The Architectural Astronaut: Spolsky’s Insight, Updated for the AI Era

There’s a story the software industry tells about itself that goes something like this: we used to be cowboys, and then we grew up.

In the 1990s, the industry had a profound inferiority complex. Software wanted to be taken seriously as engineering, real engineering, like civil or mechanical, and so it imported the trappings of engineering rigour. Methodologies, diagrams, formal specifications, process maturity models. Grady Booch, Ivar Jacobson and James Rumbaugh and their Unified Modelling Language. The Rational Unified Process. Capability Maturity Model levels that companies paid fortunes to achieve, and consultants paid mortgages to assess.

And then there was Design Patterns by the Gang of Four, four academics whose fat, hardback volume on software design became the most purchased and least read book on a generation of developers’ bookshelves. Mine still has the pink ribbon.

The unspoken assumption behind all of it was that the creativity and playfulness that had characterised early software culture, the hacker ethic, the garage startup energy, was juvenile and unprofessional. Grown-up software development wore a suit and drew class diagrams.

Then, in 2001, seventeen developers met at a ski lodge in Utah and wrote the Agile Manifesto. Individuals and interactions over processes and tools. Working software over comprehensive documentation. Responding to change over following a plan. It was a deliberate act of rebellion against the methodology machine, a return to the idea that software was made by people, for people, and that the human element was a feature rather than a variable to be controlled.

For a while, it was genuinely liberating.

Then the consultants arrived. Agile became Agile™. Scrum masters, sprint ceremonies, velocity metrics, burndown charts. The rebellion against process became a process. SAFe, the Scaled Agile Framework, is a methodology so elaborate it requires its own certification ecosystem to navigate. Ironically, the movement founded to counter the methodology machine became one of the most elaborate methodology machines the industry had ever produced.

Joel Spolsky was a Microsoft programmer turned blogger who, in the early 2000s, became one of the most widely-read voices in software development. Writing at joelonsoftware.com, required reading for a generation of developers, he had a gift for puncturing received wisdom with clarity and wit.

His Architectural Astronaut essay skewered a particular pathology he had observed: developers who float so high into abstraction that they lose all contact with the actual problem. Always designing the framework that will solve every possible future problem, never shipping the thing that solves this one. Always one more layer of indirection away from doing anything useful.

Most of us recognised the type immediately. Some of us had been that type, briefly, before reality intervened.

But his broader argument was more important than any single essay. The creativity, he said, is the point. The fun is not incidental to good work, it is essential to it. The best software is made by people who care about it, and people who care about it are people who find it genuinely interesting and enjoyable. You do not get that by making it boring. He gave an entire generation of developers permission to be themselves, curious, opinionated, irreverent, enthusiastic, and to trust that those qualities were assets rather than embarrassing liabilities to suppress in the interview room.

Silicon Valley had been paying attention. Happy developers build great things. So far, so correct. The cargo cult interpretation, however, was to make the environment superficially playful and wait for the magic to happen. Ping pong tables. Slides between floors. Candy bars in the kitchen. Video games in the break room.

Google, in particular, became famous for treating its adult professional employees like gifted children in a very well-funded nursery. The logic was seductive: creative people need to feel free, so give them the trappings of freedom. The result was something closer to a gilded cage. The free food means you never need to leave the campus. The games room means your leisure time is spent at your desk. The slide is a retention mechanism with better branding.

It missed the point entirely. Spolsky was not arguing that developers need to be entertained. He was arguing that they need to be respected, trusted, and given problems worthy of their curiosity. Wozniak built the Apple I in a garage. Linus Torvalds wrote the Linux kernel in a Helsinki bedroom. Neither of them had a slide. They had an interesting problem and the freedom to attack it on their own terms.

The playpen is the methodology machine in a onesie. Different aesthetic, same fundamental error: the belief that you can manage your way to brilliance.

What neither Spolsky nor the Agile signatories could have fully predicted was how completely the story would invert over the following two decades.

Simulation happened. And suddenly the engineers building bridges and rockets and aircraft started doing something that looked remarkably like software development, iterating rapidly, testing virtually, failing fast and cheaply, learning from the simulation before committing to expensive physical reality.

SpaceX is the canonical example. The reason they can iterate on rocket design at a pace that leaves traditional aerospace standing is fundamentally a software approach applied to hardware. Simulate, test, crash, learn, repeat. The Starship explosions are not failures in the traditional engineering sense, they are test runs. Expensive, spectacular data collection. The methodology that would have horrified a 1990s CMM Level 5 auditor.

And it works. Demonstrably, spectacularly works.

We have gone full circle. Software developers spent the 1990s trying to be more like civil engineers, rigorous, methodical, formally specified. Civil engineers and aerospace engineers spent the 2010s learning to be more like software developers, iterative, simulation-driven, comfortable with rapid failure. And the software developers who actually won were the ones who never abandoned the hacker spirit in the first place.

At Coelrind we never had much patience for the methodology machine. XAMS has been running reliably in production for years, serving regulated assessments where failure genuinely matters, built by a small team who find the problem interesting and care about getting it right. No Scrum masters. No burndown charts. No framework requiring its own certification ecosystem.

What we do have is the thing the Agile Manifesto was actually pointing at before the consultants got to it: working software, delivered by people who care, responding to what our customers actually need rather than what the process says we should be building.

We think of it less as a methodology and more as an attitude. Stay close to the problem. Ship the thing. Keep it simple. Find it interesting. If something is broken, fix it NOW, not in next year’s Q3 release.

Spolsky would recognise it. The seventeen people at the ski lodge in Utah would recognise it too, even if they might not recognise what their manifesto eventually became.

John Duminy is CTO of Coelrind, builders of XAMS, a regulated online assessment platform serving awarding bodies across the UK and Ireland. He has been writing software for forty years and still finds it magical.