Could AI research speed up sharply? The evidence is still conditional
One story is about friction in putting AI to work. Another is an Anthropic filing about what might happen if AI helps speed up AI research. Zvi Mowshowitz sets them side by side, but neither source proves a broad adoption rate or confirms that research has entered a self-reinforcing cycle.
What the source reports
Zvi Mowshowitz puts two kinds of evidence side by side. His newsletter gives examples of workplaces struggling to use AI. It also cites an Anthropic filing that says AI research might speed up sharply. The workplace examples are accounts, not a count of how many groups use AI well. The filing describes a possible future, not a turning point that has happened. These two stories can both be true: research may move fast while daily use remains hard. But this pairing cannot tell us how fast most workplaces are adopting AI. It also cannot show that AI is already making its own research speed up again and again.
Original source
- Title
- AI #193: Progress May Dramatically Accelerate
- Author
- Zvi Mowshowitz
- Publication
- Don't Worry About the Vase
- Date
- Thursday, November 5, 2026