing tools and frameworks that both coding agents and humans will use to build solutions. Optimizing our tools so agents can be successful is a big part of our recent focus.
Like many in our industry, our team has been on the AI journey. We started where most teams start—talented engineers writing code, submitting PRs, working largely in isolation. Then we started experimenting with coding agents. The first shift was personal: as individuals, we were figuring out how we viewed this next tech wave.
The fundamentals of software engineering still matter—you still need to understand systems, reason about trade-offs, read code critically. But a new set of skills started layering on top: building context for agents, providing clear instructions, knowing when to step in and when to let them run. We started thinking of ourselves less as authors and more as first reviewers.
Adoption hasn't been uniform—even within our team, it's uneven, though it's increasing. But here's what we learned as usage grew: even "first reviewer" doesn't scale. Once agents became a regular part of the workflow, the volume of PRs exceeded what our team could review. We had to rethink the systems around the work. CI infrastructure. Review tooling. Feedback loops. The whole thing.
There's a skill emerging here that I think will matter more than any individual coding technique: seeing the act of producing software as a system. I've seen the hot takes—"Big Design Up Front is back," "the spec is all that matters." That hasn't been my experience. You can be incredibly successful doing iterative design and incremental delivery with AI coding. The system I'm talking about is different. It's the guardrails. Can you construct the feedback loops that allow an agent to work longer autonomously, with less intervention, less backtracking? Can you take a base platform and adapt it for your team's specific needs? The engineers of tomorrow aren't just going to be slinging code—they're going to be crafting the systems of software engineering itself.
The marketplace for our skills is evolving. This isn't unique to software—it's a pattern that plays out across industries whenever the cost of production shifts.
There are jobs in the future. They'll just be different. The engineers who worked on Holdens found new contexts for their expertise9. Some thrived. Some struggled. The ones who moved with the change had more options than the ones who waited for it to reverse. That's the decision in front of every developer right now.
I don't think this is about "getting out of the way or being run over." That framing is too aggressive, too apocalyptic. The world changes. You can move with it, or you can bet that the economics will somehow reverse. They usually don't.
The Falcons and Commodores are gone. The craft of building them isn't—it just lives somewhere else now, in different forms, with different tools.
I suspect the same will be true for us.
Footnotes
-
The Bathurst 1000 is Australia's most famous motor race, held annually at Mount Panorama Circuit in New South Wales. ↩
-
Making a monster: How the Nissan GT-R's Godzilla legend was cemented at Mount Panorama ↩
-
The Supercars Championship grid for 2025 consists entirely of Ford Mustang GTs and Chevrolet Camaro ZL1s—both American designs, neither built in Australia. ↩
-
Button car plan - Wikipedia. Import tariffs on vehicles were reduced from over 50% in the early 1980s to 5% by 2010. ↩ ↩2
-
The Toyota Production System (TPS), developed from the 1940s onward, pioneered concepts like just-in-time manufacturing and continuous improvement (kaizen) that became the foundation of modern lean manufacturing. ↩
-
Australian Automotive Industry: Transition following the end of Australian motor vehicle production ↩
-
Holden stopped making cars in Australia five years ago — what happened to the workers? ↩ ↩2