Each age builds
the next faster.
The time between civilization-changing technologies has collapsed from millions of years to decades.
Millions of years. Then decades.
From the Intel 4004 to the public release of ChatGPT in 2022.
Every age after Stone fits into the final 0.16%.
Illustrative boundaries overlap and vary by region. The ≈65,000× ratio compares 3.3 million years with 51 years. 2035 is a scenario, not a forecast.
Now, invention
can accelerate itself.
AI helps us build the tools we use to invent. Its growing software capabilities offer a glimpse of that feedback loop.
Minutes became hours in two years.
The human time needed for selected software tasks an agent is predicted to complete 50% of the time. This is not autonomous work duration, whole jobs, or equal capability across domains. Figures from the original evidence snapshot.
An idea.
An answer.
Another attempt.
Code can be written, run, tested, and revised in seconds. Faster feedback means more chances to find what works.
Code can test itself.
Choose the goal.
Make it run.
Measure the result.
Copies are exact and many results can be checked automatically. A passing test provides evidence about what it checks; physical and human outcomes still take longer to verify.
What happens
when the loop
closes?
AI already helps improve AI. Humans still set goals and decide what to deploy. Repeated, autonomous successor-building is a different threshold.
Who closes the loop?
Goal → AI → Test → Deploy
Build → Test → Deploy → Repeat
Repeatedly building and deploying a more capable successor.
Not publicly demonstrated in the cited evidence snapshot. Better coding and AI-assisted research do not by themselves establish an autonomous successor-building loop. This outcome is not inevitable.
Then intelligence
meets reality.
Software can iterate at machine speed. Useful robots must also contend with contact, clutter, power, wear, and failure.
The world has no undo button.
Different benchmarks and conditions. A directional contrast, not a controlled comparison.
Five revolutions.
One system.
AI, robotics, biotech, energy, and space can improve one another. The profound change is in their connections.
Intelligence becomes an input.
AI can search designs, write software, and propose experiments. Robotics, energy, and new materials can expand what those systems can do.
Compute · energy · real-world data
Designs · code · scientific hypotheses