The condensed highlights of hours of AI lab leader interviews, model releases, Gemini 3 Flash insights (plus it’s hidden flaw), Hassabis’ ‘proto-AGI’ and much more…

https://matsprogram.org/apply?utm_source=ai-explained&utm_medium=youtube&utm_campaign=s26

Also, do check out my new app: https://lmcouncil.ai

Chapters:
00:00 - Introduction
00:50 - Results
02:44 - But… the Flaw
04:49 - So Benchmarks are fake? No
07:37 - Spatial Reasoning + Hassabis
10:06 - Proto-AGI
12:07 - Minimal AGI
15:07 - Compute Slowdown
17:56 - New Data Paradigm

Gemini 3 Flash: https://deepmind.google/models/gemini/flash/

Hassabis Interview: https://www.youtube.com/watch?v=PqVbypvxDto
Legg Interview: https://www.youtube.com/watch?v=l3u_FAv33G0
Pre-training Lead Interview: https://www.youtube.com/watch?v=cNGDAqFXvew
Altman Interview: https://www.youtube.com/watch?v=2P27Ef-LLuQ
Brockman Video: https://x.com/OpenAI/status/2001336514786017417
Post-Training Reveal: https://x.com/OfficialLoganK/status/2001742530472534442

Hallucinations Paper: https://cdn.openai.com/pdf/d04913be-3f6f-4d2b-b283-ff432ef4aaa5/why-language-models-hallucinate.pdf
Patreon Hallucinations Vid: https://www.patreon.com/posts/blockers-to-and-139264812
AA-Omniscience Benchmark: https://artificialanalysis.ai/evaluations/omniscience
https://arxiv.org/pdf/2511.13029


lmcouncil.ai/benchmarks
https://simple-bench.com/
https://x.com/scaling01/status/1999620587744813205

5.2 Codex Drop: https://cdn.openai.com/pdf/ac7c37ae-7f4c-4442-b741-2eabdeaf77e0/oai_5_2_Codex.pdf

OpenAI Compute Trend: https://www.theinformation.com/articles/openais-350-billion-computing-cost-problem?rc=sy0ihq

Cramer Tweet/Response: https://x.com/BorisMPower/status/2001440650210976018

OpenAI Valuation: ​​https://www.theinformation.com/articles/openai-discussed-raising-tens-billions-valuation-around-750-billion?rc=sy0ihq

Indian Data: https://www.reuters.com/world/india/with-freebies-openai-google-vie-indian-users-training-data-2025-12-17/

TheInformation Data: https://x.com/theinformation/status/2001421225751351778

Genie 3: https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/
Sima 2: https://deepmind.google/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/
Veo 3.1: https://deepmind.google/blog/sima-2-an-agent-that-plays-reasons-and-learns-with-you-in-virtual-3d-worlds/

METR: https://metr.org/blohttps://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/2025-03-19-measuring-ai-ability-to-complete-long-tasks/


AI Insiders ($9!): https://www.patreon.com/AIExplained


Non-hype Newsletter: https://signaltonoise.beehiiv.com/