AlphaEvolve is not the first system to exhibit self-improvement, but it may be the most impressive yet. AI is literally improving the hardware, architectures, data and training methods of AI itself. A deep dive into the paper, drawing on two previous interviews and 5 other papers. Plus a snippet on OpenAI’s new Codex system.

Gray Swan: http://app.grayswan.ai/ai-explained

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

Chapters:
00:00 - Introduction
00:27 - AlphaEvolve
05:23 - Limitation
06:10 - Achievements
08:21 - Future Improvements
13:30 - Quirks
16:34 - Final Thoughts

AlphaEvolve release: https://deepmind.google/discover/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/

Paper: https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/AlphaEvolve.pdf

Terence Tao Quote: https://mathstodon.xyz/@tao/114508029896631083

Nature Article: https://www.nature.com/articles/s41586-022-05172-4
MIT Article: https://www.technologyreview.com/2025/05/14/1116438/google-deepminds-new-ai-uses-large-language-models-to-crack-real-world-problems/
AI Co-Scientist: https://arxiv.org/pdf/2502.18864

OpenAI Codex: https://openai.com/index/introducing-codex/


70% of Pull Requests: https://x.com/slow_developer/status/1920920456393028027

Amodei Essay: https://www.darioamodei.com/essay/machines-of-loving-grace

OpenAI Jason Wei Tweet: https://x.com/_jasonwei/status/1923091260354531612

PromptBreeder: https://arxiv.org/pdf/2309.16797
DrEureka: https://arxiv.org/pdf/2406.01967

FT DeepMind: https://www.ft.com/content/4e497a91-670a-4f69-be4a-18e247daba3e



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

A green card, o3 vs Gemini 2.5, 6 Benchmarks and a whole bunch of my thoughts on what on earth is happening in AI, from here to 2030. Plus, how AI is becoming pay-to-win, and why. Crazy times, 14 mins probably wasn’t enough.

https://app.grayswan.ai/ai-explained

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

Chapters:
00:00 - Introduction
00:33 - FictionLiveBench
01:37 - PHYBench
02:14 - SimpleBench
02:54 - Virology Capabilities Test
03:13 - Mathematics Performance
04:29 - Vision Benchmarks
05:43 - V* and how o3 works
06:44 - Revenue and costs for you
08:54 - Expensive RL and trade-offs
09:40 - How to spend the OOMs
13:27 - Gray Swan Arena

Green Card: https://techcrunch.com/2025/04/25/an-openai-researcher-who-worked-on-gpt-4-5-had-their-green-card-denied/
PHYBench: https://arxiv.org/pdf/2504.16074Virologytest: https://www.virologytest.ai/
How o3 Vision Works: https://arxiv.org/pdf/2312.14135 https://x.com/sainingxie/status/1912570624523829573
Visual puzzles: https://neulab.github.io/VisualPuzzles/
Fiction Bench: https://x.com/ficlive/status/1912863028141244850
https://geobench.org/
https://simple-bench.com/
AIME 2025: https://openai.com/index/introducing-o3-and-o4-mini/
USAMO: https://x.com/mbalunovic/status/1914398518896193747
NaturalBench: https://linzhiqiu.github.io/papers/naturalbench/
Where’s Waldo: https://uk.pinterest.com/pin/492792384225896298/
IMO and AlphaProof:https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/
Crazy Revenue: https://www.theinformation.com/articles/openai-forecasts-revenue-topping-125-billion-2029-agents-new-products-gain?rc=sy0ihq
Number of Users: https://www.theinformation.com/briefings/googles-gemini-user-numbers-revealed-court?rc=sy0ihq
Subscriptions pay to win: https://www.forbes.com/sites/paulmonckton/2025/04/23/google-leak-reveals-new-gemini-ai-subscription-levels/
GPU Trade-offs: https://x.com/sama/status/1915098951067554030
RL Scale-up Amodei: https://www.darioamodei.com/post/on-deepseek-and-export-controls
Log-linear Returns: https://x.com/bobmcgrewai/status/1895228291981943265
2030 Scaling: https://epoch.ai/blog/can-ai-scaling-continue-through-2030
Model Size: https://x.com/slow_developer/status/1874554473256997201
Adam on AGI: https://x.com/TheRealAdamG/status/1913998366632968381
Papers on Patreon: https://arxiv.org/pdf/2502.01839
https://arxiv.org/pdf/2504.13837
Chollet Quote: https://x.com/fchollet/status/1912934762580447447
OpenSim: https://opensim.stanford.edu/


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

Critical analysis of the two most powerful new models behind ChatGPT, o3 and o4-mini. Not just the system cards, benchmarks, and my own tests, but some you may not have seen before. Yes, they can whip up amazing front-end in a few seconds, but you always have to ask what is in their data. Either way, they prove the gains from RL are just beginning…

https://weave-docs.wandb.ai/?utm_source=sponsorship&utm_medium=simple_bench&utm_campaign=ai_explained

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


Chapters:
00:00 - o3 and o4-mini


https://simple-bench.com/

Plus, Teams and Pro, plus token count: https://x.com/btibor91/status/1912568994512662679

System Card: https://openai.com/index/o3-o4-mini-system-card/

Release Notes: https://openai.com/index/introducing-o3-and-o4-mini/

https://deepmind.google/technologies/gemini/pro/

https://x.com/DeryaTR_/status/1912558350794961168

https://x.com/polynoamial/status/1912564068168450396

API Pricing:https://openai.com/api/pricing/

https://aider.chat/docs/leaderboards/


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

This pod won’t just be about the release of GPT 4.1 in the last 48 hours, o3 build-up, Kling 2.0, a sneak-peak at the next OpenAI model, or even the new Dolphin language tool. It will be about 7 such stories that contextualise where we are in AI and what is happening.

https://www.emergentmind.com/

Chapters:

00:00 - Introduction

00:30 - Kling 2.0

01:35 - GPT 4.1

05:25 - o3 Build-up

07:37 - ‘Product Company’

09:31 - Safe Superintelligence

10:54 - DolphinGemma

13:16 - Data Dominance?

Kling 2.0: https://app.klingai.com/global/release-notes

Dolphin Gemma: https://blog.google/technology/ai/dolphingemma/?s=09

https://openai.com/index/gpt-4-1/

OpenAI o3 Build-up The Information: https://www.theinformation.com/articles/openais-latest-breakthrough-ai-comes-new-ideas?rc=sy0ihq

Physical reasoning: https://x.com/a_karvonen/status/1911839968990814503

Fiction Live.bench: https://x.com/ficlive/status/1911853409847906626

Altman Ted: https://www.youtube.com/watch?v=5MWT_doo68k

https://simple-bench.com/try-yourself

https://aider.chat/docs/leaderboards/

4.5: https://www.youtube.com/watch?v=6nJZopACRuQ

Geospatial reasoning: https://research.google/blog/geospatial-reasoning-unlocking-insights-with-generative-ai-and-multiple-foundation-models/

Pioneers: https://x.com/OpenAIDevs/status/1910017976256119151

Evals: https://www.youtube.com/watch?v=scsW6_2SPC4

Anthropic Updates: https://www.bloomberg.com/news/articles/2025-04-15/anthropic-is-readying-a-voice-assistant-feature-to-rival-openai?srnd=phx-ai

https://x.com/sethsaler/status/1912188383457059301

https://techcrunch.com/2025/04/12/openai-co-founder-ilya-sutskevers-safe-superintelligence-reportedly-valued-at-32b/

https://ai.meta.com/blog/llama-4-multimodal-intelligence/

https://deepmind.google/technologies/gemini/pro/

https://research.google/blog/accelerating-scientific-breakthroughs-with-an-ai-co-scientist/

https://blog.google/products/google-cloud/ironwood-tpu-age-of-inference/

OpenAI Documentary: https://www.patreon.com/posts/one-machine-to-121940490

The latest on Llama 4, and whether it signals a slowdown in AI, or solid progress. Plus, a deep dive on that viral prediction of superintelligence by 2027, and Amodei’s cautionary words on what could stop AI progress in its tracks. o3 news, and more, as well.

Weights & Biases: https://weave-docs.wandb.ai/?utm_source=sponsorship&utm_medium=simple_bench&utm_campaign=ai_explained


DeepSeek Doc: https://www.patreon.com/posts/openai-is-not-r1-125869969

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

Chapters:
00:00 - Introduction
00:47 - Stock Crash
02:28 - Llama 4
10:55 - o3 News
11:59 - OpenAI non-profit?
13:13 - AI 2027

Llama 4 Release: https://ai.meta.com/blog/llama-4-multimodal-intelligence/

Dario Amodei Comments: https://www.youtube.com/watch?v=esCSpbDPJik

Knowledge Cut-off: https://www.llama.com/docs/model-cards-and-prompt-formats/llama4_omni/

Aider Polyglot: https://aider.chat/docs/leaderboards/

Gemini 1.5: https://arxiv.org/pdf/2403.05530

Fiction-LiveBench: https://fiction.live/stories/Fiction-liveBench-Mar-25-2025/oQdzQvKHw8JyXbN87

OpenAI Valuation: https://www.nytimes.com/2025/03/31/technology/openai-valuation-300-billion.html?login=smartlock&auth=login-smartlock

OpenAI Cybersecurity: https://www.bloomberg.com/news/articles/2024-01-16/openai-working-with-us-military-on-cybersecurity-tools-for-veterans

Deep research System Card: https://cdn.openai.com/deep-research-system-card.pdf

https://openai.com/index/paperbench/

AI 2027: https://ai-2027.com/

METR Paper: https://arxiv.org/pdf/2503.14499

OpenAI non-profit: https://openai.com/index/nonprofit-commission-guidance/

NYT Piece: https://www.nytimes.com/2025/04/03/technology/ai-futures-project-ai-2027.html?unlocked_article_code=1.804._yKi.QhwOp15Q3tcU&smid=url-share&s=09

Kokotajlo predictions 2021: https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/what-2026-looks-like

https://simple-bench.com/


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

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

Gemini gets a new record on Simple Bench, and several other benchmarks. I’ll go deep to explore its nuances, including how it deceptively reverse engineers answers, does better on certain coding benchmarks than others, may have a universal ‘conceptual language’ …

https://weave-docs.wandb.ai/?utm_source=sponsorship&utm_medium=simple_bench&utm_campaign=ai_explained

… and more. Plus practical tips, a note on security and Kling vs Veo 2 guest appearance.


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

Chapters:
00:00 - Introduction
00:36 - Fiction Bench
02:41 - Practicality - YouTube urls + Security - cut-off date
03:42 - Coding
06:22 - WeirdML Bench
07:01 - Simple Bench Record High
11:23 - Reverse Engineering!
13:22 - Anthropic Paper
17:49 - 3 Caveats

Gemini 2.5 Updated: https://deepmind.google/technologies/gemini/

Fiction Live Bench: https://fiction.live/stories/Fiction-liveBench-Feb-19-2025/oQdzQvKHw8JyXbN87

https://simple-bench.com/

WeirdML: https://htihle.github.io/weirdml.html
https://x.com/htihle/status/1905014058228625542

Anthropic Thoughts: https://www.anthropic.com/research/tracing-thoughts-language-model
https://transformer-circuits.pub/2025/attribution-graphs/biology.html#dives-cot

https://aistudio.google.com/prompts/new_chat

Search Study: https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php

Live bench: https://livebench.ai/#/
Paper: https://arxiv.org/pdf/2406.19314

LiveCode Bench: https://livecodebench.github.io/

SWE-Verified: https://arxiv.org/pdf/2310.06770


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

Gemini 2.5 is out, on the same day as the new DeepSeek V3 (which should power Deepseek R2). Do both models prove AI is being commoditized? Let’s find out, on this blockbuster day of AI releases. Plus exclusives from the Information, Simple indications, Vista Bench, LM Arena and more…

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

Chapters:
00:00 - Introduction
01:15 - Gemini 2.5 Benchmarks
05:46 - Long Context, Simple indication
07:08 - New Deepseek V3 -024
09:11 - Microsoft MAI
11:48 - 90% of code but new Claude jobs

‘World’s most powerful model’: https://x.com/OfficialLoganK/status/1904580368432586975

Gemini 2.5 Release Notes: https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/#gemini-2-5-thinking

‘Commoditized’: https://the-decoder.com/microsoft-ceo-satya-nadella-says-ai-models-are-getting-commoditized/

Microsoft Information report: https://www.theinformation.com/articles/microsofts-ai-guru-wants-independence-from-openai-thats-easier-said-than-done?rc=sy0ihq

LMarena: https://x.com/lmarena_ai/status/1904581128746656099/photo/1

Free for now: https://x.com/btibor91/status/1904578053537476628

Vista Bench:https://scale.com/leaderboard/visual_language_understanding

DeepSeek V3: https://huggingface.co/deepseek-ai/DeepSeek-V3-0324

Claude Plays Pokemon: https://www.twitch.tv/claudeplayspokemon
Amodei: 100% Coding: https://www.youtube.com/watch?v=esCSpbDPJik&t=3017s

Anthropic Jobs: https://job-boards.greenhouse.io/anthropic/jobs/4020717008

Microsoft Money from Onslaught: https://www.972mag.com/microsoft-azure-openai-israeli-army-cloud/

https://simple-bench.com/

Release Date Comments: https://x.com/zacharynado/status/1904647277861318979


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

Is Manus AI the memecoin of the AI world, or legit? I’ll compare it to OpenAI’s Deep Research, Operator, Grok 3 DeepSearch and more to find out. I’ll also let you in on some of the secrets of what makes a good hype campaign, the estimated costs of Manus AI, and where it is strong. Other news (yes, Gemini image editing and research hacking, I mean you), will have to wait for a few more hours, as millions enquire about Manus AI.

https://app.grayswan.ai/arena

AI Insiders ($9!): https://www.patreon.com/AIExplained
Patreon Vid: https://www.patreon.com/posts/4-ai-trends-in-123857767

Chapters:
00:00 - Introduction
00:46 - Hype Campaign
02:40 - Single, Public Benchmark
03:12 - What is Manus AI?
04:22 - Test 1
05:12 - Cost and Rate Limits
06:15 - Test 2 vs Deep Research + Grok 3 DeepSearch
08:24 - Test 3 (not AGI)
11:10 - 4 Trends in AI in 2025
11:37 - Hype Works

Manus AI: https://manus.im/app

Xiao Hong Interview: https://www.chinatalk.media/p/manus-chinas-latest-ai-sensation

Gaia Benchmark: https://openreview.net/pdf?id=fibxvahvs3
MIT Report: https://www.technologyreview.com/2025/03/11/1113133/manus-ai-review/

Information Report: https://www.theinformation.com/articles/anthropics-claude-drives-strong-revenue-growth-while-powering-manus-sensation?rc=sy0ihq

Hype Examples: https://x.com/Saboo_Shubham_/status/1898425707401031940
https://x.com/EHuanglu/status/1899110687902978373
https://x.com/AJs_AI/status/1898756132384178291

Mistakes: https://x.com/TheXeophon/status/1898737178273829220

Tools and Code: https://x.com/peakji/status/1898994802194346408

https://operator.chatgpt.com/




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

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