Google Gemini broke my new bike. Or: How Big Tech has abandoned trust.

When you turn on the kitchen faucet, you trust that water will come out. When you type on a keyboard, you trust that letters will appear on your screen. When you use a search engine, you trust it to deliver meaningful results.

Lately, Google has been trying to integrate its Gemini AI into everything. The problem with that is printed plainly at the bottom of every chat interface and AI-powered search result: "Gemini is AI and can make mistakes."

And yes, I’ve pretty much come full circle on these tools. This is the story of how that little disclaimer translated into a broken brand-new bicycle.

The Cheat Code for DIYers

I've always preferred doing things myself. Learning a skill is usually better than being beholden to contractors, mechanics, or third-party service providers.

To give large language models credit, they can be decent troubleshooting tools. I previously used a massive, year-long chat thread to successfully (?) diagnose a persistent foundation leak in my kitchen—a problem that professionals completely missed after charging ten grand. In another instance, I used an AI tool to run command-line recovery options on a corrupted hard drive containing raw footage for an indie feature film. Even though the data itself was unrecoverable due to an incomplete original transfer, the AI-assisted troubleshooting helped eliminate variables and land on an accurate diagnosis.

For a long time, troubleshooting seemed to be one of the very few things AI was moderately decent at. Whether dealing with technology, home appliances, or bicycles, it functioned like a cheat code for DIYers.

Right up until that little disclaimer kicks you in the nuts.

Draining the Brake Lines

I recently bought a new e-bike. It was the first bicycle I've owned as an adult that actually fits me. I assembled it myself, only to discover that the front wheel had a noticeable wobble from shipping or sloppy quality control.

Truing a wheel involves adjusting the spokes until it spins straight. I wanted to handle my own maintenance, so I worked at it until the wheel was about 90 percent true (which is pretty damn good, considering I had no idea what I was doing!). However, the brake rotor was still audibly rubbing against the pads.

I turned to Gemini for help. Despite explicitly providing the exact make, model, and version of the bicycle at the very beginning of the chat, the AI very confidently told me to adjust the hydraulic brakes as if they were mechanical brakes. I had a hunch something was off, but given how reliable these tools had been for previous troubleshooting—and that I’m usually decent at getting myself out of trouble—I followed the steps anyway.

As a direct result, I accidentally drained all the mineral oil out of my brake line.

When I double-checked the manufacturer's website, the product page clearly stated the bike used hydraulic brakes. When I pointed this out to Gemini, it offered the standard, automated pivot:

"I completely apologize for the bad steer on the brake model—I was looking at the specs for an older iteration of the line, and that is entirely on me."

The Erosion of Foundational Trust

This isn't just an issue of factual accuracy; it is a fundamental problem of trust. In product development, a user needs to trust the product and the company delivering it. If that trust is gone, there is no reason to use the product.

Big Tech is forcing tools down our throats that are, by design, fundamentally untrustworthy. Through the transitive property, if you integrate an untrustworthy tool into every layer of digital infrastructure, the entire ecosystem becomes untrustworthy.

I have never been a total AI booster or a total AI hater. What I am is a technology enthusiast. And when technology and creativity cross paths, I am completely in my element. But watching big tech abandon a foundational value like basic trust makes it incredibly difficult to stay enthusiastic about the future of technology.

Maybe this is just more of a reason to brush up on tangible, real-world DIY skills. Just remember to go straight to the manufacturer's source for your troubleshooting, because Gemini can—and will—make mistakes.

###