WHITE PAPER / OUTLINE
The operating system is part of the experiment.
Each chapter moves from question to mechanism to experiment to evidence, then leaves failure modes and open questions visible.
From the original research notebook. Earlier concepts are preserved here as history. See the current theory map and dispositions →
- 01
What if intelligence is being wasted?
FOUNDATION - 02
Models should think, not wait
FOUNDATION - 03
Context is a scarce resource
FOUNDATION - 04
Evidence instead of operational exhaust
FOUNDATION - 05
Semantic editing instead of text churn
FOUNDATION - 06
Measuring agent trajectories
MECHANISM - 07
Durable state instead of conversation memory
MECHANISM - 08
Scheduling without cognition
MECHANISM - 09
Concurrent agents need resource authority
MECHANISM - 10
Routing models by capability and constraint
MECHANISM - 11
Bounded recovery instead of retry loops
MECHANISM - 12
Autonomous discovery without task explosions
MECHANISM - 13
Authorization for machine actors
MECHANISM - 14
Disposable execution
MECHANISM - 15
Destroy the executor before verification
MECHANISM - 16
Independent review
MECHANISM - 17
Controlled real-agent acceptance
MECHANISM - 18
Everything that broke while building the system
MECHANISM - 19
Benchmarks
EVIDENCE - 20
Negative results
EVIDENCE - 21
What remains unproven
EVIDENCE - 22
Opsle Tasks
EVIDENCE - 23
Reproducing the experiments
EVIDENCE - 24
Future research
EVIDENCE
