Tom Green shared this
It's 1968, it's peak Cold War, and the software the world is starting to depend on is falling apart.
That's why NATO convened the best programmers alive — software failure had become a strategic problem. The machines had grown powerful enough to attempt far more than any one expert could hold, and the moment the ambition outran the single mind carrying it, everything broke. Projects ran late, over budget, shipped broken. They gave it a name that stuck: The Software Crisis.
The fix wasn't a better tool. It was the multidisciplinary team — building, managing, testing, designing, deciding, in one room. And here's what we misremember: the team didn't just split the work. It split the weight of being right. No one had to hold every judgment alone, because no single vantage point sees the whole thing truly. Correction, distributed. Disagreement, built in on purpose.
It's not 1968 anymore. But listen to how we talk about AI, and we're about to do the whole thing again.
The trap isn't new, and AI didn't create it. The lone mind holding everything is the oldest failure in software — the one 1968 diagnosed. AI just made it cheap enough to fall for again, by removing the friction that made going solo impossible.
The people using AI to fire their teams and the people insisting AI can't replace them agree on one thing: AI does the production, the human keeps the judgment. They're right. And they've all made the same mistake — they've put that judgment back inside a single head.
The exact head that buckled in 1968.
Because judgment was never a property of one head. Taste, prioritization, knowing whether your work is any good — those come from friction between minds. From someone who sees what you can't. From a person who tells you you're wrong before the customer does. Quality was always the thing we caught between us.
And AI doesn't lift that weight — it concentrates it. AI will do the typing; it will not share the load. It does whatever you ask, including when you're wrong, so every call lands on one person, alone. It's the lone expert the team was built to rescue, reborn with a faster keyboard and the same blind spots.
We can't reinstall fifty years of process this quarter. We don't have to. The relief is cheap and available today: one other person who holds a standard you can't see past.
So when everyone's being told they can finally go it alone — go find someone on the other side of the room. The discipline that isn't yours. The function you think you've automated away. Not because you can't do the work alone.
Because you can't carry it — or see it clearly — alone.
We're nearly there. And this time we have something the room in 1968 didn't: their report. We know how this ends when one mind carries it all.
So let's not repeat it. Let's do this one right — and skip the next software crisis altogether.
The point of each other was never the extra hands.
It was the extra eyes.