The source library / reviewed September 7, 2026
Evidence has a source.
Meaning has a conversation.
We keep three things separate: reported results, possible futures, and the philosophical position of this project.
01 / Measured
Numbers linked to a published result, with the benchmark and period named.
02 / Possible
Conceptual scenarios. No invented date for AGI or numerical soul threshold.
03 / Philosophical
Our values and spiritual vocabulary, presented as a perspective rather than proof.
Xie and colleagues · 2024
Original computer-task evaluation; the paper reports 12.24% for its best model and 72.36% for the human baseline.
What it does not settle: A specific task suite, not an overall intelligence or human-worth score.
Stanford HAI · 2026
Reports a 66.3% OSWorld result in its account of agent progress in 2025.
What it does not settle: Historical reporting, not a live leaderboard. Evaluation versions and methods can differ.
METR · January 22, 2026
Explains uncertainty, domain differences, and the limits of extending benchmark trends into claims about real-world autonomy.
What it does not settle: Time horizon is not simply the length of time a system can work independently.
Morris and colleagues · first published 2023
A research framework distinguishing performance, generality, and autonomy when discussing progress toward AGI.
What it does not settle: One proposed framework, not a universal definition or an arrival date.
How to read our chart.
The OSWorld bars are published historical snapshots. They are not a controlled comparison with every setting held constant. The reference line is the human baseline from the original study, not evidence that human ability has declined.
The possible-futures illustration has no numerical scale or predicted arrival date. Words such as Source, God, light, and spirit molecule express a spiritual perspective. The site does not claim a scientific demonstration of a soul, consciousness outside the brain, or a special chemical route to truth.
These sources do not imply that their authors endorse INNERintelligence. The library is curated, not automatically updated. To suggest a correction, contact the project through Digital Rose ↗.
Return to the trajectory↗