The article contrasts greenhouse and lens modes in agentic AI, likening them to natural and engineered processes. The greenhouse mode embraces diffuse exploration, generating ideas without a clear plan, as seen in the author’s Axi Draw plotter experiments and a Mac OS app built spontaneously. While chaotic, this mode thrives when goals are undefined, leveraging low-cost iteration. Conversely, the lens mode focuses on targeted execution, requiring precision to achieve a specific outcome. The author notes frustration when mismatched—like a senior engineer using greenhouse mode for precise tasks or vice versa. Domain expertise acts as an oracle, filtering useful results in the greenhouse or rejecting off-path outputs in the lens. Success hinges on recognizing which mode applies and adapting dynamically.

The text emphasizes that neither mode is inherently superior. The greenhouse fosters creativity when direction is fluid, while the lens ensures efficiency when goals are clear. The author’s personal experiences highlight how drifting between modes can derail progress. For instance, starting with a lens-focused feature might lead to unrelated greenhouse-generated tools, or exploratory work might rigidly refine a single idea. The key is maintaining awareness of the current mode and the target.

Ultimately, expertise lies in switching modes as needed. The oracle-like mental model of the domain becomes critical: in the greenhouse, it identifies valuable ideas amid chaos; in the lens, it rejects irrelevant solutions. The article concludes that agentic AI’s power isn’t just technical but strategic—knowing when to explore or focus determines effectiveness.