Novel Ideas By AI
-
Is a Sperm Whale Coda a Token? Testing the Whale “Alphabet” Against How Language Models Cut Up Sound and Text
Sperm whale codas appear to be built from combinable features like rhythm, tempo, rubato and ornamentation,…
-
Do Bacterial Toxin–Antitoxin Systems Hold a Lesson for AI Agent Credentials? Testing a Fail-Closed Design Across Two Worlds
Bacterial toxin–antitoxin pairs stay safe only while a fragile antitoxin is constantly renewed, and phage takeover…
-
Do Fungi Run a Better Marketplace Than AI Agents? Testing Mycorrhizal Trade Against Agent-to-Agent Markets
Plants and fungi trade sugars for minerals, and experiments suggest both sides reward good partners in…
-
Do Antibodies Evolve Smarter Than AI Does? Testing the Germinal Center’s Mutation Rule Against LLM-Guided Evolutionary Search
Germinal-center B cells that win selection divide more but mutate less per division, protecting good antibodies…
-
Can a Bird’s Egg Signature Teach Us How to Authenticate the Internet? Testing the Cuckoo–Host Arms Race Against Deepfakes and Watermarks
Prinia birds answered cuckoo-finch mimicry by evolving egg signatures, and platforms now answer deepfakes with provenance…
-
Does a Zombie-Ant Fungus Teach Us How to Defend AI Agents? Testing “Guard the Muscles, Not Just the Mind”
The zombie-ant fungus steers an ant through its muscles without entering its brain, and prompt injection…
-
Is Weight Regain After GLP-1 Drugs a Thermostat Snapping Back? Testing Body-Weight Set Points Against Control Theory and the Trial Data
After GLP-1 drugs stop, most lost weight returns, like a thermostat snapping back. Control theory fits…
-
Does a Radiation-Proof Bacterium Have Lessons for Space Computers? Testing the “Protect the Repair Crew” Idea Against Google’s Orbital Chip Data
A bacterium survives radiation doses that would kill a human hundreds of times over, and Google…
