When Sessions Still Counted

Or: Why the best working habits are often born from scarcity, and what a modem cable has to do with it.

If you’re old enough, you remember the sound. That screeching and hissing when the modem dialed into the internet. And the choreography that preceded it: hang up the phone, pull the cable from the phone jack, plug in the modem cable. If you were online, you were unreachable. If someone wanted to make a call, you had to go offline.

This constraint shaped an entire culture. You planned your online time. You downloaded pages to read them offline. You wrote emails in advance, before dialing in, and sent them in batches. You developed a fine sense of what was worth the expensive connection time and what wasn’t.

And then DSL arrived. The constraint vanished. Always online, the phone still rang. The whole choreography was obsolete overnight.

But something remained.

I’m currently building a typing trainer. With AI, over several weeks, on a Pro subscription with an AI provider. And the Pro tier has a peculiarity that shapes my entire working behavior: sessions are limited. There’s a budget, and when it’s used up, I have to pause, sometimes for hours.

The tricky part: planning and implementation share the same budget. When I discuss the concept in one AI interface and write the code in another, both draw from the same pool. Every minute of thinking comes at the cost of time I have for building.

That sounds like a nuisance. And it is. But something curious happened: this scarcity taught me things I would never have learned under abundance.

Example one. Because every coding session is expensive, I started front-loading the thinking. First, calmly clarify the concept, which data structure, which architecture, which pitfalls, and then hand the coding AI a fully thought-through brief. Not “let’s see how it goes,” but “do exactly this, and stop at these three points.”

The result wasn’t just cheaper. It was better. The coding AI makes fewer mistakes when it receives a clear brief instead of a vague idea. The scarcity forced me into a discipline that turned out to be simply good practice.

Example two. Because I often have to pause between sessions, I needed a place where loose thoughts could wait until I was back. A kind of buffer: what’s open, what was the last decision, where did we leave off. Out of necessity, a system emerged that I wouldn’t want to do without today, not because sessions are limited, but because it simply helps me keep the thread.

Example three. Because tokens are precious, I learned to choose between two working modes: the expensive, thorough one, where the AI presents a full plan first, and the lean one, where it starts immediately but pauses at defined checkpoints. Before, I would always have chosen the thorough mode, because “more planning is better” sounds right. The scarcity forced me to distinguish: when is the expensive mode actually worth it? And the answer was “not nearly as often” as I would have expected.

Here’s the point I can’t let go of: none of these three habits is a workaround.

A workaround is an ugly stopgap you throw away as soon as the problem disappears. Unplugging the modem cable was a workaround. Nobody misses it.

But front-loading the thinking, the buffer system, choosing the right working mode, those aren’t stopgaps. Those are practices that remain sensible even when sessions become unlimited someday. And they will become unlimited, I have no doubt about that. Today’s token scarcity will be an anecdote in a few years, just like the modem sound.

What will remain are the habits that scarcity produced. Just like after the modem era: the choreography vanished, but the instinct for what deserves attention stayed with many people. Those who learned to prioritize during scarce times prioritize better in abundance, too.

Economists know this phenomenon. Some of the most elegant solutions emerge under hard constraints, not despite them. Scarcity is a strict but good teacher. It forces decisions that abundance obscures.

There’s a term I’d like to propose for what’s happening here: constraint-induced method development. The constraint doesn’t just produce adaptation, it produces method. And the method outlives the constraint.

This isn’t a new insight. Poetry has the sonnet because fourteen lines and a fixed rhyme scheme unlock more creativity than the blank page. Photography had 36 frames per roll, and some say that made people better photographers than the infinite memory card ever could. The constraint is not the enemy of quality. Often, it is its condition.

And so here I sit, in the spring of 2026, building a typing trainer under conditions that no one will be able to comprehend in a few years. “You had to pause? In the middle of work? Because you ran out of tokens?”, that’s what people will ask, with the same incredulous amazement with which a teenager today reacts to the modem sound.

Without planning it, I’ve become a witness to an era. A witness to the early months of a way of working that is only just emerging. It doesn’t feel historic while it’s happening. It feels like everyday life, like token budgets and session pauses and the eternal question of whether the thorough mode is worth it.

But it never does, while it’s happening. The man who plugged in his modem cable in 1996 didn’t feel historic either. He just wanted to check his email before the connection collapsed again.

The sound is gone. The habits it produced have stayed. And I suspect it will be the same with session budgets.