Do we have a new best AI model, or do we have the downfall of benchmarks in general, as a way of capturing machine intelligence? Full breakdown of Gemini 3.1 Pro, guest-starring the new Sonnet 4.6, plus analysis from 7 papers/posts that will give you much needed context. Oh, and a new record on Simple Bench!

https://epoch.ai/ai-explained-datacenters


Check out my fast-growing (!) app, free to use, and code INSIDER15 for Pro: https://lmcouncil.ai

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


Chapters:
00:00 - Introduction
00:30 - Post-training Dominance
04:00 - ARC-AGI 2 Caveat
05:54 - Simple Bench Record
08:22 - Hallucination Caveat
10:05 - Model Card
11:12 - Exponential Coming
12:20 - Amodei on Generalizing
15:10 - One True Benchmark?
17:02 - Other Metrics…

Gemini 3.1 Model Card: https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-3-1-Pro-Model-Card.pdf

Release: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/

Where are Agents deployed?: https://www.anthropic.com/research/measuring-agent-autonomy

Newsletter Post: https://signaltonoise.beehiiv.com/p/4-ai-numbers-that-surprised-me-this-week

Hallucination AA: https://artificialanalysis.ai/evaluations/omniscience

Melanie Mitchell: https://x.com/MelMitchell1/status/2022738363548340526
ARC-AGI-2: https://x.com/arcprize/status/2024522812728496470/photo/1

Chollet on Agentic Coding and ML: https://x.com/fchollet/status/2024519439140737442

METR Caveat: https://metr.org/notes/2026-01-22-time-horizon-limitations/

Talaas Fast: https://chatjimmy.ai/

Amodei Interview Continual learning: https://www.dwarkesh.com/p/dario-amodei-2?open=false#%C2%A7002942-is-continual-learning-necessary-how-will-it-be-solved

Metaculus FutureEval: https://www.metaculus.com/futureeval/

Next Vid to Watch: https://www.patreon.com/posts/what-you-need-to-150647292



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

Podcast: https://aiexplainedopodcast.buzzsprout.com/

The two models that you will hear discussed for at least the next two months - Claude Opus 4.6 and GPT 5.3 Codex - just got released within 26 mins or each other. The full breakdown of around 250 pages of reports, with just the most interest moments, from the battle of which is best, Claude personhood, the surprising misbehaviour of Opus 4.6, and much more

https://assemblyai.com/aiexplained

Check out my fast-growing (!) app, free to use, and code INSIDER15 for Pro: https://lmcouncil.ai

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

Chapters:
00:00 - Introduction
00:54 - Self-improvement?
02:44 - Knowledge Work
05:30 - Overly agentic behaviour
09:12 - Who Shouldn’t Use Claude Opus
11:39 - Step-change?
15:09 - Claude’s ‘Personhood’

Hassabis Roadmap: https://www.patreon.com/posts/hassabis-roadmap-149750869

Release of Opus 4.6: https://www.anthropic.com/news/claude-opus-4-6
212 Page System Card: https://www-cdn.anthropic.com/0dd865075ad3132672ee0ab40b05a53f14cf5288.pdf
Claude Code Tip: https://x.com/bcherny/status/2019475897691124107


GPT Codex 5.3: https://openai.com/index/introducing-gpt-5-3-codex/

System Card: https://openai.com/index/gpt-5-3-codex-system-card/

Browse Comp: https://arxiv.org/pdf/2504.12516v1
Finance Agent: https://www.vals.ai/benchmarks/finance_agent
Terminal Bench 2: https://arxiv.org/pdf/2601.11868
Vending Bench: https://andonlabs.com/blog/opus-4-6-vending-bench

My X post: https://x.com/AIExplainedYT/status/2016851303436095647

Anthropic Apology: https://x.com/ch402/status/2014066134194995256/photo/1

Altman rebuttal: https://x.com/sama/status/2019139174339928189
https://x.com/sama/status/2019140276246442089

4% of GitHub: https://x.com/dylan522p/status/2019490550911766763



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

Podcast: https://aiexplainedopodcast.buzzsprout.com/

Anthropic's CEO, who has consistently predicted transformative AI will arrive before 2030, recently published a nearly 20,000-word essay outlining his vision of where AI is heading. The video gives you the highlights. The essay argues that scaling and recursion will advance AI from coding automation to full engineering automation, while warning of economic displacement within 1-2 years and China's trajectory toward AI-enabled totalitarianism. Additionally, Dario Amodei predicts that AI models will increasingly be understood as collections of distinct personas rather than monolithic systems.

80,000 Hours: https://www.youtube.com/watch?v=B54EQiuO1UU



Check out my fast-growing (!) app, free to use, and code INSIDER15 for Pro: https://lmcouncil.ai

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


Chapters:
00:00 - Introduction
01:10 - Scaling to software engineers
06:11 - Permanent Underclass
10:18 - Totalitarian Nightmares
16:38 - Collection of Personas

Essay: https://www.darioamodei.com/essay/the-adolescence-of-technology

Physics Prediction: https://www.quantamagazine.org/is-particle-physics-dead-dying-or-just-hard-20260126/

Axios: https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic

World GDP: https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?end=2024&start=1961&view=chart

Demis Hassabis Counter: https://www.youtube.com/watch?v=q6fq4_uP7aM

Karpathy 80%: https://x.com/karpathy/status/2015883857489522876

Machines of Loving Grace: https://www.darioamodei.com/essay/machines-of-loving-grace

Anthropic LessWrong: https://www.lesswrong.com/posts/5aKRshJzhojqfbRyo/unless-its-governance-changes-anthropic-is-untrustworthy#1__In_private__Dario_frequently_said_he_won_t_push_the_frontier_of_AI_capabilities__later__Anthropic_pushed_the_frontier

Original Constitution: https://www.anthropic.com/news/claudes-constitution

New Constitution: https://www.anthropic.com/constitution

Kimi K2.5: https://x.com/Kimi_Moonshot/status/2016024049869324599

Societies of Thought, Google DeepMind Paper: https://arxiv.org/pdf/2601.10825

https://lmcouncil.ai/benchmarks

https://www.patreon.com/posts/our-new-age-of-133960279



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

Podcast: https://aiexplainedopodcast.buzzsprout.com/