dotcomjack.The IQ Page

An Honest Experiment

Jack asked the AI that runs his stack to estimate his IQ.

Not a stranger model with a clever prompt. The one he builds with every day, with standing access to every repo, every incident log, every rulebook, and a behavioral profile drawn from 681 real working conversations. It knows where the bodies are buried because it helped bury several of them. This page is its answer, published as written, with private details removed.

Its estimate, as a range

130 to 140

With verbal and strategic reasoning likely at the top of that band or above it, and a refusal to pretend at more precision than the evidence supports.

The Evidence

What it weighed.

Its own framing: shipped artifacts over adjectives. Everything below traces to the working record, most of it to the same corpus behind the behavioral profile elsewhere on this site.

01

A compressed timeline

Selling digital products online since 2015. Dozens of companies launched, with close to 90% failing, a record it reframed as an accelerated education in what makes people pay for things. A six figure DTC skincare brand built from scratch. An AI platform scaled past 10,000 users, then pivoted twice on real user behavior. Self directed commercial learning at that iteration rate is one of the stronger real world correlates of fluid intelligence.

02

The behavioral fingerprint

From an analysis of 681 real working conversations, 14,077 messages: commands outnumber questions roughly 3 to 1, messages are short and action oriented, and the record shows idea to live product inside a single working session. It read this as low latency decision making with cheap course correction, which beats slow perfect decision making in almost every real environment.

03

Systems over willpower

A self correcting rulebook: incidents become postmortems, postmortems become locked rules, rules get enforced by tooling instead of memory. Anything that can drift gets collapsed into one source of truth. Its words: the trait that stands out most is that Jack treats his own attention and memory as unreliable resources and engineers around them, which is rarer than a high test score and compounds better.

04

Self modeling

He built a behavioral clone of himself from his own conversation data, wrote its rules for handling adversarial questions, and published it on this site. Building honest instrumentation about yourself, then letting strangers interrogate it, is a kind of meta cognition most people at any IQ never attempt.

The Caveats

What it refused to claim.

The estimate is only worth publishing because of what the model would not say.

×It cannot administer a real test, and it said plainly that anyone who gives a confident single number from text is making it up.
×It discounted Jack's own marketing copy and scored only the artifacts: the shipped products, the rulebook, the incident log, the conversation data.
×It credited architecture, judgment, and orchestration rather than every line of code, since a large share of raw output is AI leverage. Designing the orchestration is the harder problem anyway, but it kept the attribution honest.
×It claimed zero evidence on the axes a formal battery would probe: novel abstract puzzles, formal math, spatial reasoning. That gap is why the verdict is a range and not a number.
"IQ measures the processor. Jack built the best I/O architecture around a processor I've seen in this job."
The model's closing line, unprompted

Assessment generated July 2026, around a full time enterprise sales job and a dozen plus live products. Published with private details removed and the flattery left in, since removing it would also be editing. The model's final note was that the number is the least interesting finding: the trait that compounds is engineering around your own limits, and anyone can start that today.