AI interviews and creator videos, organized around the questions that matter to you. Read the takeaways. Watch the original.
Key takeaways
Key takeaways
Key takeaways
Key takeaways
No matching content. Try another topic or clear the filters.
Videos play here when you click. If a publisher limits playback, use the original link.
One person can have more than one question. Switch whenever you need.
Start with the task, then explore the forecast.
Al-Dahle describes choosing models using application-specific evaluations of performance, latency and cost.
Al-Dahle describes Airbnb teams moving from sequential handoffs to collaborating on working prototypes.
Weinstein emphasizes agents retrying and debugging after problems, beyond merely starting a task.
Handa describes a decision interface for classification and routing inside applications.
Shihipar discusses software that users can continually reshape with agents.
Shihipar stresses understanding the problem and communicating requirements to coding agents.
Huang argues that agentic systems should operate with restrictive rights.
Huang frames AI safety as an engineering problem that can be addressed through system design.
Graham describes career change as involving identity, loss and relationships as well as new opportunities.
Graham distinguishes delegating to AI from developing people, emphasizing human judgment and direction.
Germanidis emphasizes references, steerability and iteration within creative workflows.
The discussion highlights distribution and access as important parts of the AI market.
Germanidis discusses real-time video and world models as a direction for interactive experiences.
Atallah and Midha challenge the idea of a single model winning every use case.
Platt emphasizes domain knowledge and the practice involved in developing scientific judgment.
Platt discusses the importance of choosing objectives that reflect useful scientific progress.
Almeida argues that AI can strengthen existing software by becoming a small, programmable part of it.
Brown explores parallel agent work and the challenges of coordinating many contributors.
Brown considers how more capable models might organize increasingly large groups of agents.
Kvist argues that trust and accountability can constrain adoption even when AI capabilities improve.
Khan highlights communication and collaboration as skills that education should develop alongside technical work.
Khan discusses proactive feedback instead of relying only on learners to ask a chatbot questions.
Kvist emphasizes testing failure cases beyond a successful demonstration.
Belsky discusses learning through making solutions to concrete problems.
Belsky’s conversation examines motivation, accountability and human relationships in AI-supported education.
The panel debates whether research and engineering bottlenecks could constrain recursive AI improvement.
The discussion examines the difficulty of transferring simulated success to complex real-world tasks.
Amodei expects very powerful AI within a few years and discusses how capability may spread through the economy.
Karpathy describes agents as a decade-long engineering effort, pointing to gaps including continual learning.
Karpathy contrasts answering questions with a tutor who understands a learner and chooses an appropriate challenge.
No matching views yet. Try another question or clear your filters.
Your saved views and plans, kept on this device.
Save a view, then record one action you want to try.
When you want to keep the work
Reading, video access, local saves and exports are free. Use the same plan on another device with AGI’s existing cloud workspace and version history.
No paid interview collection or automated personal-report subscription is offered here.
9 USDT / 30 days
Plus the order-matching decimal and network fees. 50 workspaces, 10 versions per workspace, 64 KB per workspace and 5 MB in total. No auto-renewal or extra AI allowance.