EDITORIAL RECORD / RETIRED ESSAY
We took this
essay off the shelf.
Published · Retired
Why it was removed
The 85 percent reliability estimate and universal 99.9 percent product threshold had no task set, measurements or account of the cost of failure. The claimed first-hand agent results had no supporting record in the essay.
ORIGINAL TITLE
AI agents are overhyped but also inevitable
The earlier text is preserved in Git history. It is available as a record of the experiment, with its original wording.
Read the preserved sourceFor work with examples and results, start with the notebook. To try something yourself, open the tools.
Find a problem with this decision? Suggest a correction.