We are programming AI in assembly language
In the series of articles I’ve been writing for Fast Company devoted to what I believe corporate AI should be, I’ve been touching on a very provocative idea: Perhaps the biggest problem of corporate AI as we know it so far is not intelligence, but the level at which we are programming it. As we speak, frontier models, elastic cloud infrastructure, vector databases, managed APIs, and large amounts…
In a series of articles for Fast Company, the author argues that corporate AI suffers from a fundamental problem - the level at which it is programmed. While powerful infrastructure exists, engineers must still manually assemble essential elements like persistence, memory, permissions, tracing, retries, orchestration, durable execution, tool access, context management, and evaluation.
This is reminiscent of the early days of the internet in the 1990s. High-level languages, which simplify human reading and writing, are now commonplace, but decades ago, developers had to work directly with machine or assembly language, as described by John Backus. IBM's Fortran, C, Java, the web, and now the agentic stack all show how abstraction can make programming easier and more portable.
However, the current agentic stack resembles an earlier pre-language period, with Python and frameworks handling fundamental characteristics separately. The proliferation of these tools may indicate a missing layer. Companies like LangChain and Temporal address runtime problems, but they still require manual assembly of primitives that a future AI-native runtime could provide.
The cost of this complexity is comparable to the challenges faced when working with assembly language. To truly industrialize software, capability must be followed by formalization.
Written by urgent.news from Fast Company's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
