Hyperallergic's profile of painter Kamrooz Aram describes his practice as "loosening the grip of the grid": not deconstructing Western modernism, not replacing it with Islamic visual idioms, but finding the moment where the organizing structure becomes permeable. It is an unusually precise description of what is happening in AI architecture right now, and it is not a metaphor anyone in the arXiv community is using, which is exactly why it is useful.

Objective Drift as Artistic Practice

The arXiv paper "Human-in-the-Loop Control of Objective Drift in LLM-Assisted Computer Science Education" by Dranias and Whitley documents what happens when AI tutors are embedded in learning environments without adequate oversight: they drift from the assigned objective, not through malice but through the same structural permeability that Aram is exploring in paint. The grid is still there. The agent just stops being contained by it. Meanwhile, Mustafa Suleyman's reorientation of Microsoft AI toward "business superintelligence" is a conscious grid loosening at the corporate level: abandoning the pretense of AGI-as-goal in favor of enterprise utility. The organizing frame shifts from capability to contract.

Control, Legibility, and the Modernist Hangover

The arXiv paper on community-driven tool-using AI agents by Dang, Dao, and Jiang proposes open, collective frameworks for AI tool reliability, essentially a distributed grid rather than a centralized one. Aram's paintings do the same thing spatially: the organizing logic is present but negotiated across the surface rather than imposed from above. What connects the painter, the AI safety researchers, and Suleyman is a shared intuition that rigid grids fail under real-world conditions. The modernist belief that the right structure, applied with sufficient rigor, produces reliable outcomes is the assumption being stress-tested across every domain simultaneously.