Pietro Schirano: From Human Prompting AI to AI Prompting Humans in 6 Months
Designer Pietro Schirano stated that in just 6 months, he evolved from "human prompting AI" to "AI self-prompting" and then to "AI prompting humans."
Schirano lamented the "crazy" speed of AI development, reflecting the rapid shift in human-computer interaction paradigms. The traditional model of "human inputs commands → AI executes" is being replaced by "AI actively suggests → human confirms."
Funding is shifting from prompt engineering to autonomous agents, benefiting companies in AI workflows. Traditional human-computer interaction design is under pressure due to increased AI autonomy, leading to a decline in the value of "prompt optimization" skills.
By 2026, AI autonomy will significantly improve, evolving from needing detailed prompts in January to being able to autonomously plan tasks and actively inquire about human preferences by July, reflecting the rise of the "human-in-the-loop" paradigm.
Source: Public Information
ABAB AI Insight
Historical Behavior
Pietro Schirano, as a designer and early adopter of AI, reflects the evolution of the industry. In 2023, he relied on detailed prompts ("Write a 300-word copy, in a professional tone, including 3 key points"), in 2024 he used simplified prompts ("Help me write copy"), in 2025 he used AI to autonomously generate ("Help me complete this task"), and by 2026 he entered the "AI prompting humans" stage (AI actively asking "What do you need me to help you with?"). Comparing ordinary users (still in the "detailed prompts" stage in 2026) and developers (who have entered the "AI autonomous planning" stage in 2026), Schirano's "6-month evolution" represents the trajectory of "frontier users."
Capital Path
The capital impact of the evolution of human-computer interaction is clear: in 2023, the market size for "prompt engineering" was $500 million (courses, tools, consulting), which is expected to drop to $100 million by 2026 (due to increased AI autonomy and decreased demand for prompts). The market size for "AI autonomous agents" is projected to be $15 billion in 2026 (Devin, Cognition, OpenClaw), with an expected rise to $50 billion by 2027. Funding is shifting from "prompt optimization" (one-time learning, marginal cost 0) to "AI subscriptions" (ongoing payments, monthly fees of $20-500). In the long term, as AI can autonomously plan tasks, the value of "prompts" approaches zero, while the value of "task definition" rises—humans shift from "writing prompts" to "defining goals," while AI shifts from "executing commands" to "planning paths."
Analogy + Industry Position
Similar cases include: in the 2010s, "search engine optimization (SEO)"—the market size was $5 billion in 2010, dropping to $1 billion after Google's algorithm upgrade in 2020, and expected to fall to $100 million after the rise of AI search (Perplexity, You.com) in 2026. In the 2020s, "App interface design"—the market size was $20 billion in 2020, expected to drop to $5 billion after the rise of AI-generated UI (Figma AI, Galileo AI) in 2026. Currently, human-computer interaction is in the "prompt extinction phase"—compared to 2023 (prompts necessary, 90% users), by 2026 (prompts optional, 50% users), and by 2030, prompts are expected to disappear (10% users, only in professional scenarios). The industry position is at the "AI autonomy inflection point"—users shift from "commanding AI" (2023) to "collaborating with AI" (2026) and then to "delegating to AI" (2030), with interactions shifting from "explicit commands" to "implicit intentions."
Structural Judgment
This fundamentally represents a technological substitution leading to a reconstruction of human-computer relationships. Traditional human-computer interaction is "human-led"—humans define tasks, plan steps, and execute commands, with AI merely assisting (searching, calculating, generating). AI autonomous interaction is "AI-led"—AI defines tasks, plans steps, and executes commands, with humans merely supervising (confirming, correcting, handling exceptions). This explains why Schirano laments the "crazy speed"—it is not about "functional upgrades" but rather about "power transfer." When AI can autonomously complete 90% of tasks (writing code, designing, analyzing), humans only need to handle 10% of exceptions, shifting the "dominance" from humans to AI. This drives the "polarization of human-computer relationships"—high-end scenarios (complex decisions, high risks) retain human dominance, while low-end scenarios (repetitive, low risks) are AI-led, and the middle layer (semi-automated) disappears. As AI autonomy becomes widespread, "prompts" shift from a "necessary skill" to a "historical relic," with "task definition" becoming the new skill—humans transition from "operators" to "supervisors."
ABAB News · Law of Cognition
Prompts are a transition; intent is the endgame.
The speed of transfer of human-computer dominance is faster than technological iteration.
The higher the autonomy, the more abstract the human role.