RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
0:40
x.comZhengyao Jiang
RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
CLS (@ChengleiSi). 17 views. RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
1.1M views2 weeks ago
Python Tutorial
Learn Python - Full Course for Beginners [Tutorial]
4:26:52
Learn Python - Full Course for Beginners [Tutorial]
YouTubefreeCodeCamp.org
49M viewsJul 11, 2018
Python for Beginners – Full Course [Programming Tutorial]
4:40:00
Python for Beginners – Full Course [Programming Tutorial]
YouTubefreeCodeCamp.org
4.6M viewsAug 9, 2022
Python Full Course for Beginners | Complete All-in-One Tutorial | 9 Hours
8:38:56
Python Full Course for Beginners | Complete All-in-One Tutorial | 9 Hours
YouTubeDave Gray
1.1M viewsJul 25, 2023
Top videos
RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
0:40
RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
x.comZhengyao Jiang
279K views2 weeks ago
You can now run recursive language model (RLM) workflows in Deep Agents.Everything you need to know in 6 minutes from @sydneyrunkle.
6:36
You can now run recursive language model (RLM) workflows in Deep Agents.Everything you need to know in 6 minutes from @sydneyrunkle.
x.comLangChain
4.4K views3 weeks ago
How we trained it: RSI (recursive self-improvement).Each iteration, a training agent plans a target capability, trains the model, evaluates the graded benchmark trajectories, and data agent synthesizes the next data mixture from the failure modes it finds.Humans gate every keep-or-revert decision. Reverted rounds stay in the record. One raised headline scores while quietly damaging a held-out subset, so we threw it out. The kept rounds add up: +14.6 on HealthBench Hard, +15.9 on HealthBench Prof
0:11
How we trained it: RSI (recursive self-improvement).Each iteration, a training agent plans a target capability, trains the model, evaluates the graded benchmark trajectories, and data agent synthesizes the next data mixture from the failure modes it finds.Humans gate every keep-or-revert decision. Reverted rounds stay in the record. One raised headline scores while quietly damaging a held-out subset, so we threw it out. The kept rounds add up: +14.6 on HealthBench Hard, +15.9 on HealthBench Prof
x.comWeiran Yao
1K views2 weeks ago
Python Projects
3 Python Projects to Add in Your Resume (Beginner to Advanced + AI) | Sagar Chouksey
2:15:51
3 Python Projects to Add in Your Resume (Beginner to Advanced + AI) | Sagar Chouksey
YouTubeCoding With Sagar
72.8K views3 months ago
Watch me Do a Data Analyst Project in minutes with Python
33:47
Watch me Do a Data Analyst Project in minutes with Python
YouTubeLore So What
59.2K views4 months ago
Beginner to Advanced Job Ready Python Projects | Best Projects for Resume | Mini Projects in Python
1:21:07
Beginner to Advanced Job Ready Python Projects | Best Projects for Resume | Mini Projects in Python
YouTubeSaumya Singh
36.2K views6 months ago
RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
0:40
RT @zhengyaojiang: The first experimental evidence of recursive self-improvement (RSI). Autoresearching the autoresearch agent for eight…
279K views2 weeks ago
x.comZhengyao Jiang
You can now run recursive language model (RLM) workflows in Deep Agents.Everything you need to know in 6 minutes from @sydneyrunkle.
6:36
You can now run recursive language model (RLM) workflows in Deep Agents.Everything you need to know in 6 minutes from @sydneyrunkle.
4.4K views3 weeks ago
x.comLangChain
How we trained it: RSI (recursive self-improvement).Each iteration, a training agent plans a target capability, trains the model, evaluates the graded benchmark trajectories, and data agent synthesizes the next data mixture from the failure modes it finds.Humans gate every keep-or-revert decision. Reverted rounds stay in the record. One raised headline scores while quietly damaging a held-out subset, so we threw it out. The kept rounds add up: +14.6 on HealthBench Hard, +15.9 on HealthBench Prof
0:11
How we trained it: RSI (recursive self-improvement).Each iteration, a training agent plans a target capability, trains the model, evaluates the graded benchmark trajectories, and data agent synthesizes the next data mixture from the failure modes it finds.Humans gate every keep-or-revert decision. Reverted rounds stay in the record. One raised headline scores while quietly damaging a held-out subset, so we threw it out. The kept rounds add up: +14.6 on HealthBench Hard, +15.9 on HealthBench Prof
1K views2 weeks ago
x.comWeiran Yao
Another post about this as it’s crazy interesting - We are starting to enter a period in which progress can accelerate via self-improvement. Right now, researchers are manually modifying and improving AI agents, both the model (the "brain") and the surrounding software called the harness (the "operating system"). Weco's team has just demonstrated that an AI agent can do this autonomously - at least in relation to the harness. This is the recursive premise: Agent1 rewrites the harness of Agent2,
0:40
Another post about this as it’s crazy interesting - We are starting to enter a period in which progress can accelerate via self-improvement. Right now, researchers are manually modifying and improving AI agents, both the model (the "brain") and the surrounding software called the harness (the "operating system"). Weco's team has just demonstrated that an AI agent can do this autonomously - at least in relation to the harness. This is the recursive premise: Agent1 rewrites the harness of Agent2,
1.1M views2 weeks ago
x.comMark
Cursor 如何自动化 AI 研究并实现自我加速迭代来自 Cursor 团队 @leerob 在 AI Engineer @aiDotEngineer 的主题演讲「Recursive Model Improvement」，聚焦 Cursor 如何系统性地自动化 AI 研究流程，构建可快速迭代模型的系统，并分享了与 SpaceXAI 联合训练 Grok 4.5 的部分工作。核心框架：外循环 + 内循环简单公式“更多算力 → 更好模型”只是表象。实际存在两层循环：· 外循环：模型上线后收集真实用户反馈、在线指标、A/B 测试结果，再反哺下一轮训练。Cursor 的大量收入来自 agent 使用数据，这些交互数据本身就是高质量训练信号。· 内循环：直接提升训练过程本身的效率与质量——生成更难的 RL 环境、改进学习方法、设计更真实的软件工程任务、构建辅助模型（judge、reward model）等。Cursor 已在大规模训练模型约一年。Composer 2.5 成为产品内最受欢迎的模型，主要得益于更多 RL 环境、更激进的任务难度以及新训练方法。自动化研究与 Agent 系统瓶颈逐渐从
19:36
Cursor 如何自动化 AI 研究并实现自我加速迭代来自 Cursor 团队 @leerob 在 AI Engineer @aiDotEngineer 的主题演讲「Recursive Model Improvement」，聚焦 Cursor 如何系统性地自动化 AI 研究流程，构建可快速迭代模型的系统，并分享了与 SpaceXAI 联合训练 Grok 4.5 的部分工作。核心框架：外循环 + 内循环简单公式“更多算力 → 更好模型”只是表象。实际存在两层循环：· 外循环：模型上线后收集真实用户反馈、在线指标、A/B 测试结果，再反哺下一轮训练。Cursor 的大量收入来自 agent 使用数据，这些交互数据本身就是高质量训练信号。· 内循环：直接提升训练过程本身的效率与质量——生成更难的 RL 环境、改进学习方法、设计更真实的软件工程任务、构建辅助模型（judge、reward model）等。Cursor 已在大规模训练模型约一年。Composer 2.5 成为产品内最受欢迎的模型，主要得益于更多 RL 环境、更激进的任务难度以及新训练方法。自动化研究与 Agent 系统瓶颈逐渐从
316 views1 week ago
x.commeng shao
Excited to help deliver Grok 4.5, where we pushed RL scaling to the next level and reached top-tier agentic coding performance.What excites me even more: Grok is starting to help us improve itself — analyzing its own weaknesses, discussing and implementing new recipes. It's an early, primitive form of recursive self-improvement, but it's real and already in my daily workflow. With a strong coding agent laying the foundation, so many exciting explorations are ahead.It's amazing that we made this
0:55
Excited to help deliver Grok 4.5, where we pushed RL scaling to the next level and reached top-tier agentic coding performance.What excites me even more: Grok is starting to help us improve itself — analyzing its own weaknesses, discussing and implementing new recipes. It's an early, primitive form of recursive self-improvement, but it's real and already in my daily workflow. With a strong coding agent laying the foundation, so many exciting explorations are ahead.It's amazing that we made this
10.6M views2 weeks ago
x.comBoyuan Zheng
prediction: Auto-Research is going to look much more like Iterative Agentic Map Reduceie. Auto-Research should have more breadth search + fusionAuto-Research in its basic form is a depth first recursive search through experiment spaceLook at current results, try a tweak, get metrics, loopBut human scientists often internalize a hypothesis of why something isn’t working, and come up with several possible directions. They’re often bottlenecked by their time and ability to executeEx: agent is unabl
3:29
prediction: Auto-Research is going to look much more like Iterative Agentic Map Reduceie. Auto-Research should have more breadth search + fusionAuto-Research in its basic form is a depth first recursive search through experiment spaceLook at current results, try a tweak, get metrics, loopBut human scientists often internalize a hypothesis of why something isn’t working, and come up with several possible directions. They’re often bottlenecked by their time and ability to executeEx: agent is unabl
650.9K views3 weeks ago
x.comViv
0:39
Created a full scale model of the RMS Titanic with GPT 5.6 sol by recursive prompting.What i've found is these models are extremely capable but quite lazy, so once the primary generation is done I prompted it to improve the lazyly made parts of the ship with image referenes.I'm working on a really cracked method to fully automate this and push model generation to actuall usable high quality assets, will start sharing more info about that in the coming days.The whole model is reletively low poly
171 views1 week ago
x.comAtomic
4:03
Très bonne interview.Jeffrey Ladish (ex-Anthropic) évoque la + grande crainte de ses amis restés au sein de ces entreprises: le recursive self-improvement imminent."Les gens pensent que l'IA reste confinée au niveau humain par ses données d'entraînement, mais c'est déjà faux !"
151.1K views2 weeks ago
x.comFabien
1:30
Less Noise. More Discipline. Error Reduction. Paterns Recognition.
24 views1 week ago
YouTubeSlamCrab
0:13
A healthcare-focused model just beat frontier models on some of the hardest medical benchmarks.Cura 1T:• Beats GPT-5.5 on HealthBench Hard• Beats Claude Fable 5 on HealthBench Professional• Beats Opus 4.8 on AgentClinic• Beats Opus 4.8 on MedAgentBench-v2The interesting part?It wasn't trained conventionally.It was trained using recursive self-improvement (RSI):AI trains AI.AI finds failures.AI generates new training data.Humans approve or reject each iteration.This is interesting because we're s
1.8K views2 weeks ago
x.comEntelligence AI
2:15
Ahead of a dinner with a US senator, AI researcher Nate Soares (@So8res) was told: "Don't give them any of the crazy crap. You know, play it cool."His friends opened with the concern that someone could use AI to cause a pandemic.The senator: "Oh, that's what you're worried about? I'm worried about these companies making AIs that can make smarter AIs, that can make smarter AIs, leading to recursive self-improvement that could kill literally everybody on this planet. And I'm worried that this coul
4.9K views2 weeks ago
x.comControlAI
0:19
Koi Pond was a top grossing iOS app in 2008 that made over $1 million. Now, one prompt recreates something even more fun 🐠🐟🤯We are 6 to 12 months away from broadly recursive, self-improving systems. You define a goal, and the system iterates until requirements are met.Models like Fable 5, Grok 4.5, and GPT-5.6 Sol are crossing this threshold. We are merely seeing the initial signs. The first iteration. And it's already mind-bending.Genetic optimization and evolutionary techniques will soon be
305 views2 weeks ago
x.comFranz Bruckhoff
0:10
Good morning and have a great Thursday friend, let's keep grinding together, together stronger 💪For @CNPYNetwork, I do not think a creator leaderboard should be read only as a points game.At least, that is not how I approach it.I cannot claim to know the exact scoring logic behind every movement.And I do not think creators should pretend they know more than they do.But from watching how these systems usually behave, I read the logic through a simple combination:Topic fit.Attention.Consistency.T
332 views1 week ago
x.comAquilaNera
0:10
Good morning and happy Sunday Canopy fam 🫶For me, @CNPYNetwork also opens a bigger question about language.Not only code language.Human language.Because infrastructure can be strong, but if people cannot understand it, the story stays smaller than it should be.That is why I think building and explaining in your native language matters more than people admit.I am writing this in English because I want the message to travel wider.But the point still stands.Every community should feel allowed to e
266 views2 weeks ago
x.comAquilaNera
1:52
Why is a top seed investor still backing RL after everyone else moved on?On Nebius for Startups, the Striker VC partner argues RL generalizes better than any other training domain. Train a model on one RL task and it improves at everything else, because it learns to solve for reward functions - a transferable skill.That is the substrate for recursive self-improvement and continual learning - the two hardest problems left in AI.90% of VCs now know what RL is, and most are pivoting away. That is e
2.3K views3 weeks ago
x.comPodcast Alpha
15:04
Elon Musk was asked how America competes with Chinese manufacturing. He didn't reach for tariffs or policy. He gave a blunt concession: on humans, America has already lost."We definitely can't win with just humans."That is the reframe. And it changes what the US-China race is actually about.The conventional story is that America out-innovates while China out-manufactures, and clever policy keeps the balance. Musk dismantles that with arithmetic. China has four times the population. He argues the
10.8K views2 weeks ago
x.comVikram M
See more
Static thumbnail place holder