Key Takeaways

  • In a 1985 speech, Steve Jobs predicted modern large language models by imagining a computer that captures Aristotle's worldview so future students can ask him direct questions.
  • John Coogan highlights that early personal computing was built around interactive compression: capturing expertise into conversational interfaces.
  • Aristotle's philosophy favors tools that expand human flourishing, rejecting the premise that removing human effort automatically improves life.
  • Building defensible AI products requires creative augmentation over high-volume automated content creation.
  • Product teams evaluating generative software should apply Jobs' Three Demands for AI Worth Using.

The Jobs' Three Demands for AI Worth Using

Rule 1: Give People New Abilities

Learning, designing, composing, building would probably excite him most. AI must empower human capability rather than simply automate.

Rule 2: Make It Effortless

Push for AI that understands what you want with minimal instruction, creating seamless intuitive interfaces.

Rule 3: Make Something Excellent

Value AI's ability to help someone realize an original vision much more than its ability to produce enormous quantities of generic content (pro-AI, anti-slop).

When This Works (and When It Doesn't)

This framework works when you design consumer software or workflow tools where individual human agency drives the outcome. If you are building an interface for a writer, an architect, or an engineer, your software succeeds when the user feels like they gained a superpower. Jobs saw computing as an interactive bicycle for the mind. When an AI tool acts as an active sparring partner that clarifies thinking, users stick around.

The framework breaks down in purely transactional infrastructure. If your product runs automated invoice reconciliation, spam detection, or batch data migration, nobody wants an interactive conversation. In those cases, human effort is a defect, and full automation without creative input is the entire value proposition. Trying to make an accounting pipeline feel like an artistic collaborator adds friction where users want silent execution.

What to Do With This

Audit your product roadmap tomorrow morning. Take your core generative AI feature and run it through each component of the framework:

1. Check the primary output of your tool. If your feature exists to click a button and generate thirty generic social media posts or search-optimized articles, kill it or redesign it. Ask whether it gives the user an ability they lacked five minutes ago, like testing architectural blueprints or stress-testing pitch logic.

2. Measure prompt friction. If a user needs a 300-word prompt template to get a coherent result, your interface failed. Rebuild your context layer so the model infers intent from the user's workspace with minimal manual instruction.

3. Review your quality benchmark. Check whether your model helps a user refine their specific vision or merely spits out average corporate copy. If the output looks like every other LLM wrapper on the market, refocus your engineering effort on creative precision.