JHDD Typography Report — 2026.07.27

Connie He’s work on ‘Dear Upstairs Neighbors’ with Google DeepMind demonstrates a new frontier in AI-assisted animation language.

The disparate discussions surrounding Connie He’s AI tools, Milton Glaser’s 1957 menswear catalog, and Mark Nichols’s emphasis on human curiosity reveal a core contemporary challenge: the integration of generative processes with established principles of visual communication. These accounts, spanning decades and methodologies, collectively highlight a pivot in how “creation” is defined, from a singular human hand to a collaborative, often algorithmic, partnership, influencing everything from large-scale conceptual forms to micro-typographic decisions. The common thread is the search for expressive visual language, regardless of its origin.

The excitement around tools like those developed by Connie He and Google DeepMind, which facilitate the creation of novel animation languages, often overshadows critical questions regarding their impact on foundational typographic legibility. Industry rhetoric often champions the speed and sheer volume of AI output as progress. However, this view overlooks that genuine typographic innovation often arises from a deep, human-centric understanding of reading experience, not just visual novelty. Mark Nichols of WMH&I speaks of inspiration from curiosity and the joy of making, a stark contrast to the often opaque ‘black box’ of AI generation, which may prioritize pattern recognition over the subtle, human-driven intent behind legibility and hierarchy.

While such AI systems can produce endless variations of conceptual letterforms, the inherent understanding of optical adjustments, stroke weight relationships, and counterform balance – elements painstakingly refined over centuries of human typography – is frequently absent. The assumption that AI-generated conceptual letterforms inherently possess the necessary micro-typographic sophistication for diverse applications is a misapprehension. The systems learn from existing data, but their “understanding” of reading flow, letter spacing, or the subtle grid adjustments that define true editorial elegance remains superficial, at best. For instance, an AI might generate a visually striking display typeface, but fail to account for the minute variations in kerning pairs required for consistent optical density across all characters, leading to fragmented legibility in body text applications. By mid-2027, the industry will see a rise in specialized AI ‘finishing’ tools, specifically designed to correct these micro-typographic deficiencies, attempting to graft human-learned nuance onto machine-generated forms.

The most potent opposing force to the uncritical adoption of AI in typography comes from practitioners like Mark Nichols, whose studio WMH&I grounds its work in human emotion, curiosity, and the tangible “joy of making.” This approach prioritizes iterative, thoughtful design, where decisions about conceptual letterforms and grid systems are made with a conscious understanding of their psychological impact and historical precedent, rather than through algorithmic efficiency. This human-centered resistance emphasizes the qualitative aspects of design that AI, by its nature, struggles to replicate, focusing on the subtle expressive power of type that goes beyond mere information conveyance.

A working Typography professional should, this week, experiment with taking an AI-generated conceptual letterform set and manually apply micro-typographic corrections to it across a range of applications, from display headlines to short paragraphs. Specifically, focus on adjusting individual kerning pairs, optical weight distribution, and the subtle alignment of baseline and cap-height anomalies. This exercise will expose the specific deficiencies of current AI-generated type and sharpen the designer’s human eye for the nuances of legibility that algorithms currently miss, allowing them to better integrate or critique such tools.

---

Curated References