May 22, 2026
Brains, budgets, and broken images
The Hardware Lottery
When good ideas lose because the machines picked favorites and the comments got spicy
TLDR: The essay says many “winning” ideas may have succeeded only because they matched the machines available at the time, not because they were truly best. Commenters swung from jokes about broken images to serious frustration that budgets and hardware still steer research today, which matters because it means progress may be more biased than people think.
The big idea in “The Hardware Lottery” is deliciously unsettling: sometimes a research idea doesn’t win because it’s the best idea — it wins because it happens to fit the computers people already have. In plain English, the essay argues that history may be full of smart ideas that got ignored simply because the available machines and tools weren’t ready for them. That’s a pretty dramatic accusation, and the community response instantly split into equal parts “this is deeply true” and “of course the tools decide everything.”
But honestly? The comment section brought the real plot twists. One user dryly announced, “All images are broken,” which is the most internet possible way to react to a deep essay about how technology shapes knowledge: not with philosophy, but with a busted webpage. Another commenter stole the show with an old Charles Babbage quote basically roasting anyone who thinks feeding bad numbers into a machine could somehow produce good answers. It landed like a Victorian-era mic drop and gave the whole thread a meme-worthy, nothing-ever-changes energy.
Then came the practical hot take: one reader argued this problem isn’t just ancient history or elite theory — it happens in labs right now, where budgets and available gear decide what gets studied. Their conclusion? Open-source hardware matters because otherwise research follows the money and the machines, not necessarily the truth. So yes, the essay is about computer history — but the crowd turned it into a juicier debate about power, gatekeeping, and whether progress is being quietly rigged by whatever hardware happened to be on sale.
Key Points
- •The article introduces the term “hardware lottery” to describe cases where research ideas succeed because they align with available hardware and software rather than because they are inherently superior.
- •It argues that historical separation between hardware, systems, and algorithms research has obscured how tooling influences which ideas are seen as successful.
- •The essay identifies changing hardware economics, larger deep learning architectures, and edge-device deployment as major reasons for renewed collaboration across hardware, software, and machine learning.
- •It says current collaboration is focused heavily on domain-specific hardware optimized for deep neural networks, which improves efficiency but may discourage alternative research directions.
- •The article uses historical examples such as the Difference Machine, Mark I, perceptron machine, and Jacquard loom to show that early computing systems were often specialized and difficult to repurpose.