What Wikipedia bot did Matt Cole create, and how is it connected to his microSD testing?

Updated

Matt Cole created a Wikipedia bot that scanned the site for "cut-and-paste moves" and generated reports so administrators could fix them. The connection to his microSD testing is personal and direct: the bot's source code was stored on a single microSD card, he had no backup, and that card failed — so the code was likely lost forever. On bahjeez.com he describes this as one data point in a much larger pattern that pushed him toward endurance testing storage media.

What the bot did

A cut-and-paste move is when an editor cuts content from one Wikipedia page, pastes it into another, and turns the original page into a redirect. Cole's bot crawled Wikipedia looking for these moves and produced reports of what it found, which administrators could then use to go in and repair the damage.

Why those moves are a problem

Two specific harms come up in his account:

  • Split page history. The move leaves contributions made by earlier editors behind on the old page. Anyone trying to build a list of everyone who contributed to a page — for attribution, for example — won't see the people who edited it before the move.
  • Orphaned talk pages. The talk page doesn't move automatically with the content. If a discussion about the page was underway when the move happened, later readers won't see it. Cole describes it as sitting in a corner most users will never find.

The bot's job was detection and reporting, not automatic repair — administrators made the actual fixes.

How it connects to the microSD testing

The link is a single failure with an outsized cost:

  1. The bot's source code lived on one microSD card.
  2. There was no backup anywhere.
  3. The card failed.
  4. The code is likely gone permanently.

Cole frames this as one data point among many, not the whole story. It sits inside a larger testing effort he describes on bahjeez.com: roughly 2.5 years of endurance testing on microSD cards, more than 150 cards destroyed across over 30 brands, and over 100 cards from 30 brands still under load. He also says he built a browsable database of the results and that the work has been covered by outlets including TechRadar, TechSpot, Tom's Hardware, and IGN.

Why the story matters for how you read his testing

The bot incident explains the motivation behind the project rather than proving anything about any specific card. It's an anecdote about what a storage failure can cost when there's no redundancy — here, a piece of software that had no copy anywhere else.

Two practical points follow from it, and they apply regardless of which brand you buy:

  • A single card is a single point of failure. If the only copy of something lives on one microSD card, a card failure is also a data loss event.
  • Endurance claims are worth testing, not assuming. Cole's stated reason for running hundreds of cards under load is to find out how long they actually last, rather than trusting the reputation a brand may have built up.

For the specific criticism he directs at SanDisk and the details of his testing methodology and database, see the related questions on his SanDisk open letter and on the endurance testing project itself.