A USB drive that stores only 64 bits is enough to make most people roll their eyes. Polymatt, a tech YouTuber who builds and documents tinkering projects, calls it the “world’s worst USB device,” and by capacity alone, he is not wrong. The twist is that the 64 bits are not stored in flash. They live in magnetic-core memory, a mid-century storage technology that predates silicon-based norms by decades.

In Polymatt’s design, 64 rings of ferrimagnetic material correspond to 64 bits of information. The device uses tiny rings and multiple woven wires, with a magnetic field generated by currents through the wiring. Flipping the ring’s magnetic polarity encodes binary data, with polarity direction representing 0 or 1. Because the polarity does not require continuous power to remain set, magnetic-core memory is non-volatile, similar in that respect to NAND flash. But where modern drives win instantly on density, Polymatt’s build tops out at the humble scale of 64 bits, a constraint the video makes very clear.

Why obsess over something that is millions of times smaller than anything you can buy as an external SSD right now? The answer is less about replacing storage and more about understanding a lost design era. Magnetic-core memory was a real workhorse during the 1950s and 60s, when engineers did not have the same silicon options we treat as obvious today. The source points to recognizable historical anchors: ENIAC, IBM 704, and the Apollo Guidance Computer used magnetic-core memory. That matters because it frames the project as an engineering time capsule. You are not just watching a “gadget.” You are seeing how early computing systems stored state with physics, not cell arrays.

At the technical level, magnetic-core memory works with those “tiny rings” of ferrimagnetic materials and the woven wires that interact with them. A magnetic field alters each ring’s polarity. In modern storage language, it is a different mechanism for holding bits, but the practical outcome is familiar: you can store binary information without power after writing. Polymatt’s build also uses an Espressif ESP32 microcontroller to handle the USB interface and manage overall read and write operations. Here is the delicious irony. The ESP32 includes embedded flash storage that is millions of times larger than the magnetic-core array, yet the core of the “point” is the magnetic-core part, not the convenient onboard silicon.

If you want a gut check on how far density has come, the source gives a comparison: IBM’s 1957 magnetic-core unit had a capacity of 147,456 bits. That is 147,456 bits, not 64. But it came with crushing practical costs in size and money. The IBM unit weighed several hundred kilograms and apparently cost $6,000 per month to rent. That contrast is the real story. Even at its best, magnetic-core memory was an industrial-scale, hardware-heavy solution. Polymatt’s approach shrinks the concept to a DIY scale, even if it cannot escape the fundamental density ceiling.

The project’s own math makes the same point in a way anyone can feel: 64 rings equals 64 bits. Then the source jokes that it would take about 16 million times more to reach 1 GB at the same “ring per bit” approach. That is not a casual estimate, it is a density reality check. Modern flash and SSDs packed with silicon can squeeze vastly more bits into each square millimeter. So comparing the magnetic-core device to an external SSD is not just unfair, it is meaningless. The source explicitly says there is no real point in comparing it that way, and the build does not pretend otherwise.

Still, the timing of the project is not random. The source opens with the idea that AI data centers are swallowing up memory storage capacity everywhere. That pressure pushes engineers and product teams toward faster and higher-density solutions, mostly driven by silicon. But when storage gets scarce or expensive, the industry naturally looks for alternatives. In practice, “alternatives” usually mean better compression, different architectures, smarter caching, or new NAND and controller designs, not a ring-and-wires revival. Polymatt’s device sits in a different lane: it is a demonstration of non-volatile storage principles and an artifact of engineering elegance, even while it remains firmly impractical.

For executives, the second-order implication is the same one that always shows up when a technology runs into physics. Storage strategy is not only about what is available today, it is also about what tradeoffs your customers are willing to tolerate tomorrow. The source even nods at the aesthetic and “vibe” of silicon versus magnetic-core hardware, but the deeper business takeaway is that capacity, speed, and cost are not the only axes. Reliability, power behavior, system complexity, and the ability to adapt to changing constraints matter too. Polymatt’s “world’s worst” USB drive is not a product threat to SSD vendors. It is a reminder that the playbook has changed, and that constraints have consequences. The world may be moving toward ever-denser silicon, but the moment someone hits a ceiling, the industry will go digging for new ways to hold state, even if the inspiration starts with 64 bits and a pile of tiny ceramic rings.