In an interview from the nineties, Steve Jobs described the computer as a “bicycle for the mind.” It’s an apt metaphor, or it can be when computers are at their best. Computing is a long way from its best. More often than not, today’s operating systems are locked down, and the applications so reliant on internet access that people are pressured into paying a subscription for continued access to the tools and the data they work with. More and more applications have moved from working perfectly fine on your computer to all but requiring the cloud (a synonym for someone else’s computer). The cloud and the data centers around the world are what enabled the software as a service (SaaS) model that dominated the 2010s. If you follow the stock market, go on LinkedIn, or have sat in a corporate meeting at many jobs, the primary collection of ideas dominating the 20s appears to be large language models (LLMs).

Where the SaaS model says as long as you have a browser, you can simply focus on using our software, the LLM method seems to be moving towards you will have nothing, you will know nothing because you can always just ask a bot, and you will be happy. Anyone seriously working with LLMs knows that they require so much cajoling that it is often better to do it yourself, find someone suited for the task, or live with the output you were given even if you are unqualified to assess its quality.

Both of these trends in consumer computing applications are a trend in a perilous direction. I am not here to argue that SaaS companies have or have not delivered good value to businesses and maybe even individuals, but the computing arrangement necessary for software as a service to work converts computers from bicycles into automobiles for the mind. When the computer becomes the mind’s automobile, the incentive becomes to force everyone onto toll roads.

In order for a computer to be a bicycle for the mind, it largely needs to be user serviceable. From the operating system to the hardware to the software we install. There should be a certain ergonomics that someone with the time, motivation, and access to the tools and parts should be able to take a machine that’s limping along and make it roadworthy once more.

Again, that path isn’t for everyone, and computers should exist in such a way that any neighborhood has the infrastructure to maintain one’s electronics affordably and with ease. Be it a computer shop, a library meet-up group, a repair café, or all of the above where you can get help with troubleshooting, maintenance, and recommendations.

That world has existed in the not too distant past. You still hear echoes of it when you find yourself in a strip mall and see an independent operator running a repair shop or manage to hear something about Best Buy’s Geek Squad. Let’s assume for a second that we still have the communal knowledge to bring that world back. What should we ask of software?

Frankly, we should ask for a lot more, but what I am thinking about today is how software might strengthen our skills to the point where we do not need it or, in its absence, operate fine without it.

I think an application that utterly fails to enhance our cognitive abilities is maps software. GPS plus mapping software is truly incredible and has come a long way. Regardless of how I think something like Google Maps should exist as infrastructure (something that is commonly held and maintained as such), the client software we use to interact with GPS leaves much to be desired. This tool could operate in a way that enhances our ability to navigate in the world.

For example, say over the course of a month you put directions in to go somewhere seven times. The client should have built in a feature for its guidance to exponentially back off. Maybe it starts by guiding you start to finish, but the next time, it stops a quarter mile from the destination. The time after that, it goes further, and so on. Day by day your mind’s path-making is strengthened because we have developed our tools to know when to provide assistance and when to take it away.

Before LanguageTool went in on the LLM hype, I found that its advice on grammar and punctuation was helpful. My personal method for editing usually involves copying paragraph by paragraph into a tool to check for spelling and grammar mistakes and then hand-editing those because I want to train myself to be better and retain the intuitive understanding (comma usage has long plagued me).

I learned Photoshop long before generative AI (LLMs) was integrated into the product, and I learned that software hand in hand with the basics of photography on a Nikon camera. That there was similar terminology between my camera settings and the software I used to edit expanded my understanding of how my camera worked and what it meant to take a photo. Photoshop acted as a set of training wheels for understanding my camera. It was a manifestation of the computer as bicycle for the mind.

A tool that has executed this training wheels approach well is the Cura slicing software. Every setting has a tooltip that explains what it does. When I started 3D printing, I relied on Cura, an Ender 3, and a book on common printing failures. Through these, I learned a lot about printing in a short amount of time because the book had a page with pictures of different failure modes, the Cura software had all the settings labeled so that I could quickly learn the jargon of printing, which enabled my ability to search, and the Ender is such a do-it-yourself device that there is a huge community to seek out when you are at your wit’s end.

When computers are designed in an accessible way with standardized parts and easily modifiable software, it drives people into the ecosystem. People become knowledgeable about it, software can enhance books, books can enhance devices, and people can enhance themselves. People build communities around software and electronics that even offer a glimmer of this vision. These gestures, intentions, and actions make the world richer.

When you open many applications today, the first time you use them, a first-time walkthrough guides you along. Some tools have a daily tip that pops up. My experience of initial tutorials is trying to get through them quickly and onto the one thing I downloaded the software for. Daily tips are a nice affordance, but a pop-up when you open an application for the first time in a given day can come off as annoying. The best experiences are planned from the ground up and baked into the background. On some level it shouldn’t feel like we are learning nor that the tool is asking more of us and less of it.

The transition from novice to expert with any given software should be a transition so seamless that we do not notice we have gone from a stabilizing wheel on each side to one side to a person providing gentle support at your back before they let go and you are gliding alone, aided by your own center of gravity, by the two wheels of the bike, and by the power of your pedaling.