Get involved
AGI Strategy Reading Group
Applications for Fall 2026 are closed. Follow us on Instagram to hear when the next cohort opens. Follow @alignmentmcgill (opens in a new tab)
Sessions
Readings by session
The required readings from BlueDot Impact's AGI Strategy course (opens another website). About three hours a week. Do them before the session; the discussion assumes you have.
Session 1The Race to AGI
About 100 min of reading · Open the unit on BlueDot (opens another website)
What’s at stake?
- Personal Statement on AI Risk (opens another website)Daniel Selsam20 min
Competition and coordination problems
- Racing through a minefield: the AI deployment problem (opens another website)Holden Karnofsky20 min
- The Scramble: getting in position to pace the frontier (opens another website)Peter Wildeford25 min
Responding to the Hugging Face incident
- The Rise and Fall of Agent Civilizations (opens another website)Dwarkesh Patel35 min
Session 2Drivers of AI Progress
About 190 min of reading · Open the unit on BlueDot (opens another website)
What drives AI progress?
- The case for AGI by 2030 (opens another website)Ben Todd35 min
- What will AI look like in 2030? (opens another website)David Owen15 min
What can increasingly capable AI systems do?
- Trends in Artificial Intelligence (opens another website)Epoch AI10 min
- Epoch Capabilities Index (opens another website)15 min
- Benchmarks — Epoch AI (opens another website)Epoch AI10 min
- Frontier Risk Report (February to March 2026) (opens another website)METR20 min
- Measurements for understanding the pace of AI development inside frontier labs (opens another website)Marina Favaro and Phillie Wright (Anthropic)20 min
How could progress accelerate from here?
- AI 2027 (opens another website)Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, Romeo Dean35 min
- Q2.5 2026 Timelines Update: Uplift and Revenue (opens another website)Eli Lifland, Daniel Kokotajlo, Brendan Halstead20 min
- The Economics of Recursive Self-Improvement (opens another website)Tom Cunningham, Lukas Althoff, Basil Halperin, Brian Jabarian, Andrew Koh, Arjun Ramani, Phil Trammell, Parker Whitfill, and Cheryl Wu10 min
Session 3Pathways to Harm
About 210 min of reading · Open the unit on BlueDot (opens another website)
Loss of Control
- AI Could Defeat All Of Us Combined (opens another website)Holden Karnofsky20 min
- The Loss of Control Playbook: Degrees, Dynamics, and Preparedness (opens another website)Charlotte Stix, Annika Hallensleben, Alejandro Ortega, Matteo Pistillo (Apollo Research)60 min
- What failure looks like (opens another website)Paul Christiano20 min
- Why alignment could be hard with modern deep learning (opens another website)Ajeya Cotra25 min
Authoritarianism
- Extreme Power Concentration (opens another website)Hadshar35 min
Biological Misuse
- On Civilization-Threatening Bioweapons (opens another website)Kevin Esvelt30 min
- Building a Defense-in-Depth Biosecurity Strategy for the AI Era (opens another website)Steph Guerra, Aurelia Attal-Juncqua, John P. Tarangelo, Casey Aveggio, Katie Dammer, David Glickstein20 min
Session 4Strategies for the AGI Transition
About 245 min of reading · Open the unit on BlueDot (opens another website)
Orienting to the strategy
- AI 2040: Summary of Plan A (opens another website)Thomas Larsen; Romeo Dean; Brendan Halstead; Eli Lifland; Ryan Greenblatt; Daniel Kokotajlo10 min
The technical challenge
- An Approach to Technical AGI Safety and Security (opens another website)Rohin Shah et al.55 min
- Alignment remains a hard, unsolved problem (opens another website)Evan Hubinger30 min
- Teaching Claude why (opens another website)Anthropic15 min
Following the transition
- AI 2040: Plan A (opens another website)Thomas Larsen, Romeo Dean, Brendan Halstead, Eli Lifland, Ryan Greenblatt, Daniel Kokotajlo60 min
Pacing and alternative strategies
- AI 2040: Capability Scaling Strategy (opens another website)Eli Lifland20 min
- AI 2040: Comparing Possible Plans (opens another website)Eli Lifland15 min
- Pacing the Frontier (opens another website)Employees of frontier AI companies5 min
Examining the objections
- Plan A's problem with dry tinder (opens another website)Tom Davidson15 min
- Selective Optimism: a critique of AI 2040 (opens another website)Richard Ngo20 min
Session 5Start Contributing
About 150 min of reading · Open the unit on BlueDot (opens another website)
Choose your focus
- Become a person who Actually Does Things (opens another website)Neel Nanda5 min
- Project Tailwind: Initiatives (opens another website)Coefficient Giving15 min
Make a plan
- Personal Action Plan Template (opens another website)Dewi Erwan10 min
Create your 1-pager (optional)
- 1-Pager Template (opens another website)BlueDot Impact
Optional advice
- AGI safety career advice (opens another website)Richard Ngo20 min
- Career Review: AI Governance and Coordination (opens another website)Cody Fenwick50 min
- Career Review: Technical AI Safety (opens another website)80000 Hours50 min
After the reading group
Everyone who completes the reading group joins our alumni.
See our alumni