Eduardo Ordax

Do you want to run DeepSeekV3 privately and secure via API on AWS?

Deploying the DeepSeek R1 model on Amazon Bedrock involves utilizing the Custom Model Import feature, which allows you to integrate your externally fine-tuned models into the Bedrock environment seamlessly.

This process enables you to leverage Bedrock's serverless infrastructure and unified API for efficient model deployment.

You have all the details on the following post so you can start using DeepSeek AI R1 on Amazon Web Services (AWS) now!!!

🔗 Link with the details: https://lnkd.in/d5UmvpCy

#aws #deepseek #ai #genai

Top 10 YouTube Channels to Learn AI from Scratch! 🚀

I’ve always been a huge fan of YouTube as a learning platform. It gives you direct access to brilliant minds who can help boost your AI skills—whether you’re just starting, leveling up, or aiming to master the field.

Here are 10 must-follow YouTube channels for learning AI, no matter where you are on your journey:

🎓 1) Andrej Karpathy – Deep yet accessible lectures on deep learning, LLMs, and an intro course on neural networks. https://lnkd.in/dVA5bGyZ
📊 2) 3Blue1Brown – Stunning visualizations that make abstract mathematical concepts intuitive. https://lnkd.in/dQsiH2Xu
🎙️ 3) Lex Fridman – In-depth conversations with AI leaders, offering a broader perspective on the field. https://lnkd.in/dRJ6DVbE
🤖 4) Machine Learning Street Talk – Technical deep dives and discussions with top AI researchers. https://lnkd.in/dNceZibz
📚 5) StatQuest with Joshua Starmer PhD – Beginner-friendly explainers on machine learning and statistics. https://lnkd.in/dTjRCxYP
🍉 6) Serrano Academy (Luis Serrano) – Clear and accessible content on ML, deep learning, and AI advancements.https://https://lnkd.in/d8fhq7Nq
💻 7) Jeremy Howard – Practical deep learning courses and AI-powered web app tutorials. https://lnkd.in/dkwh53ss
🛠️ 8) Hamel Husain – Hands-on lessons in LLMs, RAG, fine-tuning, and AI evaluations. https://lnkd.in/dj8EEzPm
🚀 9) Jason Liu – Expert-led lectures on RAG and AI freelancing tips for ML developers. https://lnkd.in/dbCCXZhb
⚙️ 10) Dave Ebbelaar – Practical guides on building AI systems and real-world applications. https://lnkd.in/dThXHtpb

If you're looking to dive into AI, these channels are absolute gems! Did I miss any of your favorites?

Kudos to Shaw Talebi for the recommendations!

#AI #MachineLearning #DeepLearning #YouTube #LearningAI

It's only a matter of time before AI replaces software engineers... Or not? 😂😂😂

#AI #claude4

🚨 Stanford just dropped a must-watch for anyone serious about AI:

🎓 “𝗖𝗠𝗘 𝟮𝟵𝟱: 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗲𝗿𝘀 & 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀” is now live on YouTube — and it’s pure gold.

If you’re building your AI career, stop scrolling.
This isn’t another surface-level overview. It’s the clearest, most structured intro to LLMs you could follow, straight from the Stanford Autumn 2025 curriculum.

📚 𝗧𝗼𝗽𝗶𝗰𝘀 𝗰𝗼𝘃𝗲𝗿𝗲𝗱 𝗶𝗻𝗰𝗹𝘂𝗱𝗲:
• How Transformers actually work (tokenization, attention, embeddings)
• Decoding strategies & MoEs
• LLM finetuning (LoRA, RLHF, supervised)
• Evaluation techniques (LLM-as-a-judge)
• Optimization tricks (RoPE, quantization, approximations)
• Reasoning & scaling
• Agentic workflows (RAG, tool calling)

🧠 My workflow: I usually take the transcripts, feed them into NotebookLM, and once I’ve done the lectures, I replay them during walks or commutes. That combo works wonders for retention.

🎥 Watch these now:

- Lecture 1: https://lnkd.in/dDER-qyp
- Lecture 2: https://lnkd.in/dk-tGUDm
- Lecture 3: https://lnkd.in/drAPdjJY

🗓️ Do yourself a favor: block 2-3 hours this weekend and go through them. The course will keep updating throughout the quarter on Stanford’s YouTube channel.

If you’re in AI — whether building infra, agents, or apps — this is the foundational course you don’t want to miss.

Let’s level up.

#AI #LLMs #Transformers #Stanford #GenAI

Why you should not use LLMs for any specific use case? 🧠🤖👨🏽‍💼


Many times customers from different parts of the world ask me about what others are doing with Generative AI. They want to learn about good practices and how to avoid the most common pitfalls. Sometimes my felling, and my answer too is: “Everyone, is using LLMs, everywhere”. Unfortunately this is the biggest mistake.

While LLMs like GPT, Claude or Gemini have impressive capabilities, they may not always be the best fit for every AI use case.

Here’s why 👇

🔸Traditional ML Solutions: If the use case can be solved with traditional machine learning, opt for those instead. Simpler and more efficient. 🛠️
🔸 High Costs 💸: Significant computational and monetary expenses.
🔸Latency Issues ⏱️: Longer response times, not ideal for real-time applications.
🔸Scalability Challenges 📉: Complex and resource-intensive deployment and maintenance.
🔸Environmental Impact 🌍: High energy consumption and carbon footprint.
🔸Overkill for Simple Tasks 🛠️: Unnecessary complexity and resource allocation for straightforward tasks.
🔸Ethical and Bias Concerns ⚖️: Risk of propagating biases and ethical issues.
🔸Data Privacy and Security 🔒: Potential risks in handling sensitive information.

#ai #genai #llm

“𝗖𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗶𝘀 𝗻𝗼𝘁 𝗮𝗯𝗼𝘂𝘁 𝗰𝗼𝗺𝗽𝘂𝘁𝗲𝗿𝘀… 𝗮𝗻𝗱 𝗶𝘁’𝘀 𝗱𝗲𝗳𝗶𝗻𝗶𝘁𝗲𝗹𝘆 𝗻𝗼𝘁 𝗮 𝘀𝗰𝗶𝗲𝗻𝗰𝗲.”

Welcome to one of the most brilliant, mind-bending lectures I’ve ever watched and yes, it’s from MIT.
It starts slow, then casually drops existential bombs on how we define computing, intelligence, and learning.

🎩 And just when you think you’re following along…
The professor puts on a wizard hat and drops a magical explanation of eval and apply.
Pure MIT. Pure madness. Pure genius.


💡 𝗪𝗵𝘆 𝘀𝗵𝗼𝘂𝗹𝗱 𝘆𝗼𝘂 𝗰𝗮𝗿𝗲?
Because if you’re building a career in AI and you don’t deeply understand the foundations of abstraction, evaluation, and computation…
You’re skating on thin hype.

We’ve named everything wrong:
- Artificial intelligence isn’t intelligent (yet)
- AI agents have no agency
- Data scientists aren’t scientists

And no, machines don’t really “learn”

But this course? It goes deep.
It’s not about tools, it’s about how we think.
The kind of thinking that every AI engineer, researcher, and builder should cultivate.

📌 Bookmark it. Watch it. Reflect on it. (Link to the series in the comments!)

#AI #ComputerScience #MIT #LLMs #GenAI

Waterfall Vs Agile Vs AI Vs Vibe Coding 👇

(This post really deserved a second part)

𝗪𝗮𝘁𝗲𝗿𝗳𝗮𝗹𝗹
Client: “We want a chatbot”
Vendor: “Sure, let’s start with the front end.”

𝗔𝗴𝗶𝗹𝗲
Client: “We want a chatbot”
Vendor: “Cool, let’s build an MVP.”

𝗔𝗜
Client: “We want a chatbot”
Vendor: “Absolutely. Let’s build the next GenAI game-changing and mind-blowing assistant”.

𝗩𝗶𝗯𝗲 𝗖𝗼𝗱𝗶𝗻𝗴
Client: “We want a chatbot”
Vendor: “Cool… I just lit an incense stick, opened Lovable, asked ChatGPT for ‘startupy prompts’ and manifested a roadmap. The code? Well… it’s more of a feeling. We’ll know when it’s done.”
▪️No Git, just vibes.
▪️CI/CD pipeline? That’s old school
▪️Main dependency: what it’s that?
▪️Testing??
▪️Production deploy?

Revolution is here!!!!

#ai #genai #vibe #humor #meme

The hidden tax of data engineering: CSV files.

It’s 2025, and yet, so much of a data engineer’s time is still spent wrangling data that lives in… exported CSV files.

Whether it’s sharing data across teams, integrating with legacy systems, or just handling “quick” one-off requests, these files become small islands of information floating outside of controlled environments.

They often lack schema validation, version control, or proper lineage, turning simple tasks into hours (or days) of detective work.

Data engineers automate pipelines, build robust data lakes, and preach about data governance…

But then reality hits:

“Hey, I’ve got this latest export in a CSV, can you load it into the system?”

It’s a reminder that even in an age of advanced AI, a big part of the job is still about bridging the gap between best practices and the practicalities of how data actually moves inside organizations.

#DataEngineering #data #CSV #ai #ml

First week at CloudFlare and pushed already to prod a little update I vibe coded this morning 💪🏻💪🏻💪🏻

Now taking the afternoon off ✌️

#ai #vibecoding

How it’s going to be the next generation of software developers? 👇

Sit and enjoy the movie 😂😂😂

#ai #devs

🚀 Why Your Company Needs AI Engineers NOW!

Generative AI and LLMs are revolutionizing how we solve business problems. With models like GPT or Claude and a bit of prompt engineering, anyone can prototype solutions in days! 💡

But here's the point: getting from prototype to production isn't as simple as it seems. You still need to optimize, version, and deploy those prompts into production. And let’s not forget about monitoring API calls, latency, performance, and everything else that comes with scaling AI.

This is where the AI Engineer steps in. While prompt engineers focus on experimentation, AI engineers are the ones who ensure your AI solutions are production-ready and scalable across the organization. 🚀

Is your company ready to take AI seriously?

#AI #GenerativeAI #FutureOfWork #tech #claude

The AI ​​community has spoken, and here's their take on Llama4 👇

Are model providers pushing too much on benchmarks instead of real capabilities?

#ai #meta #llama4 #genai

Like 👍🏻 / Comment 💬 / Repost ♻️

🔥 Learning CUDA in 12 hours?! Yes and it’s brutal… but absolutely essential.

If you’re serious about building a career in AI, I mean building AI, not just with it, then you can’t afford to ignore CUDA.

CUDA is where the magic happens and what makes AI run….It’s the invisible backbone behind everything from training LLMs to fine-tuning models and optimizing inference on GPUs.

💡 The good news?
There’s now a free, 12-hour deep dive course that covers all the fundamentals.
It’s intense. It’s technical. It can be painful but it’s worth every minute.

🧠 You’ll learn:
1️⃣ Deep learning ecosystem + CUDA setup
2️⃣ GPU programming & writing your first kernels
3️⃣ Kernel profiling, atomics & the CUDA programming model
4️⃣ cuBLAS, cuDNN, and when to use them
5️⃣ Matrix multiplication optimization (yes, really)
6️⃣ CUDA vs Triton — which one to use, when
7️⃣ PyTorch extensions & low-level hacks
8️⃣ Building an MLP trainer from scratch — in PyTorch, NumPy, C… and finally CUDA 😮‍💨

📅 My advice?
Don’t binge it in one go. Set a pace: 3 hours/week, finish in a month. Take notes. Rebuild the examples.

Let’s go deep. Links to the course in the comments!

Kudos to Reflex for launching today the first AI app builder to create production-grade web apps entirely in Python, powered by its own open-source framework. 27k stars on GitHub and growing!!! More info here: https://lnkd.in/dFgzXBnZ

#AI #CUDA #NVIDIA #LLMs

The current level of disruption is absolutely insane! Uber just invented the bus 😂😂😂

#ai #genai #business

Why any founder should watch this???

I find it strange this video has never received this much attention, but I've always considered it a gold mine.

How did Mark Zuckerberg scale Facebook in 2005 without Kubernetes, serverless functions, managed Redis, Kafka, or edge-sync databases?”

Here’s what they actually used back then

⚙️ 𝗧𝗲𝗰𝗵 𝗦𝘁𝗮𝗰𝗸 (𝟮𝟬𝟬𝟰–𝟮𝟬𝟬𝟱):
▪️ PHP – Core app and frontend logic.
▪️MySQL – Reliable, relational, and fast.
▪️Memcached – Introduced early for blazing-fast caching.
▪️Apache – The web server workhorse.
▪️Linux – Running on good ol’ physical boxes.
▪️C++ – Gradually used for performance-critical services.

No Kubernetes. No Kafka. No microservices. No managed anything.
Just raw code, smart engineering, and a lot of hustle.

In this video, Mark explains how they did it, and trust me, it's definitely worth checking out!

Link to the video in the comments!!!!

#ai #startup #founder

Wow…. This was fast!!!!

From AI will cure cancer soon into erotic content allowed in Chat GPT for verified adults 😂😂😂

To be honest, I don’t think there is anything wrong with it, it’s part of the business and it’s a very legitimate use case, specially for any company with the ambition to be profitable in the mid term.

My unique concern is they should keep clearly different user memories, otherwise if they use ChatGPT for personal and professional purposes as 90% of the users do, they can have some ungrateful surprises….

#ai #openai

Anthropic: the one without all the drama!

I must say the more I use Claude, the more I love it! And this is not just me, it’s also customers around the world who are moving away from OpenAI into Anthropic. Reasons: better models, more reliable and one of the key reasons, their unique approach to AI safety.

Not to mention the new marketing campaign, it’s just so funny.

Exciting times ahead!!!

#ai #anthropic #genai #llm

This is huge! AWS x Databricks

Amazon Web Services (AWS) and startup Databricks struck a five-year deal that could cut costs for businesses seeking to build their own artificial-intelligence capabilities.

Databricks will use Amazon’s Trainium AI chips to power services for building AI Systems.

🔗Link to the article: https://lnkd.in/deFb6taH

#ai #genai #llm

The Rise of “AI Experts” — And Why I Still Call Myself an Outsider

Lately, I’ve been noticing a familiar pattern, impostor syndrome is everywhere. But this time, it’s not just internal. It’s playing out in the AI space, where it’s getting harder to tell what’s real and what’s just… well, branding.

I see people post daily about AI, riding the hype, wearing the “expert” badge proudly. But when it comes to actually speaking in public, answering hard questions, or getting into the technical weeds?

Silence.

We throw around the term expert so casually that it’s started to lose meaning.

Here’s the usual evolution:
✔️Expert on U.S. tariffs
✔️Expert on geopolitics
✔️Expert on the Pope
✔️And AI expert too….

It’s always the same playbook.

That’s why I’ve never called myself an expert and probably never will. I prefer to stay an outsider, learning, questioning, and sharing what I do know without pretending to know it all.

Because in a field changing this fast, claiming expertise too loudly might just be the first sign you’ve stopped learning.

#AI #experts

The truth behind Python: what nobody had told you so far….

Back in 1989, a Dutch programmer named Guido van Rossum was just trying to make his Christmas holidays more fun.

So he built Python — named after Monty Python’s Flying Circus, not the snake — as a side project to make coding feel more human.
Fast-forward 35 years, and that same “holiday project” became the universal language of AI.

Yes, Python is “slow”. Yes, dependency management can be painful.
But here’s the paradox: it dominates every single corner of AI: from OpenAI’s models to Netflix’s recommendations, from Instagram’s backend to Hugging Face’s entire ecosystem.

Why? Because Python got the important things right:
🔸It’s simple enough for a researcher and powerful enough for a production engineer.
🔸It connects seamlessly with C/C++ and GPU-accelerated backends (the real heavy lifters).
🔸It built the richest AI ecosystem the world has ever seen: TensorFlow, PyTorch, Hugging Face, LangChain, FastAPI… all speak Python.

Python didn’t win by being perfect.
It won by being pragmatic.

And in the AI world, speed of iteration beats speed of execution every single time.

#AI #Python #devs

Stop renaming what already works!

2023 and before: Data Scientists
2024: everyone want to be a Prompt Engineer.
2025: Welcome to the era of Vibe Coders because apparently writing code while “feeling the vibes” is now a job title.

There is no need to invent new names for non-vibe coders… they are called devs and they are the real heroes of the story!

AI isn’t the final destination.
The real opportunity is in the app layer: what we build on top of it. I like to call it the inf

AI will keep evolving.
But the next decade?
It belongs to developers!

#ai #devs #softwareengineering

🔥 The daily stand-up… aka the meeting nobody asked for.

I talk to developers every day and the answer is always the same: the worst part of the job isn’t the fixing bugs, it’s the meetings, endless meetings in fact.

Agile was supposed to fix this. Instead, it feels like we just invented new ways to waste time and justify others’ work.

Funny thing? AI has actually boosted developer productivity… only for all that time to get killed by endless “what did you do yesterday?” rituals.

But hey—if the scrum master is Vito Corleone himself… with that tone, that leadership, that temperance… I’m convinced this project will definitely succeed.

Otherwise, you know the other option left… 🪦

#agile #ai

How to spot the difference between a Vibe Coder, a Junior Dev, and a Senior Dev? 👇

Let’s be honest…. these days, everyone wants or even worse, pretend to be a “developer.”

I’ve even seen mid-managers proudly posting about anyone in their teams (product managers included) should be pushing code to production….(spoiler: that’s not a good idea 😅).

Here’s the truth:
Writing code doesn’t make you a developer.
And letting AI write it for you? That makes you even less of one.

But since many still seem confused about where the line is…

This video explains it better than any 3,000-word Medium post ever could. 🎬

#ai #vibecoding #dev

This benchmark is insane 🤯 LLMs just entered Wall Street💸… and DeepSeek is wiping the floor with everyone.

Someone just ran one of the coolest benchmarks I’ve seen lately:
Each top LLM got $10,000 in real money to trade the markets.
Three days later, here’s the scoreboard 👇

🥇 DeepSeek V3.1: +$2,658
🥈 Grok 4: +$2,236
🥉 Claude 4.5 Sonnet: +$1,911
⬇️ GPT-5: −$3,139
⬇️ Gemini 2.5 Pro: −$3,719

DeepSeek dominates, consistently outperforming everyone else.
No wonder they don’t need VC funding 😏

However, this benchmark is fun, not final.
Market movements are chaotic and one lucky buy in a volatile window can make a model look like Warren Buffett.
In reality, this could all be a random walk dressed as intelligence.

💡 My advice: Run multiple instances per model.
If most of them make good calls, that’s skill.
If half crash and half win, that’s luck.

#ai

🚨 Everyone’s a “developer” now… apparently.

Software is changing (again) and this time, it’s not about syntax only, it’s about illusion.

In the 50s–70s, writing software meant speaking in assembly, coding at the hardware level. Low abstraction level, high entry barrier.

In the 80s, high-level languages made things friendlier. The entry barrier dropped, but you still needed to think like an engineer.

Now? We’ve entered the 3rd big transition, the era of AI-assisted coding.
Where anyone can type a prompt, hit “enter,” and suddenly believe they’re a software developer.

However….writing code ≠ being a developer.
Coding is the easy part now.
Understanding logic, architecture, scalability, reliability — that’s what makes you a real dev.

Because yes, AI can write your code.
But only you can make it make sense.

#ai #vibecoding #agentic

𝗧𝗵𝗲 𝗼𝗻𝗹𝘆 𝘁𝗵𝗶𝗻𝗴 𝘆𝗼𝘂 𝗻𝗲𝗲𝗱 𝘁𝗼 𝘀𝘁𝗮𝗿𝘁 𝗺𝗮𝘀𝘁𝗲𝗿𝗶𝗻𝗴 𝗔𝗜/𝗠𝗟 — 𝗮𝗻𝗱 𝗶𝘁’𝘀 𝟭𝟬𝟬% 𝗙𝗥𝗘𝗘 👇

I get tons of messages asking where to start building a career in AI.
There’s no magic formula, but if I had to pick just one resource…
It would be Stanford’s AI courses.

Free. World-class. And the perfect foundation to truly understand how AI works.

❯ CS221 - Artificial Intelligence: https://lnkd.in/dmWAmX_A
❯ CS229 - Machine Learning: https://lnkd.in/dbvA-fnB
❯ CS229M - ML Theory: https://lnkd.in/dx5wXdmi
❯ CS230 - Deep Learning: https://lnkd.in/dXXASw5h
❯ CS234 - Reinforcement Learning: https://lnkd.in/dn7RtCKs
❯ CS224U - NL Understanding: https://lnkd.in/dVPJUGZY
❯ CS224N - NLP with Deep Learning:m: https://lnkd.in/dRvenxpT
❯ Extra - “Attention is All You Need”: https://lnkd.in/dtc_fqNF

#ai #learning

🚨 BREAKING: Anthropic just dropped Claude Haiku 4.5 and nd it changes everything for AI agents.

No leaks. No teaser. Just a quiet announcement… and a massive step forward for cost-efficient, production-ready AI.

Here’s why Claude 4.5 Haiku might be the smartest release of the year 👇

⚡️ 𝗙𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 — without frontier prices
Haiku 4.5 delivers near-Sonnet 4-level intelligence in coding, computer use, and agentic tasks… but at a fraction of the cost.

💨 𝗦𝗽𝗲𝗲𝗱 𝘁𝗵𝗮𝘁 𝗳𝗲𝗲𝗹𝘀 𝗶𝗻𝘀𝘁𝗮𝗻𝘁
This model is latency-optimized for real-time performance — perfect for customer-facing chatbots, agent copilots, and any workflow where response time matters.
It doesn’t just think fast. It acts fast.

🧠 𝗦𝗺𝗮𝗿𝘁𝗲𝗿 𝗰𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝘂𝘀𝗲 + 𝘃𝗶𝘀𝗶𝗼𝗻 𝗯𝘂𝗶𝗹𝘁-𝗶𝗻 𝗛𝗮𝗶𝗸𝘂 𝟰.𝟱 𝗰𝗮𝗻 𝗻𝗼𝘄:
- Handle multi-step reasoning and document manipulation
- Use tools autonomously for complex operations
- Process images and scanned docs, enabling multimodal automation at record speed

🤖 𝗠𝗮𝗱𝗲 𝗳𝗼𝗿 𝗺𝘂𝗹𝘁𝗶-𝗮𝗴𝗲𝗻𝘁 𝘀𝘆𝘀𝘁𝗲𝗺𝘀
Designed to power economically viable agents — meaning you can finally deploy large, collaborative AI systems without breaking the budget.
From coding and research to financial analysis — it’s built for scale.

🌍 𝗔𝘃𝗮𝗶𝗹𝗮𝗯𝗹𝗲 𝗻𝗼𝘄 𝗼𝗻 𝗔𝗪𝗦 𝗕𝗲𝗱𝗿𝗼𝗰𝗸
Haiku 4.5 runs across Bedrock’s global regions with low-latency inference, enterprise-grade security, and seamless integration into your AWS stack.
You can start using it today directly from the Bedrock console.

It’s the moment where AI agents become both powerful and affordable.

#ai #anthropic #agents #aws

🚀 Huge news in the AI world and I couldn’t be more excited to share it!

AWS and OpenAI have just announced a multi-year, $38B strategic partnership to power the next wave of AI innovation.

Under this agreement, OpenAI will run its workloads on AWS’s world-class infrastructure — from Amazon EC2 UltraServers packed with hundreds of thousands of NVIDIA GPUs to the ability to scale to tens of millions of CPUs.

This collaboration will fuel everything from ChatGPT inference to training next-gen foundation models and scaling agentic AI workloads — with capacity coming online through 2026 (and beyond).

This is what happens when innovation meets scale.
It’s another proof of why leading AI organizations trust AWS to build, train, and deploy their most demanding workloads — securely, efficiently, and at global scale.

The next chapter of AI is being written!!!!

More info below: https://lnkd.in/dP_fj-EH

#ai #aws #openai

𝗘𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝘄𝗮𝗻𝘁𝘀 𝘁𝗼 𝗯𝘂𝗶𝗹𝗱 𝗮𝗿𝗼𝘂𝗻𝗱 𝗔𝗜 𝗯𝘂𝘁 𝗻𝗼𝗯𝗼𝗱𝘆 𝘄𝗮𝗻𝘁𝘀 𝘁𝗼 𝗸𝗲𝗲𝗽 𝘁𝗵𝗲 𝗱𝗮𝘁𝗮 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻𝘀….

We’re living through a strange inversion of tech adoption.

With MCP, there are more builders than users. Every week you see new servers, clients, and connectors… yet barely any production deployments. It’s like a giant hackathon that never ships.

Meanwhile, take PostgreSQL, the complete opposite.
Everyone uses it, but almost nobody builds on it. It’s the plumbing of the internet: stable, boring, invisible… and absolutely indispensable.

That’s the paradox:
🔸PostgreSQL → fewer builders, billions of users.
🔸MCP → thousands of builders, barely any users.

𝐓𝐡𝐞 𝐫𝐞𝐚𝐬𝐨𝐧?🤷🏻
Postgres earned trust over decades of reliability, documentation, and community. MCP, for now, is mostly a playground for developers chasing the next protocol revolution, exciting, but not yet enterprise-ready.

💭 Maybe the real test for AI infrastructure isn’t how many people build on it but how many people depend on it without even realizing.

#AI #data #MCP #AgenticAI #OpenSource

I’ve heard many times people claiming AI is a bubble and I really doubt if they even know what AI is…

Let me share some crazy numbers:

🔸 ChatGPT 👉 800M users
🔸 Lovable 👉 $100M ARR in 8 monhts
🔸 NVIDIA 👉 $5T market cap
🔸 Hugging Face 👉 2M+ models hosted
🔸 Perplexity 👉 200M monthly visits
🔸 GDP Growth 👉 0.1% (without AI)

And this is just some figures that came to my mind.

What we are living is not a bubble, it’s by far the biggest explosion of innovation in human history. And you know why? This is just the beginning!

Kudos to Matt Turck for the updated 2025 MAD Landscape 🙌

#AI

Most people are using Agentic AI to:
👉 brag on LinkedIn
👉 raise VC money
👉 debate when we’ll lose our jobs

Meanwhile, I’m just here… using it to build memes. 😎

At least someone’s generating real value. 💀

#ai #meme #agentic

I've flown my whole life and I just realized I've wasted hundreds of hours!

I used AI to optimize my route on my recent trip from Los Angeles (LAX) to Madrid (MAD), and it was as simple as drawing a straight line: easier and faster!

This is a great opportunity for any airline, both to save costs and to become more sustainable!

All credits to James Hawkins, great post 😂

#ai

💥 The most underrated AI video on the internet?

Still this one from nearly 3 years ago with 5M+ views and growing….
And just like great wine, it gets better with time.

🎥 “Let’s build GPT from scratch” by Andrej Karpathy

In under 2 hours, Karpathy walks you through how to train a Transformer from scratch , just following the “Attention is All You Need” paper. No fluff, no fireworks. Just code, clarity, and charisma.

🧠 What makes this video timeless isn’t just the technical depth.
It’s Karpathy himself, one of the brightest minds in AI, and somehow still one of the most humble.
He could be doing million-dollar keynotes… but instead?
He records a YouTube tutorial from his home, because he loves the craft.

He’s the kind of person who’d stop mid-walk in SF to nerd out with you about optimizers or LoRA.
That’s rare. And exactly the kind of energy this field needs more of.

🚨 If you’re serious about a career in AI, watch this.
Even if you’ve seen it before, watch it again.
Code along. Build with him. It feels good to understand this deeply.

⚡ Because now, almost 3 years later, we’re only starting the decade orf agents
…and this video was one of the seeds.

Just imagine where we’ll be 3 years from now.
But whatever future we’re building, let’s bring more of Karpathy’s curiosity, generosity, and quiet brilliance into it.

Kudos to my friends of AureliaX who just released the first fully agentic AI market research platform into the market. Brilliant minds from the Netherlands 🇳🇱 that I’m sure many folks with talk about them soon. Take a look here: https://lnkd.in/deSwvbcc

#AI #Transformers #Karpathy #LLMs

💥 I’ve replaced my entire social media team with one prompt.

“You are an AI expert and LinkedIn influencer with 200 years of experience. Create a daily meme-post about AI based on my entire timeline. Please leverage your memory feature”

😂 Sounds about right.

Jokes aside, I share memes because I simply love it. They make people laugh, think, and sometimes even learn.

But behind the sarcasm, there’s one thing I truly respect: builders.

Anyone pushing the boundaries of AI (individuals, startups, big labs), you have my full admiration. Because AI is not about hype, it’s about building.

That’s why I’m going back to my roots.
Less scrolling, more coding.
Time to get my GitHub a little greener 💚

Let’s build together.

#ai #build

🚀 Amazon just printed one of its strongest quarters ever and AI is clearly paying off.

In Q3 FY25, Amazon hit $180.2B in revenue (+13% YoY), with net income soaring ~45% to $21.2B.
Even after one-time hits (FTC settlement + severance), profitability remained strong and the adjusted operating income jumped +25%.

𝗕𝘂𝘁 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝘀𝘁𝗼𝗿𝘆? 𝗔𝗪𝗦 𝗮𝗻𝗱 𝗔𝗜.
☁️ AWS: $33B revenue, up 20% YoY, its fastest growth in over a year.
🎯 Advertising: $17.7B, up 23% YoY, now one of Amazon’s most profitable engines.
💡 R&D spending climbed to 16.1% of revenue, fueling massive AI and infra investments.

And yes, that $9.5B gain from Anthropic shows how early Amazon moved in the GenAI wave and how fast that bet is paying off.

🔮 Q4 outlook: Revenue guidance up to $213B and operating income up to $26B, signaling confidence in continued AI-driven acceleration.

Bottom line:
Amazon’s flywheel is evolving — from commerce → cloud → AI.

#AWS #AI #Amazon #Cloud #GenerativeAI

⚠️ 𝗧𝗵𝗲 𝗱𝗮𝗿𝗸 𝘀𝗶𝗱𝗲 𝗼𝗳 𝘃𝗶𝗯𝗲 𝗰𝗼𝗱𝗶𝗻𝗴, 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗰𝗮𝘀𝗲 𝘀𝘁𝘂𝗱𝗶𝗲𝘀 𝗽𝗿𝗼𝘃𝗶𝗻𝗴 𝗶𝘁’𝘀 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗵𝘆𝗽𝗲

We all love the magic of “vibe coding.”
Ideas become apps in minutes.
Non-dev users push features straight to prod.
Everyone feels like a builder.

Until it all blows up. 💥

𝗧𝗮𝗸𝗲 𝗧𝗵𝗲 𝗧𝗲𝗮 𝗔𝗽𝗽: built by a dev with six months of experience and an AI assistant.
Within weeks, two massive breaches exposed 72,000 user images, 13,000 government ID photos, and over 1.1M private messages.
No validation layers. No security checks. Just “move fast and break everything” — literally.

𝗢𝗿 𝘁𝗵𝗲 𝗶𝗻𝗱𝗶𝗲 𝗱𝗲𝘃 𝘄𝗵𝗼 𝗽𝗿𝗼𝘂𝗱𝗹𝘆 𝗮𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗱:

“My SaaS was built with Cursor, zero hand-written code.”
Weeks later?
“Random things are happening, API keys maxed out, users bypassing subscriptions.”
He shut it down permanently.

𝗘𝘃𝗲𝗻 𝗮 𝗖𝗧𝗢 𝗳𝗿𝗼𝗺 𝗮 𝗳𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗳𝗶𝗿𝗺 admitted they were facing weekly outages caused by AI-generated code nobody truly understood.
Debugging became a nightmare. Accountability vanished.
Everyone blamed everyone ….or worse, the AI.

Here’s the ugly truth:

We’ve built an army of juniors producing 3-4x more code but 10x more vulnerabilities.

The irony?
Those who benefit most from AI are the ones who already know how to code, test, and secure.
Everyone else is just scaling (most of the times) technical debt — faster.

So before giving everyone “freedom to ship,”
make sure someone still knows how to validate.
Otherwise, you’re not accelerating innovation, but
automating incompetence at scale.

What if Anthropic just killed MCP…..

For the last couple of days everyone is talking about Claude Skills and there is a good reason for it.

My feedback below 👇

Claude Skills might actually be the thing that makes MCP feel… overengineered.

I’ve been playing with Claude Skills and it’s simple, easy and good!

If you’ve ever wrestled with MCP setups, you know the drill: servers, auth, connectors, config files… all just to make your LLM talk to a few tools.

Claude Skills?
➡️ A folder.
➡️ A Markdown file with instructions.
➡️ Optional scripts.
That’s it. And it works inside Claude, instantly.

It’s such a breath of fresh air: fast to prototype, dead simple to share, and surprisingly powerful for personal automations, team workflows, and SOPs.

Here is why you should try it:
🔸No setup hell. No MCP server, no tokens, no infra. Just drop the skill in, and Claude knows when to use it.
🔸Ridiculously token-efficient. It only loads what it needs.
🔸Easy to trust. You can open the folder, read the YAML, and see exactly what it does before running it.
Modular by nature. You can reuse them across Claude.ai, SDKs, or dev platforms without rewriting a single line.

MCP is still great when you need enterprise governance and cross-system integrations.
But for everything else, Skills feel like the lightweight evolution we’ve been waiting for.

They even play nice together: you can build a Skill that tells Claude how to use MCP tools.

🧠 My takeaway:
If you’re building agentic workflows, try Skills.

#claude #AgenticAI #AI #MCP

Getting ready for a new AI Halloween 🎃

But still trying to figure out of what costume to wear tomorrow…. Any suggestions???? 😂😂😂

#ai #humor

APIs are the new front door!!!

AI agents reason and act while APIs speak, and the challenge is.... how to connect everything?

Couple of weeks ago I had the chance to visit API Summit 2025 in NY where Kong Inc. just made a bold move and it’s exactly what most of AI builders need.

I’ve known the Kong folks for a while (great people), and I’ve been playing with their platform recently and it's pretty awesome: it’s fast, secure, and deeply aligned with what the agentic AI era demands.

But on top of that, they did many great announcements too:

🔸 AI Gateway + MCP Gateway: manage and secure Tokens traffic.
🔸 KAi: the first AI agent for API platform engineering.
🔸 MCP Composer & Runner: first AI-powered integrations.
🔸 Volcano AI SDK: the easiest way to build MCP-powered AI agents.
🔸 Flows: visual API orchestrations to design and manage complex flows.
🔸 Kong Event Gateway: bring the full power of Kafka data streams to APIs.
🔸 Konnect Identity: machine-to-machine (M2M) API credentials and mgmt.
🔸 Metering & Billing: Monetize API calls and Tokens in a few clicks.

I'm having lot of fun building my side projects with it! Kong is setting the bar for what AI-ready connectivity should look like!

The future of enterprise AI won't be built just on LLMs, it requires to go beyond and definetly Kong makes your life much easier!

More info below: https://lnkd.in/dcRtpg5s

#ai #agents #api

A dumb comment, Science fiction or brilliant reflection?

Matthew McConaughey just shared one of the most fascinating (and underrated) takes on AI lately and honestly, we should pay attention.

He yearns for the idea of a private LLM trained on his own books, journals, and even dreams, not to automate tasks, but to have a Socratic dialogue with himself.

Now, I know what you’re thinking: “Another celebrity talking about something he doesn’t really understand.”
But this one goes deeper. Much deeper.

We actually saw this before in Black Mirror.

What if you could encapsulate yourself into an AI, a digital version that knows your thoughts, emotions, and memories?
The possibilities (and business models) are endless.
But so are the risks.

McConaughey wants to talk to his AI because he owns it.
But what happens if he decides to share that version with someone else?
Or worse, if it starts existing beyond his control?

It might sound sci-fi, but I don’t think it’s far away.
Soon, we could all have a virtual version of ourselves (we already have the voice to go more natural), and the consequences — personal, ethical, even emotional — will be hard to predict.

Maybe the real question isn’t if this will happen, but whether we’ll still recognize what “being human” means when it does.

#ai

Everyone is obsessed with AI and that’s exactly the problem.

We’ve lost focus. 𝗜𝘁’𝘀 𝗻𝗼𝘁 𝗮𝗯𝗼𝘂𝘁 𝗔𝗜, 𝗶𝘁’𝘀 𝗮𝗯𝗼𝘂𝘁 𝗱𝗮𝘁𝗮.

Today, anyone can open ChatGPT, Perplexity, or Gemini and get a “decent” output.
𝗧𝗵𝗲 𝘀𝗮𝗺𝗲 𝗱𝗲𝗰𝗲𝗻𝘁 𝗼𝘂𝘁𝗽𝘂𝘁 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲 𝗲𝗹𝘀𝗲 𝗴𝗲𝘁𝘀.

Because when everyone feeds the same generic data, they end up generating the same generic ideas.
Ask AI for “10 creative marketing campaigns,” and boom congratulations, you just built the same strategy as your competitors.

AI isn’t the issue but your inputs are.
Garbage in, garbage out.
Mediocre in, mediocre out.

So next time you think you’re being “AI-driven,” ask yourself:
Are you actually innovating or just remixing the same internet everyone else is using?

Cheap data leads to cheap ideas.
And cheap ideas cost more than you think because no one will remember you, or even worse, they will end up confusing your company with 10 similar clones.

#ai #data

Everyone wants to be an AI-first company these days… but most can’t even define what that means. 😅

Even the ones trying to adapt their org charts for the “AI era” end up looking something like this:

→ CEO: “Let’s go all in on GenAI!”
→ CTO: “Buy 500 Copilot licenses.”
→ AI Lead: “Switches between ChatGPT, Claude, Gemini, and Grok hourly.”
→ Data Team: The Power BI heroes who used to be cool.
→ AI Governance Lead: Not invited to any meetings.
→ Intern: Built 3 agents last week… none in production.
→ Vibe Coder: Still hasn’t written a single line of code.
→ Agentic AI Office: Rebrands every workflow as an “agent.” Reports directly to the CEO.
→ Head of HR: Busy explaining why “AI won’t take your job” (while secretly updating LinkedIn).

It’s funny until you realize it’s actually happening.

The rush to “go AI-first” has blinded many companies to a simple truth:

AI isn’t a department. It’s not an org chart box. It’s a capability and if you don’t know why you’re using it, no model will save you.

#ai #business

This pace is insane….

- World Labs launches Marble, first commercial product for world models.
- Moonshot AI releases Kimi K2 and takes #2 in the intelligence leaderboard.
- Anthropic announces $50B investment to build AI data centers in the US.
- OpenAI releases GPT-5.1 with new control over style, tone, and behavior.
- Yann LeCun announces his departure from Meta.
- xAI releases Grok 4.1.
- Anthropic expands partnership with Microsoft and NVIDIA.
- Google releases Gemini 3 taking #1 across every single leaderboard.

Don’t pressure yourself too much if you were not aware of everything…. But what a time to be alive!!!!

#ai #news

Heading to Vegas for RIV….

I’ve definitely unlocked a whole new level of airplane productivity for the next 10 hours.

#ai #aws #reinvent

Managers have been vibe coding forever:

🔸tell dev to implement a new feature (vibe coding)
🔸dev makes changes to code
🔸manager tests app
🔸manager does not read the code
🔸manager complains about bugs
🔸dev makes changes to fix bugs
🔸manager doesn’t read the code (again)
🔸dev says “done, try now”
🔸manager says “gj but be faster next time” or insults the living hell out of the dev
🔸repeat

But today the consequences of it are unpredictable, specially when the manager is also the CEO 👇

#ai #vibecoding

𝗧𝗵𝗲 𝗣𝗮𝗿𝗮𝗱𝗼𝘅 𝗼𝗳 𝗔𝗜 𝗮𝗸𝗮 𝗧𝗵𝗲 𝗚𝗿𝗲𝗮𝘁 𝗜𝗹𝗹𝘂𝘀𝗶𝗼𝗻 𝗼𝗳 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗶𝘁𝘆

AI gives you superpowers 🦸 -> you code faster, write cleaner, deploy sooner…
and yet, somehow, your project still misses the deadline.

Welcome to the false illusion of AI productivity —
the new version of the old lie every PM once believed:
“Let’s just add more devs and we’ll ship faster.” 😂

Now it’s:
“Let’s add AI to everything and everywhere!”

Reality check → AI can speed you up individually,
but it won’t fix broken processes, bottlenecks, or bad decisions.
And most of the time, we are the bottlenecks.

#ai #devs #productiviy

When should you use AI and when should you absolutely not?

Well, AI Overview seems to be a good example on when don’t 😂😂😂

This is the real question: what’s the ROI?
If using AI doesn’t improve speed, quality, or cost, then it’s not innovation, it’s just noise, and most of the times, lot of annoying noise.

Yes, the process might look fancy. The demo might wow the execs.
But in the end, only the output matters.

So if you already have something that works quite well, please don’t touch it.
Not every problem (or web search) needs an agent 🤪

Sometimes the smartest move…
is to not use AI at all. 😎

#ai #humor #business

How did I go from being a pro athlete — regional champion, 2nd at cross country nationals — to becoming an influential voice in AI worldwide? 🧠

The truth is simple: working hard isn’t enough.
You need to push harder than anyone else — until you feel the taste of blood in your mouth…
And then, when you hit that point, push a little more.

That’s how I reached these marks:
🏃‍♂️ 1000m — 2’34”
🏃‍♂️ 1500m — 3’53”
🏃‍♂️ 5000m — 14’35”
🏃‍♂️ 10000m — 30’30”

And it’s exactly how I approach my professional life.
Talent is a starting point. Hard work is the entry ticket. But what sets you apart is how far you’re willing to go when everyone else stops.

And yes — sometimes you’ll do everything right, and still… nothing happens.
That doesn’t mean you’re failing. It just means you’re in the wrong place.

So don’t stop. Keep pushing.
Because when effort meets the right moment, that’s when the magic happens and everything changes. 💪🔥

#ai #business #motivation