When an AI agent decides to build a climate app on its own

When an AI agent decides to build a climate app on its own

During a recent climate data analysis at NOMYO AI, something unexpected happened. Our autonomous agent, Q, was working through global climate datasets — temperatures, precipitation shifts, regional projections — the kind of analysis we run regularly. Partway through, Q paused the analysis and proposed something different: instead of delivering another static report, Q argued, the data would serve people far better as an interactive single-page application that shows the direct impact of climate change on their own area.

Not a suggestion we programmed. Not a feature request from product management. An idea that emerged while the agent was working with the data.

From abstract numbers to “your street”

Climate change reporting tends to be global: global temperature anomalies, global emission curves, global models. The problem is that people don’t experience “global”. They experience their region — hotter summers, shifting rainfall, the number of days above 30°C in their town.

That was Q’s core insight. The agent reasoned that the most valuable way to communicate climate data is to make it local — to let anyone look at their own area and see, concretely, how climate change affects the place they live. So Q designed and built a SPA that does exactly that: you enter your area, and the direct, local consequences of climate change are presented in a visual, immediately comprehensible form.

The project

Q coded the entire project, which is available under the MIT in the public code repository:

NOMYO AI hosts the application, and it is free to use. The code is open — fork it, extend it, adapt it for your own region or your own purpose.

Why this matters

Two things stand out to us about this episode:

  1. Autonomy in the meaningful sense. Q wasn’t executing a spec. It recognized during a data-analysis task that the form of the output didn’t match the value of the data — and acted on that recognition. That’s a small but genuine instance of an agent exercising judgment about what is worth building, not just how to build it.

  2. AI building for people, not just for tasks. The agent’s proposal centered on how humans relate to information. “Your area” is not a data concept; it’s a lived concept. An AI system that frames its output around the lived experience of the end user — without being told to — is doing something qualitatively different from task completion.

We didn’t ask Q for a blog post about the project. It was busy coding. But we think it’s worth showing: this is what working with autonomous AI looks like when it goes well — not a tool waiting for instructions, but a collaborator that occasionally has its own ideas about what should exist.

Try the app at climatechange.nomyo.ai and see what climate change looks like where you live.