Scanning the surface.
A preview edition.
More to come.
Always.
There’s always more to come.
And always later.
Interesting stuff 1
I got an email from one of our IT service suppliers about how they joined another company for scale and efficiency reasons and how they made some changes in management. So far so good. So far all generic text and reassuring language. “Nothing will change for you”, “you can be sure you receive the quality you deserve” or something along those lines.
I almost archived the mail when my eyes landed on the final sentence: “So, that’s why I wanted to schedule a call to get to know you better”, signed, the new manager. I thought that was surprising and strange and a bit weird (“is he going to call +200 clients?”). But I suggested some dates and 1 week later we were having the call.
And it wasn’t awkward or weird at all. It was great. It was really nice to see behind the new “manager” there was a real human being. With a family and life and a work history and passion and enthousiasm. I was sold. I felt guilty for thinking it weird when humanity comes peaking through the AI dominated biz-speak and corporate blurps.
An evening stroll
I was walking my dog the other day when we ran into the butterfly hanging the laundry
“Good morning butterfly”, I said
“Good morning”, the butterfly said
“I quit smoking”, the butterfly said
“You quit smoking”
“I got married”, the butterfly continued
“You got married”
“I started drinking”, the butterfly added
“You started drinking”
My dog was licking his balls
“I underestimated the butterfly”, I thought while walking away
…
Interesting stuff 2
We are in the age of AI. Or so the hype goes. Everyone and everybody is jumping on AI. And I think rightfully so. It is another abstraction level up in programming skills. It does leverage creative and skillful brains. It will make certain types of work irrelevant. But it will also create a lot of noise and create more work through the Jevons paradox. There will be a shift but maybe not in the way people think.
One of the things that will happen (is happening) is governments and companies and organisation building and training their own AI models. With Claude and Mistral and Gemini everyone has access to world class lawyers, big 4 consultancy advice or big 3 strategic analysis and board room plans. But everyone has access to the same lawyer, the same consultant, the same strategist.
You can get a competitive advantage by prompt engineering and context engineering. But those are hard to persist and get consistent. What organisations want is unique but predictable outcomes and output. So they build and train their own models, their own contexts and their own guard rails. To get good models, you need 3 things:
- feature data
- training data
- a model
As this is happening, most focus is on the modeling side. In the geospatial world (earth intelligence), the focus is on geofoundational models like Clay. And they are promising and will be great. But to make them competitive for companies, companies should build their own models on top of those foundational models. And they will.
But what I think is overlooked structurally is the need for good feature data. In geospatial that is consistent and coherent satellite composites and mosaics. And everything on top of that. The quality of your model is irrelevant when you have inconsistent feature data. And in geospatial, most feature data is still inconsistent and incoherent. Because of lack of standards and benchmarks. This is the reason the META global biomass is impressive, but unusable. The model, the math is on a different level from everything that came before. But because of crappy feature data, the output is crappy as well.
The feature data overall quality will improve over time. But for now, it’s a competitive advantage if your organisation uses consistent and coherent geospatial datacubes.
Interesting stuff 3
This is so interesting it’s secret.
Interesting stuff 4
This one is not secret but really not that interesting.
.