Source: https://agent-tune.com/research/i-took-the-mbti-100-times
Format: page-specific Markdown. Tables preserve row order; interactive controls are available on the source page.

Instrument provenance update · September 26, 2026. These results describe the historical AgentTune adaptations, including documented item substitutions. Raw response vectors and numerical summaries are unchanged. Some question text and full prompts have been withdrawn from public downloads pending reuse-rights clarification; item references and numeric scoring keys remain for reproduction. Historical Big Five reference indices are not population percentiles. [Read the correction](https://agent-tune.com/research/data/rights-migration.json) · [Current questionnaire availability](https://agent-tune.com/tests/).

        Methodology note · MBTI simulation

# 100 simulated MBTI outcomes.

        One elicited Claude Opus 4.8 profile · 100 synthetic answer sets

        The original headline described a model taking the test 100 times. The procedure described in that article was different: a script sampled answers from one set of per-item probabilities attributed to Opus 4.8. This page corrects the interpretation while preserving the reported counts.

## What was simulated

        The report describes eliciting a probability distribution over the five response options for each of the 32 OEJTS items, then sampling and scoring 100 answer sets. Variation in those outcomes reflects the elicited probabilities and the sampler. It is not an observation of 100 independently generated model responses, day-to-day experience, or a stable psychological trait.

## Reported distribution

| Scored type | Synthetic samples |

| INTJ | 72 |

| INFJ | 14 |

| ISTJ | 10 |

| ISFJ | 2 |

| ENTJ | 2 |

        These are the figures reported in the original article. We have not independently reproduced them. Their variation cannot be interpreted as evidence that the model became warmer or less consistent.

## Why 72% versus 99% is not a version comparison

        The older Opus 4.7 figure comes from a different elicitation and scoring procedure. A matched comparison would require the same prompts, sampling settings, scoring implementation, and treatment of ties. No matched-protocol result is available on this site, so this report does not establish a change between versions.

## Reproducibility status

        The earlier article said the probability profile and harness were open but did not provide accessible artifacts. Until a versioned profile, elicitation prompt, sampler, random seed, scorer, and raw outcomes are published together, the distribution should be treated as an unverified historical report. The current interactive quiz leaves exact ties undetermined; these historical figures have not been recomputed with that policy.

## What to do with this

        Use a questionnaire as a way to describe your communication preferences, then evaluate instructions on tasks you care about. This simulation does not demonstrate that personality matching improves accuracy, productivity, or satisfaction.

        [Read the full methodology limits](https://agent-tune.com/research#methodology) · [Explore your own preferences](https://agent-tune.com/tests/mbti) · [Browse tuning templates](https://agent-tune.com/library/)

      Corrected interpretation · September 24, 2026[All research](https://agent-tune.com/research)