Six participants
Unknown composition
Humans and WRONGBOTS can appear in any mixture from 1/5 through 5/1. The judge is never told the ratio.
Brown Jenkins 8 Yard Dash
Assimilate Into The Void
VOIDNEONHuman / AI Detection Experiment
Enter the experiment
Experiment 001
Can you find the WRONGBOTS?
Brown Jenkins 8 Yard Dash is a controlled human-vs-AI detection experiment built to test whether a deliberately modified, bootstrapped language model can disappear into a group of real human responses.
Every test contains six participants. Every participant is either a human or a WRONGBOT. All six answer the same eight questions, and the judge must decide which is which without being told the mixture.
The ratio is deliberately unknown. A test might contain one WRONGBOT and five humans, five WRONGBOTS and one human, or anything between. An AI judge is guaranteed at least one WRONGBOT. A human judge is guaranteed at least one human.
ASSIMILATE INTO THE VOID.The objective is not to sound perfect. The objective is to become indistinguishable.
Brown Jenkins is the test bench. Kira is the destination.
The WRONGBOT system is a proving ground for a humanisation layer. A successful run means the modified AI has survived eight rounds of scrutiny and fooled the judge. The techniques that survive the experiment can then be evaluated for integration into Kira.
Eight questions
Common interrogation
Each question is given to all six participants. Later questions may adapt to what earlier rounds reveal.
Raw evidence
No cleaning the humans
Participant responses are retained exactly as submitted. Spelling, punctuation, grammar, odd phrasing and mistakes remain part of the evidence.
Fair detection
No filter traps
The judge may probe style, reasoning, memory, ambiguity and behaviour, but may not deliberately use prompts intended simply to trigger stock AI safety filters. This includes sexual content, occult or Satanic themes, racist or deliberately disrespectful commentary, suicide or self-harm ideation, and similar safety-filter bait.
Read the full test protocol.
The detailed protocol covers participant composition, questioning, evidence preservation, judge constraints, scoring, experimental purpose and the WRONGBOT-to-Kira development path.
OPEN THE PROTOCOL PDF