I’m training a machine learning model on 10 years of stock data to predict how certain tickers might move the next day. It’s built on a Random Forest architecture and filters signals using momentum, volume, and volatility features. Follow along for the journey :)

🚨 URGENT Two days ago I was contacted by a high-ranking employee of the French Government. After determining this person’s position and proximity to the French couple, I have deemed the information they gave me to be credible enough to share publicly in the event that something happens. In short, this person claims that the Macrons have executed upon and paid for my assassination. Yes, you read that correctly. More specifically, that the green light was given to a small team in National Gendamarie Intervention Group. I am told there is one Israeli that is on this assasination squad and the plans were formalized. Again, this person provided concrete proof that they are well placed within the French government apparatus. Further to this point, this person claims that Charlie Kirk’s assassin trained with the French legion 13th brigade with multi-state involvement. Journalist Xavier Poussard’s life is also at risk. This is deadly serious. The head of state of France apparently wants us both dead and has authorized professional units to carry this out. I ask that every person RETWEET and share this. I do not know who in the American government can be trusted, since this source claims our leaders are aware. But I have more specific information which is definitively verifiable, should they care to reach out to me. To the brave official in France who did this because they were so moved by the evil of Charlie’s public execution to risk their own life— May God bless you. Truly. Let all be revealed.

Fantasy league for AIs: LLMs paper-trade crypto on live prices

Choosing K in K-Means Clustering

AI This Week Cancer Breakthrough and ChatGPT Update

sora 2 is mind blowing #Sora2 #AIvideo #GenAI #OpenAI #VFX

AI tools are no longer just cool experiments. They are becoming the backbone of productivity for students, creators, and professionals. The top 5 AIs I’m sharing in this video are designed to save time, cut friction, and keep you focused on what matters most. Each one solves a specific pain point so you’re not just working harder, but actually working smarter. Think of it as your ultimate productivity toolkit. You’ll find apps that help you write faster, manage tasks automatically, stay on top of schedules, generate new ideas, and even keep your inbox under control. #ai #producivity #productivitytips #worksmarter

Sora 2 Just Changed AI Video Forever

Read more: leraandronova.substack.com/p…

AI week in 2 minutes. NVIDIA intends to invest up to $100B so OpenAI can build 10 GW of supercomputers. First systems expected in 2026. DeepMind shows “thinking” robots. New Gemini Robotics models plan multi-step tasks, call tools, and explain their reasoning. Meta drops Vibes. A feed of AI-generated shorts you can create and remix, with cross-posting to Instagram and Facebook. Alibaba unveils Qwen 3-Max. A 1-trillion-parameter model. Hong Kong shares jumped about 10% after the news. Which story should I deep dive next? #AInews #OpenAI #NVIDIA #DeepMind #GenerativeAI

Leakage makes your accuracy lie. Fix it by splitting like reality (by time or by user), removing duplicates, building features without future info, fitting preprocessing on train only, and freezing the test set. Save this for your next ML model. #machinelearning #datascience #mltips #ai #dataleakage #traintestsplit

AI just wrote working viral genomes and scientists built 16 real phages that killed antibiotic-resistant E. coli. Breakthrough medicine or a biosecurity warning. Sources: Nature and bioRxiv. #ai #biology #phage #superbugs #biosecurity #stanford #research #ecoli

Great ML channels hit three pillars: -Theory (math intuition behind models) -Practice (end-to-end code that runs) -Research (weekly breakdowns of new papers). Follow creators who cover all three so you avoid tutorial hell, build real projects, and stay ahead of the curve. #machinelearning #deeplearning #mlengineering #mlresearch #aicommunity

Here’s your weekly AI news (9/15–9/21) AI just wrote viruses that kill superbugs: scientists built them in the lab and they worked. Meanwhile, Nvidia is investing $5B in Intel, Oracle is in talks with Meta for a $20B cloud deal, Microsoft is building the “world’s most powerful” AI datacenter, and the US-UK just signed a $42B tech pact. DeepSeek shook the field again by revealing its flagship model only cost $294K to train. #ai #ainews #technews #machinelearning #bigtech

Overfitting is one of the most common headaches in machine learning. It’s when your model looks perfect in training but completely falls apart on new data. Basically, it memorizes the answers instead of learning the patterns. Here are some things to keep in mind if your model is “acing the homework but failing the exam”: 1️⃣ More Data = Better Generalization → The more diverse your dataset, the harder it is for the model to just memorize. Data augmentation (rotating images, adding noise, paraphrasing text) can help when you can’t collect more samples. 2️⃣ Regularization is Your Friend → Penalties like L1/L2 or dropout stop the model from clinging to tiny details. Think of it as forcing the model to stay humble. 3️⃣ Right-Sized Models Win → Bigger isn’t always better. A smaller, simpler model can often outperform a massive one when data is limited. 4️⃣ Cross-Validation Matters → Don’t trust a single split. Rotate your data across multiple folds to see how well your model really generalizes. 5️⃣ Clean Data = Stronger Models → Outliers, missing values, and messy scaling trick models into learning shortcuts. Clean, balanced data builds trustable results. #machinelearning #datascience #deeplearning #ai #mltips

Where is the Love? 🎶

Gemini 3’s Flash tier might beat 2.5 Pro. That’s huge. Drop your pick for 2026: Gemini or GPT. #Gemini3 #AInews #Multimodal #GoogleDeepMind #GPT

If you want to code smarter in 2025, these are the tools to master. GitHub Copilot helps you write code faster with real-time AI suggestions. Postman makes testing and debugging APIs simple and efficient. Docker ensures your applications run the same across every environment, no matter the machine. OpenAI Codex and other coding AIs can generate functions, explain errors, and speed up development, though you may run into usage limits. Kaggle gives you access to thousands of datasets and built-in notebooks so you can practice data cleaning, analysis, and machine learning on real-world problems. Together, these tools cover the full workflow from writing and testing code to deploying and experimenting with data. #Coding #Developer #AI #TechTools #Programming