About Pythology · Manawatū, New Zealand
What is actually causing this?
I did not start Pythology with a grand plan to build an intelligence company. I kept building things, asking harder questions, and running into the same problem: software could often give me an answer, but it was much harder to see why the answer deserved to be believed.
Why I kept going
I wanted the system to be able to say “here is why I think that” — and later admit when reality disagreed.
That idea kept following me from one project into the next. What changed first? Which evidence is real and which part is inferred? What else could explain it? If the explanation is right, what should happen next? And, perhaps most importantly, what would prove us wrong?
Eventually those questions stopped looking like features and started looking like the architecture itself. EarthNet became a place to test the idea against the messy outside world. Prometheus became the part willing to write a forecast down before the answer was known. Sentinel became the idea that a human should be able to challenge the machinery instead of simply accepting whatever appears on the screen.
Pythology is still early. I would rather say that plainly than dress six months of building up as a twenty-person laboratory. The interesting part is that there is now enough working machinery to test the idea properly.
How the path unfolded
The projects were not a perfectly planned roadmap. Each one exposed a problem underneath the previous one, until the common thread became difficult to ignore.
Start with real data, real users and software that has to work outside a laboratory.
Bring many changing planetary signals together without confusing observation with explanation.
Ask which mechanisms could connect the evidence and which alternatives remain alive.
Say what should happen next before the answer is known, then keep the record.
Let a human ask what the system sees, doubts, assumes and would need to change its mind.
What comes next
The longer-term ambition is to test the same evidence discipline in problems where the hidden mechanism matters even more — particularly biology. That does not mean importing EarthNet confidence into medicine. Every biological claim has to earn its own evidence, validation and expert scrutiny.
Biological Intelligence is our research direction for connecting reviewed evidence, biological state, causal mechanisms, molecular interrogation, experiments and later outcomes. The goal is not autonomous diagnosis. It is a more inspectable way to reason through complicated living systems.
Explore the biological research →Migraine is a good example. New Zealand data shows a very large burden, especially for women, while many people remain undiagnosed and the country still lacks the data needed to measure the full economic cost. Problems like that make us ask whether better evidence integration and mechanism-focused research could expose useful gaps.
See the NZ Health Survey evidence →The eye is our preferred first regenerative proving problem: cornea, lens, retina, retinal pigment epithelium, retinal ganglion cells and optic nerve all provide different bottlenecks with unusually measurable function. A regenerative claim would have to survive the full chain from molecular change to structural restoration, functional restoration, durability and safety.
See the wider research direction →The rules I do not want to lose
The models will change. The software will change. Hopefully the company will get much bigger. These are the habits I want to survive all of that.
Measured, reported, modelled and inferred information should not silently become the same thing.
Missing evidence and disagreement tell us where the next question lives.
A preferred explanation should earn its place against plausible competing mechanisms.
Forecasts, experiments and later observations decide whether confidence deserves to rise or fall.
Talk to Pythology
If the architecture overlaps with something your organisation needs to understand, I would rather start with the problem than give you a generic sales pitch.
.jpeg)