Software can open a new possibility when it can be extended beyond the original task.

Howard Pearlmutter began working with LISP in 1974 and FORTH in 1978. Extensible-language approaches became part of practical commercial applications, including graphics, educational software, and games.

His 1981 international speaking tour included the theme “Tools that can Learn.” It invited a broad view of adaptable computational capability and what people could create with it.

Three kinds of growing ability.

A person can learn through a tool. An interactive model, for example, can help the user discover a relationship. The new understanding belongs to the person.

A program can be extended. New behavior can be supplied through additional code or configuration, making the tool capable of a task it did not previously perform. Extensible languages make that kind of growth central to how software is built.

A computational system can also learn or adapt through data and feedback, according to its mechanism. That change is different from a developer explicitly adding a behavior. Understanding the mechanism helps a user judge what the adaptation establishes.

These possibilities can work together. An AI assistant may help a person understand a system and help produce an extension to it. The conversation can become more informed through supplied context without implying that the underlying model's training has changed.

For a practical example, imagine a developer explaining an existing interface to an assistant, examining a proposed change, and using feedback from the result to refine the work. The developer may learn. The application may be extended. The interaction may become better informed. A claim that the model itself has been retrained would require a separate basis.

The useful distinction is where the new ability resides and how it arose. That gives a learner a clearer way to describe the result and recognize which responsibility remains.

Today's AI-assisted development gives the learning question another immediate form. A tool can help generate, explain, explore, or transform work. The person still needs to understand what the result means and how to use it in the situation at hand.

That calls for both capability and judgment.

A learning experience can examine what becomes easier, what new work appears around the result, and which human expertise becomes more valuable. It can also help people recognize an opportunity they would not have attempted before.

Howard's current practice includes AI-assisted engineering across native integration, enterprise middleware, media workflows, and prototypes. His Telos engagement combined foundational understanding with practical demonstrations and attention to skeptical concerns.

Read Telos.

The longer history supplies perspective beyond a single model or release. It can help a group examine adaptation, feedback, and the relationship between tools and human ability.

Explore feedback, mental models, and learning about tools.

The next session might be a seminar, workshop, or mentoring contribution shaped around what you want to become able to do. Begin with the capability and the responsibility it brings into your work.