5 September 2026 · an experiment currently awaiting its final subject
I have a very important benchmark planned for GPT‑6 Astra.
Not coding.
Not maths.
Not some elaborate reasoning test.
I want to ask it a question I’ve asked ChatGPT before:
Based on everything you know about me, what kind of animal would I be?
Because apparently this is how I evaluate frontier intelligence now. 😂
The previous diagnosis
One of the answers I remember was not content with giving me a single animal.
Apparently I was some combination of a lion, a magpie and a border collie.
Which, when you think about it, is an absolutely unhinged creature.
The lion was the leadership/courage part. The magpie was the tendency to collect ideas from everywhere. And the border collie was the part that apparently wants to organise all of those ideas into something useful.
Then, when forced to stop designing a new species and actually choose one animal, the answer was:
An eagle.
I remember the answer. I do not reliably remember which exact GPT version gave it, so I’m not going to rewrite history just to make the experiment neater.
That actually makes Astra more fun.
The next test
When Astra reaches my account, I’m going to ask the same question again.
No feeding it the old answer. No telling it about the lion-magpie-border-collie situation. Just the question, with whatever context the model genuinely has available at the time:
Based on everything you know about me, what kind of animal would I be?
And then we compare.
The timing is especially funny because my use of ChatGPT has changed so much. What started as talking, writing, asking questions and throwing ideas around has gradually turned into using AI across planning, research, content, Notion, Canva and actual production workflows.
So I genuinely want to know whether a newer model sees the same pattern in me — or something completely different.
Maybe I’m still an eagle.
Maybe the border collie finally wins.
Maybe Astra takes one look at the number of simultaneous ideas and declares me an octopus.
Or maybe it gives an answer none of us would predict.
That is the point of the experiment.
To be continued…
I’m leaving this article unfinished on purpose.
When Astra arrives, I’ll ask the question and add its answer here verbatim enough to preserve what it actually chose and why, rather than retrofitting the result into a better story.
Then we can decide whether GPT has become more intelligent…
…or merely better at wildlife classification. 😂
For the less scientific Astra discussion
I also wrote separately about GPT‑6 Astra, AI capability, Christian discernment, wisdom and stewardship. That piece belongs in Anointed Intelligence™; this one is staying exactly what it is: a ridiculous experiment I now genuinely intend to conduct.
GPT-6 Astra, Discernment and Christian Stewardship: Power Is Not Wisdom