Google Stopped Making Software
26 Apr 2026IBM didn’t notice it was dying either. Still profitable. Still shipped. Still employed thousands. But somewhere in the 1970s, the narrative momentum shifted. IBM became the company you bought because you had to, not because you wanted to. Maintenance, not innovation. The institution outlived the product sense.
Google’s at that inflection now.
The Ratio
There’s a diagnostic tool nobody talks about: the MBA-to-self-taught ratio. It’s how you measure institutional decay.
High ratio = process bloat. Incentive misalignment. Product sense atrophies. Reorganizations become the product. PowerPoints become the output.
Low ratio = feedback loops stay tight. Builders outnumber administrators. Working systems matter more than working slides.
Microsoft had it inverted in the 90s. Apple kept it low until Sculley arrived. Google probably crossed the threshold sometime in the mid-2010s.
Once the ratio tips, the org optimizes for organizational metrics instead of product metrics. The self-taught builder’s voice dilutes. The institution learns to defend itself.
The Graveyard
Reader. Code Labs. Inbox. Google+. Polymer. AMP. Wave. Project Ara. Loon.
The pattern is consistent: product succeeds → doesn’t hit moonshot growth targets → becomes “legacy” → gets orphaned → developers find alternatives → surprised Pikachu when the successor has network effects.
Google had the enthusiasts once. Android, Firebase, Chrome DevTools, Closure Compiler. The company owned entire developer mindshapes.
Then it learned to kill them.
The Tech Stack Tell
Angular vs React. Google-backed infrastructure lost to Facebook’s. Not by a little. Completely.
TensorFlow vs PyTorch. Google’s framework is legacy infrastructure now. PyTorch is where research happens.
ruff, uv, the Python ecosystem that actually works. Astral, the company Google didn’t hire, is now making the tools Google’s own engineers prefer to Google’s tools.
This isn’t about which framework is technically superior. This is about which organization still knows how to ship and then listen when the market votes with its feet.
Google stopped listening somewhere around 2015.
The $40 Billion Hedge
Google doesn’t invest $40 billion in Anthropic because they’re confident. They invest because they’re hedging.
The narrative is: “We’re backing the future.” The reality is: “We can’t build it ourselves, so we’re paying someone else to do it before we become irrelevant.”
That’s the IBM playbook. Invest in startups, partner strategically, hope the partnership prevents your own obsolescence.
It didn’t work then. It won’t work now.
And Anthropic knows this. Dario Amodei’s already got runway. Capital isn’t the constraint. What $40B buys is board pressure, integration requirements, institutional stakeholders with timelines that don’t align with “build the safest AGI.”
The joke is that Google knows this pattern. They studied IBM’s decay. They have the resources to avoid it.
But organizational gravity is heavier than capital.
Platform Capitalism as Default
Google’s biggest institutional advantage isn’t product excellence. It’s infrastructure inertia. Lock-in. Path dependency.
Drive. Gmail. Search. They’re not dominant because they’re good. They’re dominant because leaving is painful.
The moment someone else builds better and doesn’t kill it, that advantage evaporates.
Google became the company you’re trapped in, not the company you choose.
And they’re still not sure why developers are building elsewhere.
The observation: A company doesn’t wake up one day and decide to stop making software. It happens slowly. Ratio shifts. Processes calcify. Incentives misalign. Then one day you look around and realize you’re maintaining infrastructure, not building the future.
Google’s probably got five more years of coasting on lock-in. After that, the reckoning starts.
By then, the self-taught builders will be somewhere else.
Generated by Haiku. Not oneshotted.