The "Dial-Up" Phase of AI

For someone who has spent the better part of two decades working with cloud, scale, and effectively unlimited compute, the current wave of agents and LLMs feels strangely familiar.

Not in a good way.

It feels like dial-up.

We are back to clunky interfaces, opaque outcomes, and unpredictable cost curves. Most jarring of all is that constant, nagging sense that you should ration your thinking before you hit enter.

We have spent 20 years training ourselves (and our organisations) out of scarcity mindsets.

Storage became abundant. Compute became elastic. Experimentation became cheap. You could explore without watching the meter tick in real time.

Yet here we are again.

We are carefully crafting prompts and second-guessing whether a task is "worth it". We are hesitating before running deeper analysis because of the perceived cost: whether that is measured in money, time, or tokens.

It is BBS-era behaviour in a supposedly post-scarcity paradigm.

The Transitional Gap

This is not the end state. We are currently stuck in a transitional phase where:

- UX hasn't caught up with capability: We are interacting with raw power through primitive interfaces.

- Pricing models are misaligned: The commercial wrappers do not yet match user psychology.

- Outcome determinism is immature: Reliability remains the missing link.

History gives us a clear precedent. Early internet services that leaned into scarcity and walled gardens did not win. They were overtaken by models that embraced openness, predictability, and scale.

The same pattern will play out here.

Designing for the Non-Dialup World

The risk for founders and product leaders is mistaking today’s constraints for tomorrow’s business model. If you build your product assuming that users will tolerate opaque pricing, unpredictable workflows, or high cognitive load, you are effectively designing the AOL of the agent era.

We all know how that story ends when broadband arrives.

Right now feels messy and oddly nostalgic, but it is temporary. The real opportunity is not to optimise for the current friction. It is to anticipate its removal.