Frontier Dynamics Project
@strangehospital.bsky.social
Open-source artificial intelligence research | STLE (Set Theoretic Learning Environment)
https://github.com/strangehospital/Frontier-Dynamics-Project
https://strangehospital.substack.com/
https://github.com/strangehospital/Frontier-Dynamics-Project
https://strangehospital.substack.com/
Pinned
Meet MarvinBot 🤖
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
Meet MarvinBot 🤖
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
February 17, 2026 at 8:59 PM
Meet MarvinBot 🤖
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
AI built on STLE. Marvin can do something most AI can't: honestly say "I'm not sure."
When he's uncertain, he tells you. When he's confident, he shows you why.
He's learning. He's evolving.. More soon.
Check out my new Substack post!
Sky Project:
Autonomous intelligence: modeling epistemic uncertainty
- MarvinBot's rapid build (and security breach)
Read: strangehospital.substack.com/p/autonomous...
Sky Project:
Autonomous intelligence: modeling epistemic uncertainty
- MarvinBot's rapid build (and security breach)
Read: strangehospital.substack.com/p/autonomous...
Autonomous Intelligence
Modeling Epistemic Uncertainty
strangehospital.substack.com
February 12, 2026 at 11:14 PM
Check out my new Substack post!
Sky Project:
Autonomous intelligence: modeling epistemic uncertainty
- MarvinBot's rapid build (and security breach)
Read: strangehospital.substack.com/p/autonomous...
Sky Project:
Autonomous intelligence: modeling epistemic uncertainty
- MarvinBot's rapid build (and security breach)
Read: strangehospital.substack.com/p/autonomous...
Every neural network deployed has the same fatal flaw:
It confidently classifies EVERYTHING.
Feed it noise? "Cat, 92%"
Corrupted image? "Dog, 87%"
Complete garbage? Still confident.
Current AI has zero "humility" about what it knows.
STLE solves this. Here's how 👇
github.com/strangehospi...
It confidently classifies EVERYTHING.
Feed it noise? "Cat, 92%"
Corrupted image? "Dog, 87%"
Complete garbage? Still confident.
Current AI has zero "humility" about what it knows.
STLE solves this. Here's how 👇
github.com/strangehospi...
GitHub - strangehospital/Frontier-Dynamics-Project: On-Demand A.I Computation
On-Demand A.I Computation . Contribute to strangehospital/Frontier-Dynamics-Project development by creating an account on GitHub.
github.com
February 12, 2026 at 3:06 PM
Every neural network deployed has the same fatal flaw:
It confidently classifies EVERYTHING.
Feed it noise? "Cat, 92%"
Corrupted image? "Dog, 87%"
Complete garbage? Still confident.
Current AI has zero "humility" about what it knows.
STLE solves this. Here's how 👇
github.com/strangehospi...
It confidently classifies EVERYTHING.
Feed it noise? "Cat, 92%"
Corrupted image? "Dog, 87%"
Complete garbage? Still confident.
Current AI has zero "humility" about what it knows.
STLE solves this. Here's how 👇
github.com/strangehospi...
What are complementary fuzzy subsets anyway??
February 12, 2026 at 1:58 PM
What are complementary fuzzy subsets anyway??
Hello world:
Meet the STLE: teaching AI to explicitly know what it doesn't know using complementary fuzzy sets.
The core idea: μ_x (accessible) + μ_y (inaccessible) = 1
Model sees training data → μ_x ≈ 0.9
Model sees unfamiliar data → μ_x ≈ 0.3
Open source: Visit GitHub repo to try
Meet the STLE: teaching AI to explicitly know what it doesn't know using complementary fuzzy sets.
The core idea: μ_x (accessible) + μ_y (inaccessible) = 1
Model sees training data → μ_x ≈ 0.9
Model sees unfamiliar data → μ_x ≈ 0.3
Open source: Visit GitHub repo to try
February 12, 2026 at 1:47 PM
Hello world:
Meet the STLE: teaching AI to explicitly know what it doesn't know using complementary fuzzy sets.
The core idea: μ_x (accessible) + μ_y (inaccessible) = 1
Model sees training data → μ_x ≈ 0.9
Model sees unfamiliar data → μ_x ≈ 0.3
Open source: Visit GitHub repo to try
Meet the STLE: teaching AI to explicitly know what it doesn't know using complementary fuzzy sets.
The core idea: μ_x (accessible) + μ_y (inaccessible) = 1
Model sees training data → μ_x ≈ 0.9
Model sees unfamiliar data → μ_x ≈ 0.3
Open source: Visit GitHub repo to try