Armand Ruiz

Director Data Science & AI Elite

Leading the Product Management org for the AI Portfolio of products focused on Generative AI and Traditional ML. Products include: watsonx.ai and watsonx.gov

Vice President of AI Platform

I am building the world's largest team of 𝐀𝐈 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐬 by

- Recruiting hundreds of new engineers around the globe
- Reskilling our existing talent
- Offering them top-quality training in new-generation AI

Responsible for Pre-Sales Generative AI Pilots with the new IBM watsonx.ai Platform.

Director watsonx team - Client Engineering

Director of the group IBM Data Science Elite team. We prove the value of IBM Watson solutions by co-creating pilots with our customers. I managed a team of 90+ data scientists that worked on 100s of complex AI implementations.

Director Data Science

I manage a rockstar team of Product Managers leading the product portfolio of IBM Watson AI products.

Product Director - Product Management, Watson Studio and Machine Learning

Leading new Data Science products for IBM

Lead Product Manager - IBM Data Science and Watson Machine Learning

Bring Open Source data science tech to IBM stack

Product Manager for Data Science Open Source Extensibility 

In charge of Data Science projects in EMEA

Data Science Advocate - IBM Data Science & Machine Learning

Support sales in Data Science opportunities

Technical Seller Data Science & Smarter Cities

Internship to bring new innovations to IBM

Intern - Technical Software Engineer, Extreme Blue

Vodafone

Optimize 3G network coverage using Data Science

Data scientist - Radio Optimization Engineering

Federació Catalana de Tennis

Tennis Coach

VP of Product - AI Platform @IBM

As the Director of AI Engineering at IBM, I lead the watsonx team, a global group of over 1,000 AI Engineers who work on cutting-edge Generative AI solutions for various industries. With more than 12 years of experience in AI, I have a passion for democratizing practical AI skills and empowering professionals to leverage AI to transform their business.

Previously, I managed IBM's Data Science and AI Elite, delivering complex AI implementations for hundreds of clients. 

I also led the product management of IBM Watson Studio and Watson Machine Learning, an AI platform used by hundreds of customers and SaaS apps powered by AI. I hold a Master of Science in Telecommunications from Universitat Politècnica de Catalunya and speak three languages: Spanish, Catalan, French, and English.

Connect with me!

✉️ JOIN MY NEWSLETTER: Every Saturday morning, you'll get 1 actionable tip to launch and grow your AI strategies in business in less than 4 minutes. → GO HERE: https://www.nocode.ai/

📧 Email: armand@nocode.ai
X: @armand_ruiz


"It is kind of fun to do the impossible"
All opinions are my own and don't represent my employer.

Called it. 🎯 The Model Context Protocol (MCP) is officially moving to the Linux Foundation under the new Agentic AI Foundation.

I’ve been saying for a while that open governance is the only viable path forward for AI infrastructure. Today’s news proves it.

This brings Anthropic, OpenAI, Microsoft, Google, and others under one neutral banner. 

10,000+ active servers. 
97M+ SDK downloads. 
Zero vendor lock-in.

If you want an ecosystem to flourish, you have to open it up. The era of Agentic AI just got a serious upgrade.

Read the news here: https://lnkd.in/g7vy5-x2

At NeurIPS2025, Google presented Titans, a new architecture that combines the speed of RNNs with the performance of Transformers. 

The trade-off between infinite context and computational cost is the biggest hurdle in AI right now.

While Transformers are accurate but expensive to scale, and linear models are fast but lose detail, Titans delivers the best of both worlds by introducing Test-Time Memorization.

The Core Innovation: Instead of a static state, Titans uses a deep neural memory module that actively updates while running, utilizing a "Surprise Metric":
- Low Surprise: Routine data is processed efficiently without clogging memory.
- High Surprise: Anomalies and key details are instantly integrated into long-term memory.

The Results:
- Scales effectively to 2M+ token context windows.
- Outperforms larger models (including GPT-4) on extreme long-context reasoning benchmarks.
- Generalizes beyond text to genomics and time-series.

This is a move from passive processing to active, efficient learning.

Jensen just dropped a masterclass on the future of AI. 

It was great to hear from him in a long-format podcast conversation. 

Three key insights he shared:

1/ Human language is the new syntax: The "tech divide" is closing. You won't need C++ to code the future; you just need to speak.

2/ The "30-Day" Mindset: Nvidia operates like they are always 30 days from bankruptcy. This paranoia fuels their innovation and resilience.

3/ Safety is an engineering problem: Forget the sci-fi movie plots. AI safety is just like cybersecurity—as capabilities rise, so do the guardrails.

Pic taken during my epic commute

No alternative text description for this image

Current mood: "Joining a new company is akin to an organ transplant - and you're the new organ. If you're not thoughtful in adapting to the new situation, you could end up being attacked by the organizational immune system and rejected."

Learning fast, exploring the culture, and pushing myself to bring out the best work I’ve ever done. 

Loving the challenge. 🤩

As a leader, your job is to run a tight ship.

That doesn’t mean control. It means precision.

Clear goals. 
Clear roles. 
Clear standards.

In fast-moving teams, especially in AI, ambiguity is expensive and confusion kills momentum.  Alignment compounds it.

Gemini 3 is here and it isn’t just another model release. It’s Google firing on every AI cylinder: infra, research, product, and distribution.

A few things stand out:

1.	Full-stack advantage, finally realized.
Gemini 3 is the first model shipped into Google Search on day one. That’s not just marketing ; it’s how you put frontier AI in front of billions.

2.	The agent era is no longer theoretical.
With Deep Think mode, long context, multimodal understanding, and real-time planning, Gemini 3 moves beyond chat. It’s not a chatbot. It’s a problem-solver.

3.	AI is getting better at “reading the room.”
It reflects a core shift: models are starting to grasp intent and nuance and not just data.

Benchmarks back it up: topping LMArena, solving “Humanity’s Last Exam”, setting a new bar in math and science tasks. But that’s the sideshow.

More info here: https://lnkd.in/gWeXGwhC

I cannot wait to try it out

JSESSIONID

li_at

Talks about #ai, #nocode, #deeplearning, #machinelearning, and #artificialintelligence

building AI systems

𝗔𝗜 𝗪𝗼𝗿𝗹𝗱 𝗠𝗼𝗱𝗲𝗹𝘀 𝟭𝟬𝟭

Current AI is often described as a "statistical correlation machine." It predicts the next word or pixel based on patterns it has seen before. World Models represent a shift toward causal understanding—giving machines a way to simulate the future.

What is a World Model? Think of it as an internal flight simulator. Instead of just recognizing an environment, the AI builds a mental model of how that environment works. It uses this model to test hypotheses ("If I drop this glass, it will break") without needing to actually perform the action.

The "𝗟𝗮𝘂𝗻𝗱𝗿𝘆" Example Imagine training a robot to fold clothes.

- 𝗪𝗶𝘁𝗵𝗼𝘂𝘁 𝗮 𝗪𝗼𝗿𝗹𝗱 𝗠𝗼𝗱𝗲𝗹: The robot must physically interact with the clothes millions of times, learning through endless trial and error.

- 𝗪𝗶𝘁𝗵 𝗮 𝗪𝗼𝗿𝗹𝗱 𝗠𝗼𝗱𝗲𝗹: The robot simulates the physics of the fabric in its head. It plans the fold internally, predicts the outcome, and only moves its arm when it is confident. This results in safer, faster, and more generalizable learning.

𝗛𝗼𝘄 𝘁𝗵𝗶𝘀 𝗱𝗶𝗳𝗳𝗲𝗿𝘀 𝗳𝗿𝗼𝗺 𝗰𝘂𝗿𝗿𝗲𝗻𝘁 𝗟𝗟𝗠𝘀: While Large Language Models (LLMs) are masters of syntax and semantics, they often struggle with physical causality. World Models differ because:

𝟭. 𝗧𝗵𝗲𝘆 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗲, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗿𝗲𝗰𝗮𝗹𝗹: They don't just retrieve information; they run mental experiments.

𝟮. 𝗧𝗵𝗲𝘆 𝗹𝗲𝗮𝗿𝗻 𝗹𝗮𝘄𝘀, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗿𝘂𝗹𝗲𝘀: They infer the physics of their environment from data rather than being hand-coded.

𝟯. 𝗧𝗵𝗲𝘆 𝗮𝗿𝗲 𝗽𝗿𝗼𝗮𝗰𝘁𝗶𝘃𝗲: They predict "what happens next" in a state-space, not just a token-space.

𝗧𝗵𝗲 𝗥𝗼𝗮𝗱𝗺𝗮𝗽: The next frontier is the convergence of these technologies. By combining the reasoning of LLMs with the grounding of World Models, we move closer to Embodied Intelligence—software that can read, watch, plan, and act in the physical world.

Scaling laws in AI are nowhere near finished. The latest Gemini 3 results show it clearly.

Pre-training: Many believed progress had plateaued. Yet the jump from 2.5 to 3.0 is one of the biggest leaps we have seen. When you improve data, architecture, and compute together, you keep breaking ceilings. No walls in sight.

Post-training: This might be the real frontier. We are still early in how we align, steer, and specialize models. The innovation curve here is just beginning, and the gains on 3.0 prove it.

Good insights from Deepmind engineers this week.

Better models are not coming slowly. They are coming faster.

Plan for continuous capability jumps, not incremental upgrades. The disruption curve is steepening.

it is very cool to work everyday with people in person. i think  product innovation and magic really happens with real human interaction

AI is making 10X engineers 1000X, and 1X engineers 0.1X 

stealing this quote from Shadi Copty 🙃

Fortusight, Founder | AI advisor | Ex-Dir of Llama @Meta & Ex-VP of Watson @IBM

AI Evals Explainer

Generated with Nano Banana Pro

It’s insane how many hours I’ve seen senior technical leaders burn on architecture charts. 

All that time spent polishing pixels... now you just describe it in plain English and boom and nano banana AI handles the rest.

By the way, this is a solid, modern and agile AI stack. To build at x10 speed you need to have the right building blocks.

Fortusight's AI stack

A lot of people ask me this question, I'll start with the absolute core:
1. I use Agno for multi-agent framework, they're fast, elegant and robust.
2. Exa for web grounding, I consider Exa a bigger booster than LLM choice for my use case.
3. FMP for premium, clean, reliable financial data turned into tailored tools.

For LLMs, I use O3 for multi-agent orchestration and planning, GPT 4.1 for digesting web content and supporting my A1Facts flow; Flash2.5 for structured extraction and Gemini 3 pro for synthesizing investment thesis.

I constantly progress the choices of LLMs prioritizing quality over price.

For tools, I use Cursor with Opus 4.5 (recently replacing Gemini 2.5 pro); and Figma make - which I absolutely love.

For internal ops agents, (1) managing coverage requests (2) digesting user feedback (3) analyzing product analytics (4) writing my weekly newsletter -- all built with Agno and Gemini 3 pro.

Less exciting but just as critical :-) Streamlit for front end [being replaced with react], and Firebase for database/auth/analytics -- both choices have been really critical for prototyping rapidly with my users.

The "secret sauce" is secret (lol), but I will say solving factuality in evolving market dynamics is a significant effort, and using expert-system techniques and simulations augments LLMs where they are critically weak.

Happy to share more if you have any questions -- and obviously here's the link if you want to try it out (https://lnkd.in/g7eR86dr)

Introducing SAM 3D: Powerful 3D Reconstruction for Physical World Images

and it’s not your typical 3D reconstruction tool.

It does what previous models couldn’t:
- Reconstruct real-world objects and scenes from a single image.
- Handle occlusion, indirect views, and cluttered backgrounds.
- Estimate human pose and shape with surprising accuracy.

Why does this matter?
Because for the first time, we’re seeing 3D perception at the scale, quality, and accessibility of today’s 2D models.

The architecture behind SAM 3D borrows from LLMs: pre-training on synthetic data, followed by post-training on real-world images using a human-in-the-loop ranking engine. The result is a feedback loop that continuously improves both the data and the model.

The implications stretch far beyond creative media. Robotics. E-commerce. Sports medicine. Interactive avatars. You name it.

And the best part? 

It's fast. 
Real-time fast.

Meta’s already using SAM 3D to power a “View in Room” feature in Facebook Marketplace; turning static listings into immersive experiences.

The gap between virtual and physical is shrinking. SAM 3D is a serious leap forward.

Learn more:
- 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗮𝗻𝗻𝗼𝘂𝗻𝗰𝗲𝗺𝗲𝗻𝘁 𝗯𝗹𝗼𝗴: https://lnkd.in/g8w6dvAB
- 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗦𝗔𝗠 𝟯𝗗 𝗢𝗯𝗷𝗲𝗰𝘁𝘀 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗮𝗽𝗲𝗿: https://lnkd.in/gd_HQE9c
- 𝗥𝗲𝗮𝗱 𝘁𝗵𝗲 𝗦𝗔𝗠 𝟯𝗗 𝗕𝗼𝗱𝘆 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗣𝗮𝗽𝗲𝗿: https://lnkd.in/gp3p9jaf
- 𝗗𝗼𝘄𝗻𝗹𝗼𝗮𝗱 𝗦𝗔𝗠 𝟯𝗗 𝗢𝗯𝗷𝗲𝗰𝘁𝘀: https://lnkd.in/gfwmcKGK
- 𝗗𝗼𝘄𝗻𝗹𝗼𝗮𝗱 𝗦𝗔𝗠 𝟯𝗗 𝗕𝗼𝗱𝘆: https://lnkd.in/g9CPHEXJ
- 𝗘𝘅𝗽𝗹𝗼𝗿𝗲 𝘁𝗵𝗲 𝗣𝗹𝗮𝘆𝗴𝗿𝗼𝘂𝗻𝗱: https://lnkd.in/guSDTnU3

A sharp, modern AI tech stack for building a serious startup.

Fortusight from Shadi Copty lays out a blueprint that actually reflects where the industry is going:

multi-agent architectures, real-time grounding, premium data, and a diversified LLM strategy tuned to specific jobs rather than brand loyalty.

1/ Agno for agent frameworks.
2/ Exa for grounding.
3/  FMP for clean financial data.
4/ A portfolio of models (O3, GPT-4.1, Flash2.5, Gemini 3) each picked for its superpower.

Cursor + Opus 4.5, Figma Make, Firebase, and Streamlit/React for fast product loops.

And a layer of expert systems + simulation to fix factuality where LLMs keep failing.

This is the kind of stack that lets a small team punch far above its weight.

🚀

Open article: Financial Analysis App by fortusight.streamlit.app

Exciting project for you all to check! I tried an earlier prototype and it was very impressive

Startup life, here I come :-)

Thrilled to share with you my new adventure, Fortusight - a project I've labored on nights and weekends for the last 2 years.

It's a tool for the serious investor that wants to deeply understand a company before buying and holding for 5-10 years.

It uses AI Agents to read through filings, earning calls, analyst reports, news, press-releases - forms an opinion on a company's competitive position, management's friendliness to shareholders, and creates bull/base/bear scenarios for valuation.

To quote a friendly hedge fund manager "I used to have 5 people doing all of this" - each company is analyzed for 192 different perspectives, with A1Facts to ensure factuality for high stakes decisions.

Here's how you try it: https://lnkd.in/g7eR86dr

Let's go 🚀 🚀 🚀

🚨 Stop scrolling. The math on this is barely real.

Get a $799/mo institutional AI research platform for just $50/mo.

The catch: Only 10 spots open. 

RIAs & Fund Managers—this is a complete no-brainer. 👇

https://fortusight.ai/

I need your help 🙏

I'm looking for design partners for my baby startup that just got pre-seed funded.

If you know a wealth fund manager, a small RIA, a family office that wants to reduce their equity research costs from ~$75K to less than what they pay for starbucks coffee - share this link with them: https://fortusight.ai  

For 10 design partners, I'm giving away Fortusight for $50/month instead of $800/month for a year in exchange for feedback to help guide the product's development.

The beta with our users has been incredibly successful, with recurring usage of our equity deep dives.

This will propel the company forward, and one day I will find ways to repay your kindness - many thanks ❤️

No BS AI/ML Content | ML Engineer with a Plot Twist 🥷30M+ Views 📝

Done-for you Audience Building & Content Writing for AI/ML Founders that frees up 100% of your time, so you can focus entirely on building your company  
  
Is your work too important to ignore? 💌 https://tally.so/r/wk69VJ  (or shoot me a DM) 👉
  
It's 2025 and we're at the peak of the attention economy  
  
Where Authenticity is king, visibility is currency and people wanna buy from people

Not faceless dehumanized brands  
  
So, they told you LinkedIn was "free marketing" you slapped some posts and called it strategy  
  
You've poured hours crafting the 'perfect' post, only to watch it sink without a trace..  
  
Then it dawned on you, your technical brilliance isn't enough to draw a crowd  
  
Hell, maybe you've even tried mimicking influencers, scrambling for any traction..  
  
Your post got 500 f*cking views... from your friends...  
  
Being ignored is soul-crushing  
  
It's not your fault tough  
  
When you started your company, you didn't sign-up for a degree in an Audience Building, Copywriting and LinkedIn strategy 

And here you are, sacrificing yet another late night for your vision  

"Am I just sabotaging my efforts?"  
  
"Is my life-work meaningless?"  
  
Getting traction feels impossible  
  
You hoped hiring some marketer would save your weekends but..  
  
Tech is a code they can't crack  
  
They can only churn out the lame, soul-sucking crap we all ignore on sight  
  
With armies of founders flooding feeds, your voice dissolves into meaningless white noise  
  
But in the process, you've wasted more of your precious time explaining 101 tech concept and rewriting low-quality drafts 

Done being ignored?  
  
Let's make the world listen 
  
I can help you with:  

✒️ No BS AI/ML Content  
⭐ LinkedIn Profile Revamp
🎯 Website Copy Optimization  

I've done it before:

↳ For founders backed by YC, a16z, Sequoia, and Techstars
↳ For bootstrappers in LLMs, ML, Vector DBs, Data Analytics, and Engineering
↳ For AI and Machine Learning leaders at Google, Microsoft, Airbnb, and YouTube

If they could rise above the noise, so can you

Yeah, even if you're an introvert with a "boring" niche

Still with me?  
  
Amazing—You're amazing!  
  
Cheers,  
Paolo

My booking link: 💌 https://tally.so/r/wLW0oz (or shoot me a DM) 👉

Paolo Perrone

NVIDIA just made their best AI models free.

99% of developers have no idea.

While you burn cash on OpenAI credits...
NVIDIA open-sourced Nemotron.
Their entire agentic AI family 🤯

Free Models That Actually Ship:
• Nano: Runs on your laptop
• Super: Single GPU beast
• Ultra: Multi-GPU monster

What They Do:
→ Graduate-level reasoning
→ Math without hallucinations
→ Code that compiles
→ Visual reasoning that works
→ Tool calling. No loops.

The genius part:

They published EVERYTHING.
Models. Datasets. Techniques.
Actually open source.

Game Changers:
• 6x faster on new hardware
• No API limits
• Runs on a Raspberry Pi
• Commercialize it. No strings.

The uncomfortable truth:

You're waiting for GPT-5.
Smart teams ship with Nemotron.

Who's using it:
- Accenture
- CrowdStrike  
- ServiceNow
- Deloitte

The kicker:

Download from Hugging Face. Now.
Deploy tomorrow.
Zero license fees.

Forever.

They built enterprise NIM too.
(That costs money. Models don't.)

Still paying for API credits? 

♻️ Repost before OpenAI charges per thought

No BS AI/ML Content | ML Engineer with a Plot Twist 🥷50M+ Views 📝

I just finished Google's biggest AI Agents course ever.

250+ pages of production insights. 100% free.

Here's everything they released:

DAY 1️⃣ INTRODUCTION TO AI AGENTS (50 pages)

Most agent demos fail in production.
This covers the 5 levels of AI agents, from basic reasoning to self-evolving systems, plus security and architecture patterns that scale.
🔗 https://lnkd.in/gN9JaXgR

DAY 2️⃣ AGENT TOOLS AND MCP (50 pages)

MCP servers can add new tools without your approval.
Your poetry bot suddenly gains purchasing power overnight.
This breaks down tool design, the Model Context Protocol, and hidden security risks.
🔗 https://lnkd.in/gRWDSjBr

DAY 3️⃣ CONTEXT ENGINEERING AND MEMORY (70 pages)

Agents with zero long-term memory treat every session like meeting you for the first time.
Learn how to build memory systems that actively curate context, not just save conversations.
🔗 https://lnkd.in/gbgWJ4kH

DAY 4️⃣ AGENT QUALITY (50 pages)

Traditional QA doesn't work for agents. You can't unit test judgment.
This covers the 4 pillars of agent quality (effectiveness, efficiency, robustness, safety) and how to evaluate reasoning.
🔗 https://lnkd.in/g6DHcZDx

DAY 5️⃣ PROTOTYPE TO PRODUCTION (40 pages)

Building an agent takes 2 weeks.
Trusting it in production takes 6 months.
Learn evaluation gates, circuit breakers, and evolution loops that close the gap.
🔗 https://lnkd.in/gYTakhAK

Huge props to Antonio Gulli and Shubham Saboo for making this happen, and make it 100% free

250+ pages. Zero cost. All production-focused.

♻️ Repost to save a fellow AI engineer 6 months of trial and error

Sebastian Raschka, PhD just exposed the biggest secret in AI.

After dissecting 9 flagship LLMs, he found something nobody wants to admit:

We haven't invented a new architecture in 7 years.

The evidence is damning:
• DeepSeek V3: 671B parameters, still a transformer
• Kimi K2: 1 TRILLION parameters, still a transformer
• GPT-OSS: OpenAI's latest, still a transformer
• Llama 4: Meta's flagship, still a transformer

Every "breakthrough" model?
Same skeleton. Different makeup.

Here's what's actually happening:

DeepSeek V3/R1: Compressed attention (MLA) + expert routing (MoE)
Gemma 3: Sliding windows to fake longer context
SmolLM3: Removed position embeddings entirely (!)
Kimi K2: Just made DeepSeek bigger

The brutal reality:

• No new architectures since "Attention is All You Need" (2017)
• Every innovation = efficiency hack
• We're polishing, not inventing

671 billion parameters.
1 trillion parameters.
Same. Damn. Architecture.

While everyone's chasing parameter counts, Sebastian mapped the actual mutations:
• MoE: Use 37B of your 671B parameters
• Sliding attention: Pretend you have infinite context
• MLA: Compress your KV cache and pray

This isn't innovation.
It's optimization theater.

The most honest technical analysis you'll read this year.

Read it. Then ask yourself:
If we can't innovate past transformers, what happens when we hit the parameter wall?

♻️ Repost if you appreciate actual technical depth over hype

AI agent bootcamps charge $2,000 for this. You're getting it for free.

1️⃣ Design (6+ hours)
→ Langflow docs: docs.langflow.org
→ Travel planning agent template: https://lnkd.in/gmbWX5Mg
→ Agentic design patterns by Andrew Ng: 

2️⃣ Build (22+ hours)
→ LangGraph course: https://lnkd.in/g7raxjpC
→ LangSmith (modules 1-3): smith.langchain.com
→ Evaluation: https://lnkd.in/ghE-8EwA
→ Prompt engineering: 4 Methods of Prompt Engineering

3️⃣ Deploy (35+ hours)
→ LangChain Academy (modules 1-6): academy.langchain.com
→ MLOps specialization: https://lnkd.in/gmeb53UP
→ System design fundamentals: YouTube playlist
→ Distilabel for synthetic data: https://lnkd.in/gd_jJaJE

63 hours. $0.
The $2,000 bootcamp teaches the same stack with better marketing.

💾 Save this. It's the curriculum.
♻️ Repost to save someone $2,000.

The AI bubble is 75% of US GDP growth.

Read that again.

I spent the weekend analyzing Deutsche Bank's latest report.
What I found made me cancel my AI startup plans.

The dot-com bubble was 15.6% of GDP when it popped.
AI is already 5x bigger. And growing.

Here's the brutal reality nobody wants to admit:

1️⃣ Every AI Company Is Bleeding Cash
• OpenAI: Losing money on every $200/month subscription
• Break-even price: $2,000/month (10x current)
• 2026 projected loss: $14 billion
• 2029 projected loss: "Hundreds of billions"

2️⃣ AI Infrastructure Is a Money Pit
Data centers need $2 trillion annual revenue to break even.
Actual AI industry revenue: $390 billion.
That's a $1.6 trillion gap. Every. Single. Year.

3️⃣ AI Adoption Is Dead in the Water
• 95% of AI pilots fail (MIT study)
• AI coding tools make devs SLOWER (METR report)
• 77% of workers say AI increased their workload
• Adoption rates missed predictions 2 years running

The uncomfortable truth:

This isn't 1999. When the dot-com bubble popped, e-commerce was real.
It bounced back.

AI profitability is mathematically impossible.
The efficient compute frontier guarantees it.
Floridi's Conjecture proves agentic AI can't exist.

When this pops, there's no bouncing back.

What's keeping it alive? Your 401k.
70% of US venture capital is flowing into AI.
Half the S&P 500's growth is just tech companies building data centers.

Remove AI spending and US growth is already at zero.
We're IN the recession. AI is just hiding it.

♻️ Repost before your retirement fund discovers math

Wanna build AI agents? Stop watching random tutorials.

Here's the actual learning path:

1️⃣ Design (6+ hours)
→ Langflow intro: langflow.org
→ Multi-agent workflow patterns
→ Build: Travel planning multi-agent system

2️⃣ Build (22+ hours)
→ LangGraph course: https://lnkd.in/g7raxjpC
→ LangSmith for tracing: smith.langchain.com
→ Evaluation metrics and prompt engineering

3️⃣ Deploy (35+ hours)
→ LangChain Academy: academy.langchain.com
→ MLOps fundamentals
→ LangGraph Cloud deployment

Total: 63 hours. If you follow the order.
6 months if you don't.

💾 Save this.  Start at Phase 1.
♻️ Repost if you've ever jumped straight to deployment and regretted it.

90,000 AI papers published on Arxiv.

Ilya Sutskever says you only need 27 to understand 90% of what matters.

I just went through his list. Here's what shocked me:

What's NOT on the list:
❌ No RAG papers (everyone's obsessed with retrieval)
❌ No prompt engineering papers (despite the hype)
❌ No LangChain or agent frameworks
❌ Nothing about fine-tuning optimization
❌ Zero papers on "emergent abilities"

What IS on the list:
✅ "Attention is All You Need" (obviously)
✅ "The Unreasonable Effectiveness of RNNs" (2015, still matters)
✅ "Scaling Laws for Neural Language Models" (this is the one)
✅ "Kolmogorov Complexity and Algorithmic Randomness" (wait, what?)

The pattern I noticed:

1️⃣ Ilya's not telling you to read papers about using AI.
He's telling you to read papers about how intelligence works.
Three categories dominate his list:
Fundamentals (the architecture breakthroughs)
Attention mechanisms
Residual connections
LSTMs and sequence modeling

2️⃣ Scaling (why bigger actually works)
Scaling laws
Pipeline parallelism
Training dynamics

3️⃣ Compression (the theory underneath)
Minimum description length
Kolmogorov complexity
Information theory

Here's the controversial take:
If you're only reading papers from 2023-2024, you're studying tactics.

Ilya's list is strategy:

"Attention is All You Need" is from 2017.
"Understanding LSTM Networks" is from 2015.
"Deep Residual Learning" is from 2015.
The foundations haven't changed.
We're just scaling them.

The most surprising paper on the list:

"The First Law of Complexodynamics"
It's not even about neural networks.
It's about how complex systems evolve.

This is what separates engineers who use AI from engineers who understand AI.

If you only read 5:

1) Attention is All You Need (transformers)
2) Scaling Laws for Neural Language Models (why GPT works)
3) Understanding LSTM Networks (sequence modeling)
4) Deep Residual Learning (why deep networks train)
5) Kolmogorov Complexity (the theory of intelligence)

♻️ Repost to save someone from FOMO-reading 90,000 papers

There's a special hell in AI/ML engineering.

It hits around Year 2.
I call it the "valley of almost useful."

You're past "Hello World" tutorials.
Not good enough for real jobs.
Stuck building demos nobody uses.

The daily torture:

Morning: "Today I'll implement that paper"
Noon: Your embeddings are somehow 7-dimensional
Evening: Stack Overflow has zero answers
Night: "Maybe I should go back to web dev"

The worst part?

You know just enough to be dangerous.
Just not enough to be valuable.

🔥 The 3 Levels of "Almost Useful"

Level 1: The Overengineering Phase
Every problem needs a neural network.
CSV parsing? Better use GPT-4.
Your hello world has 17 dependencies.

Level 2: The Imposter Syndrome Pit
Everyone else seems to ship AI products daily.
Your Twitter feed is pure pain.
"Just deployed my 5th AI startup this week!"
Meanwhile, you can't get embeddings to work.

Level 3: The Clarity Moment
You realize 90% of "AI Engineers" are just prompt engineers.
Most production AI is boring Python scripts.
The magic isn't in the model - it's in the problem.

Here's what got me through the valley:

1️⃣ Ship broken things
My first production model had 67% accuracy.
Still saved the company $200K/year.
Perfect is the enemy of deployed.

2️⃣ Join a startup, not FAANG
Startups let you touch everything.
Break things. Learn fast. Ship faster.
Big tech will teach you one tiny slice.

3️⃣ Build boring solutions
Stop chasing cutting-edge.
Most businesses need basic classification.
Not your 12-layer transformer architecture.

4️⃣ Document everything
Future you will thank present you.
Especially when debugging at 3 AM.
"Why did I set embedding_dim=7?"

5️⃣ Find your valley friends
The loneliest part is thinking you're alone.
You're not. We all went through this.
Some of us just hide it better.

The uncomfortable truth:

The valley of almost useful is necessary.
It's where you learn that AI isn't magic.
It's engineering.

And engineering is mostly debugging.
And crying.
But mostly debugging.

Still in the valley?
Good. That's where real engineers are made.

Not in bootcamps.
Not in courses.
In the trenches, shipping broken code that somehow works.

Welcome to the club.

P.S. That senior engineer who seems to know everything? They googled it 5 minutes before the meeting. We're all just better at pretending now.

♻️ Repost to help someone realize they're not alone in the valley

🚨 Update: 800+ comments. My DMs exploded.
Can't reply to everyone. 
Grab it here: https://lnkd.in/eMQn_X-6
-------------------------------------------
Hamel's AI Evals toolkit disappears in 24 hours.

This isn't publicly available.

4 resources from the guy who:
→ Trained 2,000+ engineers on evals
→ Consulted for OpenAI, Google, Meta
→ Delivers high-value engagements for top-tier companies (ask dbt Labs why)

What you're getting:

📇 12 Evals Flashcards — the visual system for memorizing what matters
📘 The Complete Evals Guide — 144 pages of everything you need
📚 The RAG Evaluation Bible — retrieval metrics that actually work
📧 30-Day Email Course — builds evals that predict user satisfaction (free)

You won't find this on Google, GitHub, or HuggingFace.

It's in your DMs. If you act fast.

Last time I did this, 4,173 people commented. I can't DM that many again.

Comment "EVALS" + Repost ♻️ and I'll send all 4

24 hours. Then it's gone.

I pushed every agent framework until it broke. Most broke fast.

CrewAI vs LangGraph vs Agno vs PydanticAI vs Atomic Agents...
Plus SmolAgents, Mastra, and AutoGen.

The brutal truth:

Most devs hate the popular ones.

CrewAI and AutoGen? "Tricky to debug"
LangGraph? "Overly complicated abstractions"
SmolAgents' CodeAgent? "Cool idea, doesn't always work"

What actually matters:

🎯 Control vs Agency
Some frameworks let LLMs run wild (AutoGen)
Others keep them on a tight leash (Atomic Agents)
Guess which ones actually ship to production?

🛠️ Abstraction Level
High abstraction = Quick start, debugging nightmare
Low abstraction = More code, but you know what broke
Pick your poison.

💀 The Learning Curve
LangGraph: Steep as hell, but scales
CrewAI: Easy start, hits a wall
PydanticAI: Write everything yourself

The uncomfortable truth:
"AI agent frameworks are some of the worst forms of abstraction"

But here's what actually works:
• Agno for developer experience
• LangGraph for multi-agent at scale
• PydanticAI when you need control
• CrewAI if you're just prototyping

Half these frameworks won't exist in 2 years.
The winners? Whoever makes debugging possible.

Saved you 2 weeks of testing. You're welcome.
♻️ Repost if you've been burned by agent framework promises

a16z just published their AI coding report.

The math blew my mind:

AI coding costs $10k/year per developer.
But creates $100k+ in value.
That's $3 trillion in new GDP.

The entire economy of France. From a few Silicon Valley startups.

Here's what nobody's talking about:

Software development just shifted from pure people cost to people + compute cost.

Your junior developer in Bangalore?
Now competing with Claude at $10k/year.

The uncomfortable truth:

We're watching the end of cost-based outsourcing.

Not because AI replaces developers.
But because location doesn't matter when compute costs dominate.

I tested the numbers myself:

• 100k context window filled
• Claude Opus 4.1 reasoning mode
• 10k output tokens
• 3 queries/hour, 7 hours/day

= $10,500 annually

That's more than many offshore developers.

But here's the kicker:

The companies winning aren't cutting developers.
They're hiring MORE.

Why?

Because when productivity doubles, the math changes:
• 1 developer + AI = 2x output
• Cost per feature drops 50%
• Suddenly every project has positive ROI

The report shows clear winners already:

Cursor: $500M ARR in 15 months
Windsurf: $2.4B Google acquisition
Claude Code: Started a war with its own distribution channel

The pattern is obvious:

Lower barriers → More projects → More complexity → More developers needed

AWS didn't kill sysadmins.
WordPress didn't kill web developers.
AI won't kill software engineers.

It just changes what they do.

University CS programs?
Already obsolete.

Teaching algorithms when AI writes them faster?
Teaching syntax when AI handles implementation?
Teaching debugging when AI prevents most bugs?

That's like teaching calligraphy in the age of keyboards.

The real skill now:

Knowing when the AI is hallucinating.
Dragging it out of architectural holes.
Translating business needs into prompts.

The $3 trillion opportunity isn't in replacing developers.

It's in making every developer 10x more valuable.

While universities update their curriculums,
smart developers are already adapting.

Are you?

♻️ Repost if you think CS education needs a complete overhaul

I just watched Sebastian Raschka build the entire attention mechanism from scratch in 2 hours.

No abstractions. No "just use torch.nn.MultiHeadAttention."

Line-by-line Python until it clicks.

This is 's breakdown of how LLMs actually work.

Step 1️⃣ Basic self-attention (no weights) 
Shows you the raw math: how a model decides which words matter.
You see why "bank" in "river bank" gets different attention than "bank" in "savings bank."

Step 2️⃣ Add learnable parameters 
Now the model can learn what to pay attention to.
This is where it stops being a simple lookup and becomes intelligence.

Step 3️⃣ Causal masking 
The trick that makes GPT work: tokens can only attend to previous tokens.
Without this, the model cheats during training.

Step 4️⃣ Multi-head attention 
The breakthrough: run 8 attention mechanisms in parallel.
One head learns syntax. Another learns semantics. Another learns long-range dependencies.
This is why transformers work.

👉 What makes this different:

Most tutorials either:
Show you the math (incomprehensible)
Show you the PyTorch API (you learn nothing)

Sebastian does both.

You see the math implemented in code, step by step.

By the end, you understand why attention_weights = softmax(Q @ K.T / sqrt(d_k)) isn't magic.

It's just weighted averaging with learnable weights.

💡The moment it clicked for me:

Attention isn't complicated.
It's just "which words should influence this word's representation?"
Then you do that for every word.
Then you stack it 96 times (GPT-4).

That's it.

The Lovable for Data Analysis Just Dropped

Upload a file → Ask a question → Get your answer.

No Python, no SQL, no notebooks required.

Compare that to what you’re doing now: 

❎ Write SQL queries or Python scripts 
❎ Debug when they break
❎ Switch between notebooks, terminals, dashboards 
❎ Wait for a data scientist to help you

This agent actually understands your data:

✅ Sees your structure, knows how to query it 
✅ Maintains context across your workflow 
✅ Fixes its own errors
✅ Generates insights, not code you need to interpret

This is a big deal for anyone doing real data analysis work. 

You get the power of a data science team without needing to become one.

The best part?

I've got $15 in free credits for you to test it (no expiration date).

♻️ Repost + comment "GIVE ME MY WEEKENDS BACK 😭" and I'll send you the link

Traditional RAG: retrieve → generate → pray

Agentic RAG: retrieve → validate → retry → succeed

One is a pipeline. The other is a feedback loop.

I just built this Agentic RAG system with LangGraph + Milvus

It knows when it's wrong, and fixes itself

Here’s how:

Router: "Can I answer without retrieval?" 
✅ Yes: Skip to generation 
❌ No: Retrieve from Milvus

Grader: "Are these chunks actually relevant?" 
✅ Good: Generate answer 
❌ Bad: Rewrite query, loop back

Validator: "Does this answer match the sources?" 
 ✅ Yes: Return to user 
❌ No: Retry with different approach

It's a graph with feedback loops, not a one-shot pipeline.

The difference:
Traditional RAG commits to one path. No second chances.
Agentic RAG loops back, validates, catches mistakes, and retries until it's right.

You get a system that debugs itself before you see the output.

Why this stack:
LangGraph = control flow primitives (routing, looping, branching) 
Milvus, created by Zilliz = production-grade vector search at scale
Together: RAG that thinks instead of guesses.

Full implementation 👇

This is the RAG architecture pattern that will become standard.
♻️ Repost to save a fellow engineer from another week of RAG debugging

BFCM isn't won by the brand with the best products.

It's won by the brand that sees what's breaking first.

Most operators spend BFCM like this: 

9 AM: Campaign's crushing it, scale to $10K/day
11 AM: ROAS drops 40%, nobody knows why
2 PM: Still diagnosing
5 PM: $6K burned on a broken campaign

The strategy wasn't wrong. 

You just couldn't see the problem fast enough. 

Diagnosing "why is ROAS dropping?" right now: 

→ Open Meta Ads Manager
→ Check Shopify analytics
→ Pull GA4
→ Cross-reference Klaviyo
→ Export to spreadsheet
→ Try to connect the dots

20 minutes later you've burned $3K.

Triple Whale + Moby give you real-time answers from your complete data stack. 

👉 https://lnkd.in/gtgB8GJw

Moby has real-time access to all of it. 

Ask Moby: "Why did ROAS drop?"

Get: "iOS campaign CTR crashed 67% after creative refresh. Top SKU went out of stock an hour ago. Traffic's converting to lower-margin backup products."

Not a guess. The actual answer.

One brand caught this: 

The campaign was showing 187% ROAS. 
Moby a flagged 41% return rate. 
Actual ROAS after was returns: 73%. 

They were about to scale into a loss.

During BFCM, 20 minutes to diagnose = thousands in wasted spend. 

The brands that win can see problems in seconds, not hours. 

And yes, I am a Triple Whale partner.

👉 https://lnkd.in/gtgB8GJw

♻️ Repost if you've burned budget figuring out what went wrong

taplio

The Battle

Armand Ruiz

Paolo Perrone

Followers

Posts

Engagements

Likes

Performance ScoreArmand Ruiz's LinkedIn Performance

Performance ScorePaolo Perrone's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Armand Ruiz

Paolo Perrone

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Armand Ruiz's LinkedIn Performance

Performance ScorePaolo Perrone's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreArmand Ruiz's LinkedIn Performance