Towards AI, Inc.

Over last few months I was involved as a reviewer for the course "Agentic AI Engineering" helping Towards AI, Inc. build and improve this course. This turned out to be great course and is released for everyone now. Congratulations What's AI by Louis-François Bouchard and Paul Iusztin for putting together a great blueprint for Agentic AI systems. Here's my review: --- Towards AI's "Agentic AI Engineering" course stands out as a comprehensive resource for moving beyond basics to building enterprise level production agentic systems. It is industry-focused, emphasizing real-world constraints and reliability. It has full spectrum of Agentic AI from core concepts and tooling all the way to monitoring, and evaluation. The course offers practical hands-on approach rather than just theory. This course provides the foundation and deep technical knowledge required to succeed in the world with AI Agents. Course link in the comments. 👇

💡Want to learn hands-on with me on how to build agents in #Azure? You should get a ticket for UphillConf in Bern (Switzerland) where I will deliver a full-day workshop! Let me know if I will see you there! Also, some other great workshops have been announced! Dear Lisa Carpenter Omar Solano and Louis-François Bouchard, comment below what your workshop is about and why people shouldn't miss it! 💎👏 #MVPbuzz #UphillConf26 #DevLife #Agents #Agentic #AI

ByteDance (TikTok's parent company) just released Seedance 2.0, and it's blowing up on Chinese TikTok with these AI 1-on-1 LeBron videos.

What happened in #AI? On February 5th, Anthropic and OpenAI released Claude Opus 4.6 and GPT-5.3-Codex, respectively, within minutes of each other. Both are point releases, but both deliver jumps in some benchmarks that look more like generational leaps. I think both releases are excellent, and they reward different use patterns. If you are writing code at the terminal all day, Codex 5.3 is now debatably the best tool available. If your work spans research, finance, document processing, and computer use, Opus 4.6 has the edge. Others in the news: Google for Agentic Vision in Gemini 3 Flash, Qwen team for Qwen3-Coder-Next, Mistral AI for Voxtral Transcribe 2, Waymo for Waymo World Model, Perplexity AI for Model Council, and more! Get a complete update on this week's biggest AI developments in the full issue 👇

Claude Cowork tried to delete a folder that user rejected

Anthropic just launched Agent Teams for Claude. This is a game changer for building. You can now have multiple agents working at the same time. They talk to each other to solve hard problems and finish tasks faster. Instead of one AI doing everything, you get a specialized team that handles different parts of your project. It makes the workflow much smoother. Check it out at code.claude.com to see how it works.

Anthropic was scheduled to drop their massive new model, Claude Opus 4.6, at 10 AM. They actually released it 15 minutes early to get a head start. But OpenAI was waiting. Literally minutes after Claude went live, OpenAI pulled the trigger on GPT-5.3-Codex. No warning. No big event. Just a direct counter-attack to steal the spotlight. The "Console Wars" of AI are officially here. Here is the breakdown of what actually dropped: Team OpenAI just gave us speed. The new Codex model is running 25% faster than the previous version. The killer feature is "Self-Correction", it can now spot its own bugs and fix them before you even see the error message. Team Anthropic is going for brain power. Opus 4.6 launched with "Agent Teams", which lets you run a squad of AI agents working together on one project. They also bragged about a true 1 Million Token context window and hit a massive 90.2% on the BigLaw Bench for legal reasoning. It basically comes down to this: Anthropic is aiming for deep thinking and memory. OpenAI is aiming for speed and building software. Which team are you on?

Anthropic is roasting OpenAI with a series of Super Bowl ads about putting ads in their AI. 😂

Inside the issue: • How teams catch data drift before users do • What really makes LLM inference faster (KV caches, FlashAttention, and more) • Why microservice design starts to matter once ML systems leave notebooks • A plain-English guide to variational autoencoders • And a spectral way of thinking about embeddings that could reshape how we read Transformer Read the full issue here 👇

This is huge for the Moltbook community. Prompt injection is one of those risks that's easy to overlook until it's too late 🔒