In 2008, I talked about how important it is for websites to work correctly in text browsers (Lynx at the time). And guess what? We’ve pretty much landed back there, thanks to AI!
Act one
The first time I stumbled across the rise of the command line was with the advent of AI agents. When doing their job, they try to work as efficiently as possible and therefore prefer easy to navigate websites (short click paths) and fast, lightweight pages without technical obstacles. Accessibility guidelines such as WCAG are a good guide for optimization in a text first world.
Act two
The next time I remembered what I said, was when I was researching one-click shopping within Perplexity. The oh-so-greedy crawler wasn’t messing with the shopping sites at all, but their developers were negotiating a technical, text-based interface to the online store that bypasses what we as humans experience: Away with the whole user interface, away with all the Lametta.
Final Act
And finally, Andreas’ (as always very intelligent) article on the latest developments in the vibe code scene, where Claude Code is making quite a splash. Notabene a text-based extension of the Command Line. The tool not even offers a graphical user interface, but performs it magic in a way, skilled UNIX administrator do it.
This again bases on key UNIX design principles for the CLI, which Wikipedia credits to Malcolm Douglas McIlroy in 1978 and that Peter H. Salus picked up in 1994 and formulated it like this in A Quarter Century of UNIX:
- Write programs that do one thing and do it well.
- Write programs to work together.
- Write programs to handle text streams, because that is a universal interface.
Conclusion
Large Language Models are based on text, so let’s skip anything graphical.
If it’s true that AI will take over “the work”, agents are our new persona and they don’t need bling.
Brutal statements – but I quite like this world 😉
PS: Why is everybody so eagerly writing MD nowadays?
