Not to be confused with The Gauntlet
There is no product with that name. "Run the gauntlet" is a figure of speech, used in talk about his checks, which some third-party blogs then turned into a name. The tool is SwarmForge.
Where you see it
The repo is unclebob/swarm-forge. Justin Martin made its first commit on 31 March 2026, and Robert C. Martin has made 716 of the 723 commits across its branches (git log, 5 November 2026). It has packs of two, four and six agents on branches of those names (two-pack, four-pack, six-pack). The six-pack README gives "specification, implementation, cleanup, architecture, mutation hardening, and final QA to six separate agents": specifier, approval, coder, cleaner, architect, hardender (his spelling in the files), QA. Handoffs out of the specifier are "held in Attention for operator approval". The specifier writes Gherkin, a plain-text format for test scenarios. His own report of one run, on 4 June 2026: "Outcome was a usable application with a solid architecture and overloaded with unit tests, acceptance tests, and QA tests." That was not measured. On not reading the code, his own post of 23 July 2026, which we read through a mirror because X blocks fetching: "My current strategy is to not read any of the code written by my agents." It ends: "In the end, I have very high confidence in the code they produce because they've had to run the gauntlet of all of my constraints and tests." StartupFortune misquotes that last part as all these constraints, so check before you repeat it. That report also quotes Grady Booch against: "As an experienced developer, I know the smell of what is good and what is not. And no agent has either the experience or the context to know those things." We have Booch from that one secondary report only. We found no outside measurement of SwarmForge.