Show: The Cognitive Revolution · Publisher: Turpentine / Erik Torenberg · Host: Nathan Labenz
Episode URL: https://www.youtube.com/watch?v=MrJVgw8dBhw
Publish date: 2026-04-14
Duration: 11449.0s
Default source credibility: MEDIUM — Policy + strategy crossover. Host-speculation mixed with named guest claims. Apply tier per-quote: named-guest-with-metric = HIGH, host/panel speculation = LOW, policy-analyst interviews = MEDIUM.
- The concept of “crunch time” is introduced: a period where AI rapidly accelerates AI R&D, potentially compressing decades of progress into months, creating an urgent, high-stakes scenario.
- The dominant strategy at frontier labs is to use current AI systems to align and control their more powerful successors, a high-risk approach that depends on a narrow window of opportunity.
- The speaker advocates for transparency and early warning systems beyond benchmarks, such as tracking the percentage of software pull requests both written and reviewed by AI.
- The core advice for leaders is to adopt AI aggressively, not just for productivity but to maintain an accurate understanding of the situation, as the pace of change may soon make it impossible to keep up.
Extracted quotes
| # | Credibility | Speaker | Org | Timestamp | Topic | Quote |
|---|---|---|---|---|---|---|
| 1 | HIGH | Ajeya Cotra (Senior Advisor) | Open Philanthropy | 7:15 | 08-radical-vs-tablestakes | At a panel at Dealbook in New York, the moderator asked whether we thought it was more likely than not that by 2030 we would get AGI… seven or eight hands went up… But then he asked a follow-up question… whether we thought that AI would create more jobs or destroy more jobs over the following 10 years… eight out of 10 people… thought that AI would create more jobs than it destroyed. |
| 2 | HIGH | Ajeya Cotra (Senior Advisor) | Open Philanthropy | 31:05 | 08-radical-vs-tablestakes | Meter came out with an uplift RCT, which I think was the first of its kind… They had software developers split into two groups. One group was allowed to use AI, the other group was disallowed from using AI. And they studied how quickly those developers solved issues… it actually turned out that in this case, AI slowed down their performance. |
| 3 | MEDIUM | Ajeya Cotra (Senior Advisor) | Open Philanthropy | 43:15 | 06-security-frontier | Various CEOs have said internally 90% of our lines of code are written by AIs… One thing I’m interested in is what fraction of pull requests to your internal codebase were mostly written by AI and mostly reviewed by AI. […] I’d be very interested in watching that number climb up because it’s an indication both of AI capabilities and of how much deference they’re giving to AIs. |
| 4 | MEDIUM | Ajeya Cotra (Senior Advisor) | Open Philanthropy | 1:06:27 | 06-security-frontier | Figuring out how to create a setup where we use control techniques and alignment techniques and interpretability… to get to the point where we feel good about relying on their outputs is a crucial step to figure out because it either bottlenecks our progress because we’re checking on everything all the time and slowing things down, or it doesn’t bottleneck our progress but we hand the AIs the power to take over. |
| 5 | HIGH | Ajeya Cotra (Senior Advisor) | Open Philanthropy | 2:09:52 | 06-security-frontier | We ended up making $25 million of grants through that [RFP] and then another two to three million from the companion RFP, which was just a broader like all kinds of information from RCTs to surveys about AI’s impact on the world. |
Per-quote detail
1. Ajeya Cotra — Open Philanthropy (7:15)
At a panel at Dealbook in New York, the moderator asked whether we thought it was more likely than not that by 2030 we would get AGI… seven or eight hands went up… But then he asked a follow-up question… whether we thought that AI would create more jobs or destroy more jobs over the following 10 years… eight out of 10 people… thought that AI would create more jobs than it destroyed.
- Stat: At a Dealbook panel (late 2023), 7 out of 10 panelists believed AGI would arrive by 2030, while 8 out of 10 believed AI would be a net job creator over the next 10 years.
- Credibility: HIGH — Speaker cites a specific public event (Dealbook panel) with a specific metric and denominator, highlighting a key strategic disconnect in executive expectations.
- Topic tag:
08-radical-vs-tablestakes
2. Ajeya Cotra — Open Philanthropy (31:05)
Meter came out with an uplift RCT, which I think was the first of its kind… They had software developers split into two groups. One group was allowed to use AI, the other group was disallowed from using AI. And they studied how quickly those developers solved issues… it actually turned out that in this case, AI slowed down their performance.
- Credibility: HIGH — Speaker cites a specific, published study (RCT) from a named organization (Meter) with a clear methodology and a concrete, counter-intuitive result about productivity.
- Topic tag:
08-radical-vs-tablestakes
3. Ajeya Cotra — Open Philanthropy (43:15)
Various CEOs have said internally 90% of our lines of code are written by AIs… One thing I’m interested in is what fraction of pull requests to your internal codebase were mostly written by AI and mostly reviewed by AI. […] I’d be very interested in watching that number climb up because it’s an indication both of AI capabilities and of how much deference they’re giving to AIs.
- Credibility: MEDIUM — Speaker proposes a specific, advanced metric for tracking AI autonomy in enterprise settings, contrasting it with less meaningful existing claims.
- Topic tag:
06-security-frontier
4. Ajeya Cotra — Open Philanthropy (1:06:27)
Figuring out how to create a setup where we use control techniques and alignment techniques and interpretability… to get to the point where we feel good about relying on their outputs is a crucial step to figure out because it either bottlenecks our progress because we’re checking on everything all the time and slowing things down, or it doesn’t bottleneck our progress but we hand the AIs the power to take over.
- Credibility: MEDIUM — This is a clear articulation of the core strategic dilemma facing frontier labs: the trade-off between speed and control when using AI to build more advanced AI.
- Topic tag:
06-security-frontier
5. Ajeya Cotra — Open Philanthropy (2:09:52)
We ended up making $25 million of grants through that [RFP] and then another two to three million from the companion RFP, which was just a broader like all kinds of information from RCTs to surveys about AI’s impact on the world.
- Stat: $25 million in grants for AI agent benchmarks, plus $2-3 million for other evidence types, funded by Open Philanthropy in late 2023.
- Credibility: HIGH — Speaker provides a specific dollar figure for a grant-making initiative she led at her organization, focused on measuring frontier AI capabilities.
- Topic tag:
06-security-frontier
Extracted 2026-04-14T08:26:55 via scripts/podcast_mine.py (Gemini gemini-2.5-pro).