Member-only story
Agentic AI — Prompting is the new Programming
Bahasa Jawa is the new Java Programming Language
I sometimes wonder why so many people believe that our jobs as software developers will be completely replaced by AI copilots and other AI tools.
Have these people ever tried to develop a prompting system for a specific domain? Take, for example, a chatbot for the support department of a household appliances company, dedicated solely to managing customer data. It sounds trivial — a simple CRUD system (Create, Read, Update, Delete) that allows customers to modify their address details. For developers, this would be straightforward with a classic form-based web application. But when building an AI agent, the situation changes drastically.
Why? Because we need to consider the following aspects in our prompting (source: https://chatgpt.com/share/67ad01b8-d81c-8013-a2e6-654edc6705ab):
- Clarity & Precision
- Incremental Thinking (Iterative Refinement)
- Context Awareness
- Goal-Oriented Prompting
- Leveraging AI’s Strengths (Structured Guidance)
- Experimentation & A/B Testing
- Self-Correction & Feedback Loops
- Ethical & Bias Awareness
- Role Assignment for AI
- Handling Uncertainty (Clarifying Unknowns)
When I look at these characteristics as a developer, it becomes immediately clear:
These are exactly the principles we use in programming languages.
Programming languages enable structured communication with machines, supporting all these concepts. Natural languages like Bahasa Jawa or English, however, are not designed for these requirements. That’s why we need well-defined “design patterns” for prompting.
Martin Fowler has begun collecting patterns for the Agentic AI / Generative AI area: https://martinfowler.com/articles/gen-ai-patterns
Prompting is Programming
Prompting elevates us to a higher level of abstraction. However, we have encountered this argument repeatedly in software development:
