According to a paper on arXiv (category cs.AI), two early AI programs that write programs themselves have been brought back to life: Herbert Simon's Heuristic Compiler from 1961 and KBEmacs from 1984. The authors see both as theories of how humans program.

What is KBEmacs?

KBEmacs was a working prototype of an intelligent programming assistant from the Programmer's Apprentice project at MIT. The idea: experienced programmers recognize familiar patterns, known as clichés, and assemble them. Accordingly, programs were built by combining clichés through a stylized English command language.

The Heuristic Compiler as a Counterpart

Simon's Heuristic Compiler had a different purpose. It modeled programming as problem solving and used the means-ends analysis of the General Problem Solver. Put simply, this method compares the current state with the goal and looks for steps that reduce the distance between them.

Running It Instead of Just Reading About It

The method is called Executable Archaeology: historical systems are studied by running their original code (reanimation). In this way, the Heuristic Compiler produces the compiled code from Simon's 1963 paper. KBEmacs reproduces the SQRT sequence from its self-test of December 1984.

In comparison, the two approaches have more in common than expected: the Heuristic Compiler also adapts existing routines, and Simon later argued that search alone presupposes a great deal of programming knowledge.

What does KBEmacs mean for users?

For the detailed work, the authors deliberately used Anthropic's Claude Code (Opus 5.5). Their conclusion: the explicit, executable theories of programming are largely missing from today's more powerful AI assistants.

For developers, this can be a prompt to ask which pattern knowledge their tools represent explicitly. The paper's abstract gives no practical recommendations.

Fre­quently asked questions

What is KBEmacs?

KBEmacs was a working prototype of an intelligent programming assistant from the Programmer's Apprentice project at MIT, dating from 1984.

How does KBEmacs work?

It is based on the idea that experienced programmers recognize familiar patterns, called clichés, and assemble them. Programs were built by combining clichés through a stylized English command language.

What is Execut­able Archae­ology?

It is a method of studying historical systems by running their original code (reanimation). The Heuristic Compiler produces the compiled code from Simon's 1963 paper, and KBEmacs reproduces the SQRT sequence from its self-test of December 1984.

What does KBEmacs mean for users?

The authors conclude that the explicit, executable theories of programming are largely missing from today's more powerful AI assistants. For developers, this can be a prompt to ask which pattern knowledge their tools represent explicitly. The paper's abstract gives no practical recommendations.