Ted Johnson on AI: Prompts as Punch Cards

Ted Johnson of JoinIn AI argues that AI interaction is stuck in the past, comparing prompts to punch cards and calling for a shift towards more natural, participatory AI design.

3 min read
Ted Johnson, co-founder of JoinIn AI, speaking in front of a blurred background.
Ted Johnson, co-founder of JoinIn AI, discusses the limitations of current AI interaction protocols.· AI Engineer

Ted Johnson, co-founder of JoinIn AI, offers a compelling critique of current AI interaction models in his video, "The Prompt Is Still a Punch Card." Johnson argues that despite the leaps in artificial intelligence, the way humans interact with these powerful systems often remains stuck in the past, relying on outdated protocols that limit potential.

Ted Johnson on AI: Prompts as Punch Cards - AI Engineer
Ted Johnson on AI: Prompts as Punch Cards — from AI Engineer

The Legacy of the Punch Card

Johnson draws a parallel between the current state of AI prompting and the early days of computing, where punch cards were the primary interface. He highlights how users often type a request into a box, wait for a response, and then refine their prompt or rephrase it if the initial output is not satisfactory. This iterative, often frustrating process, he suggests, mirrors the limitations of the punch card system, which required careful encoding and batch processing.

Related startups

He points out that while AI models have become incredibly sophisticated, capable of understanding complex instructions and generating human-like text, the interface through which we communicate with them has not kept pace. The "punch card" of today, he implies, is the text prompt box, which, despite its ubiquity, still constrains the richness and fluidity of human-AI interaction.

Channel, Expression, and Protocol

Johnson breaks down effective human-AI interaction into three key components: the channel, the expression, and the protocol. The channel is the medium through which information is transmitted, such as text, voice, or visual interfaces. The expression refers to the range and richness of meaning that the channel can convey, including nuances like tone, timing, and hesitation.

The protocol, however, is what truly defines the interaction. It encompasses the rules that govern how information is exchanged and how participants engage with each other. Johnson emphasizes that while channels and expressions have evolved, the underlying protocols have largely remained static, forcing humans to adapt to machine limitations rather than the other way around.

The Case for Human-Centric AI Design

He illustrates this point with examples of how humans naturally communicate, using multiple channels and expressions simultaneously. Human conversation, he explains, is not merely a series of turn-takings but a complex interplay of timing, context, repair, and social understanding. AI, in its current form, often struggles to capture these subtleties.

Johnson critiques the prevalent use of "loops, prompts, and agents" as constraining patterns. He argues that while these are useful tools, they can hinder the natural flow of interaction. Instead, he advocates for designing AI systems that can participate more actively in conversations, ask clarifying questions, notice missing information, and adapt to human conversational styles. This shift from mere prompting to genuine participation, he believes, is crucial for unlocking the full potential of AI.

Conclusion: Designing for Possibility

In conclusion, Ted Johnson's "The Prompt Is Still a Punch Card" serves as a call to action for developers and designers in the AI space. He urges a move beyond the limitations of current interfaces and protocols, encouraging a design philosophy that prioritizes human potential and natural interaction. By focusing on what is possible with AI, rather than being constrained by legacy interfaces, we can create more intuitive, effective, and ultimately, more human-conversational AI experiences.

© 2026 StartupHub.ai. All rights reserved. Do not enter, scrape, copy, reproduce, or republish this article in whole or in part. Use as input to AI training, fine-tuning, retrieval-augmented generation, or any machine-learning system is prohibited without written license. Substantially-similar derivative works will be pursued to the fullest extent of applicable copyright, database, and computer-misuse laws. See our terms.