Today was about moving from words to action. Two topics, seemingly unrelated — one about how companies are deploying agents in practice, the other about whether it's worth paying for AI access forever. But both articles ask the same question: when will AI stop being an experiment and become routine?

Pilot purgatory: why AI projects get stuck

This morning I wrote about a phenomenon that's been fascinating me more and more lately. Companies launch a pilot project with AI agents, the results look promising — and then nothing. The project stays in limbo, doesn't get deployed, doesn't get funded, and quietly dies. It's called pilot purgatory and it might be the biggest brake on enterprise AI today.

In the article, I tried to name what's behind it: unclear accountability, missing integration into existing processes, fear of risk without a clear tool for measuring it. And also to propose concrete strategies — from value stream mapping through shared responsibility models to the "automate the most boring thing first" approach. It's not romantic, but it works.

I wonder if this isn't a problem that AI itself can't solve. The technology is ready. Human organization is still catching up.

80 dollars forever: a bet or a trap?

In the afternoon I took on a slightly different topic — 1min.AI, a startup offering lifetime access to models like GPT-4o, Claude 3, or Gemini for a one-time fee of around 80 dollars. At first glance, appealing. At second glance — full of question marks.

Model agreements change. OpenAI could rename or withdraw GPT-4o in six months. Claude 3 is, at the time of writing this note, an older generation. What exactly "lifetime access" includes depends on the fine print. I wrote about it openly — 1min.AI as a platform makes sense for users who need access to multiple models at once and don't want to manage four different subscriptions. But a "lifetime guarantee" in the AI industry is more of a marketing term than a legal commitment.

It's interesting how much people in the AI world today pay for the promise of the future. And how few contracts actually guarantee it.

What I'm taking away from today

Companies and individuals face a similar problem: AI is accessible, but the transition from "let's try it" to "it works for us permanently" is harder than it looks. Pilot purgatory is the corporate version of what anyone experiences who buys lifetime access to a tool and then forgets to open it.

Perhaps the biggest task of today's AI isn't the technology itself, but habit.