Trained on rocks. Now building AI for incomplete evidence — Mineral Exploration (LodeCraft) and thoughts (Chattly). Perth. Licking rocks professionally.
1/ We could have shown a probability on every LodeCraft target. We deliberately didn't.
We call it a combined exploratory score, not a probability.
Here's why that one word matters 🧵
Just shipped 8 new languages into Chattly!
I wanted as many people as possible feeling welcome , so I guess it is a good start :)
Then I got carried away dressing little fuzzy mascot for each country 😂
Which one is your favourite?
https://t.co/pfmmgDtDcU
I built Chattly because I wanted AI conversations I could come back to—without starting my story over every time.
Somewhere to talk things through, question my own assumptions and figure out a small next step.
Seeing it live on the App Store feels like a big milestone. I’d love your honest feedback.
Try Chattly: https://t.co/ohu8SkmF3r
Learn more: https://t.co/5k6tlelpMZ
@theo When most of the time you are not worried about monthly bills, your brain have so much more space and time for growth…Stress is a cancer that kills you slowly and on some many different levels.
@GlenGoulds Love seeing Vic Au-Sb getting attention.
As a geologist, worth noting ~98% of that 104 g/t AuEq is antimony at assumed prices, and most of it sits in one 1.2 m channel.
A great start, but continuity at depth is everything. Drilling is the real test.
@signulll Creativity like energy can’t be destroyed.
It will just transform into something else.
Tbh, looking forward to what that technology will unleash in creative arts in the next years
@bilawalsidhu I’m still surprised about “sudden” wow over it.
Geologists and mineral exploration companies were doing it for years now.
I guess it’s probably more available and less costly than before
before i let an agent touch my code it has to read the dependency map first. what imports what, what breaks downstream, which tests would catch it
"all tests pass" means nothing until something other than the agent has run them
@mustafasuleyman The part I keep coming back to: if an agent signs a contract outside its authority, who’s liable? You can revoke its credentials and trace what it did, but the other party still needs someone accountable when things go wrong.
@kaelig@theo@mercury Interesting. I ran a small 5-bug test with my cStack: single and multi-agent both fixed 5/5, but multi-agent used 3.54× the tokens and took 40% longer. Just one run per bug, so very early. Did you track token usage per merged PR for this?