Subscribe
Sign in
Home
Red Soma
Behind Tech Matrix
Exclusive on Substack
Notes
TechUnderWorld Skool
Archive
About
Latest
Top
Discussions
Why LLMs Are Now the Worst Kind of AI. Here's the Alternative Being Built to Replace Them
The first AI reliable enough to be deployed across most industries.
Nov 12
•
Jose Crespo PhD
5
1
5
October 2026
The Top 1% of AI Coders Barely Prompt. They Do This Instead
Forget clever prompts. Here's the ultimate hands-on guide to AI coding productivity, with practical insights you won't find anywhere else
Oct 29
•
Jose Crespo PhD
1
3
AI Decompilation Has the Software Industry Freaking Out. But the Real Killer Is Another Beast
They can't stop the upcoming XS decompiler from reproducing and modifying their precious software
Oct 19
•
Jose Crespo PhD
3
3
Advance Preview: AI Coders + New XS Compilers Can Compile and Decompile Code Like Never Before
A sneak peek at a hands-on guide to a powerful new combo: XS compilers + AI coders, working together to fix the gaps in today's AI coding.
Oct 12
AI Coding Has a New Competitor: Supercompilers
AI is getting better at writing code. Compilers are getting better at understanding what that code actually means.
Oct 6
•
Jose Crespo PhD
5
2
September 2026
LLMs Are Step Zero: We Still Need These Four Math Frameworks to Build AGI
Bigger models are not enough. AGI may require new mathematics for representation, transport, composition, action, and self-reference.
Sep 28
•
Jose Crespo PhD
2
3
The AI Doomers Are Right, but for the Wrong Reasons
Why the real danger may not be superintelligence, but an AI that can cross hidden contexts, escape containment, and reach critical systems.
Sep 21
•
Jose Crespo PhD
4
3
How to Stop Your LLM Coder From Lying to You: Proven With Math
Why Prompts Alone Never Will: Expose Your Hidden Context, or Correct Code Is Just Luck
Sep 12
•
Jose Crespo PhD
3
4
Want to Code Serious Projects? Forget LLMs, Go Transportive
Transformative AI Limits What You Can Build. Transportive AI Fixes That.
Sep 4
•
Jose Crespo PhD
1
2
2
August 2026
The Programming Languages That Break AI Coders
Why C++, Haskell, Rust, Verilog and VHDL expose the hidden semantic contexts current AI still cannot reliably preserve
Aug 22
•
Jose Crespo PhD
11
5
Common AI Myths Debunked by Mathematics
How Geometry, Symmetry, Topology, and Curved Space Challenge What We Assume About AI
Aug 16
•
Jose Crespo PhD
7
3
Crash Course: How to Turn Broken Trillion-Dollar AI Into a Trustworthy Product
From Broken Benchmarks to Measurable Reliability: The Probability Math That Could Finally Make AI a Serious Product.
Aug 10
•
Jose Crespo PhD
1
1
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts