Founder @ Drilldown | Custom-designed business dashboards for B2B SaaS that deliver insights in one click, eliminating the need for complex spreadsheets in just 60 days.

What once felt simple and under control, now runs on many moving parts...

Marketing channels are disconnected from sales, LTV:CAC ratio is a rough guess, and your churn data is a month old.

‍The data to make sense of it all is scattered across Stripe, Hubspot, and product analytics, leaving your team to manually piece it together.

The result?

A collection of outdated reports that no one fully uses.

You’ve tried to fix it, asking people to update spreadsheets, buying new tools, maybe even hiring a specialist. But nothing sticks. Your own engineers are brilliant, but they’re busy building the product customers pay for.

In your mind, it’s simple: you just want accurate numbers so you can stop second-guessing and make the right calls.

At this stage, you face a choice.

You can continue guessing which free users will convert and which sales activities actually close deals, all while your teams debate why customers churn.

Or, you can choose a path to regaining control. The path where you:

- Increase conversion rates from trial to paid users.
- Attract customers who pay more and stay longer.
- Eliminate data debates between your teams.
- Truly understand why churn happens.
- Slash the time it takes to close deals.

We build the system that delivers these results, turning your scattered data into efficient, repeatable ARR growth.

The first step is a conversation. During a 30-minute call, we talk through your situation and what you're hoping to achieve. We'll be honest about whether we're the right fit to help. Our promise: you'll walk away with a fresh perspective, no strings attached.

Check our website for more info: www.datadrilldown.com

Thomas Varekamp

It took me 2+ years to figure this out.

And I still find ways to optimize every week.

Most B2B SaaS companies overcomplicate their data.

They scrap together some open-source tools + write tons of custom code.

A fun project for the engineer.
A nightmare for the business.

Because as soon as you scale:

→ engineers have other priorities.
→ dashboards break and stay broken.
→ maintenance takes hours every week.

The myth: data tools are expensive.
The truth: you can run a complete stack for <$100 /mo.

An engineer spending 1 hour every month on a data is more expensive than using tools.

So I always recommend to keep it simple: minimal code and managed by a single person.

Here is the toolkit we use to make that happen:

1. Extract Data
  
→ Don't write code, use managed connectors.
Examples: Fivetran, Airbyte, Stitch

2. Store Data

→ Centralize everything in one warehouse.
Examples: BigQuery, Snowflake

3. Transform Data

→ Define your logic once, not inside every dashboard.
Examples: Dataform, dbt Labs

4. Visualize Data

→ The final layer for the business.
Examples: Looker Studio, Power BI, Tableau

You don't need complex engineering for business analytics.

You just need the right tools.

Drilldown | Helping B2B SaaS leaders regain control by turning data into direction.

Landed in Bali last week. 

Mixed feelings on this place so far.

What I like:

I found a spot called Omni, a combined co-working, wellness, and gym.

It's basically a hub to get tons of work done with no distractions:

- solid gym
- top-notch service.
- great, healthy food.
- great co-working space.

Everyone is limited by time and energy. This place lets you really maximize both by having everything in one spot.

Also, prices are about 50% cheaper than in The Netherlands. Big arbitrage.

So, here’s what I don't like about Bali (so far):

1. The traffic is crazy. You literally cannot walk here. No pedestrian paths, just cars and tons of scooters.

2. Weird inequality. The moment you step outside of your nice co-working space, you see run-down buildings, broken roads, and trash. 

People that have been in Bali, any recommendations?

This weekend I’m going to explore the Gili Islands, excited for that one!

This is Mex, a Dutch entrepreneur I met here in Bangkok.

We bumped into each other at a club, and I instantly recognized the Dutch 🇳🇱.

A few days later we met for coffee, and that’s when he told me his story.

It all started years ago with Minecraft.

In Minecraft, you can create custom servers, basically private worlds with your own rules. 

Mex originally built one just for his friends, but after a while opened it to the public. The server was so good that people kept joining.

For the first 4 years it was purely a passion project. 

But as the community grew, Mex figured out how to turn this into a business.  

He began offering:

- custom game modes.
- memberships with extra privileges.
- custom skins for players to show off.

The model worked. The server grew steadily at around 20% per year. Then, 2 major events caused an explosion:

1. The pandemic
  
With everyone stuck at home, growth shot up to 100% year-over-year.
  
2. The Minecraft Movie
  
Its release brought in a whole new wave of players.

He scaled up, built a team, and hired an operator to manage the day-to-day.

After a while, he acquired another Minecraft server from a competitor (he told me they finalized the deal in a McDonald's lol)

Now he lives in Bangkok and runs the business remotely.

This is why I love talking to other entrepreneurs. 

You find out about businesses you didn’t even know were a thing.

A dashboard full of metrics means nothing if no one knows what they mean.

At many B2B SaaS companies we work with, this is common:

→ data is stored in some spreadsheets.
→ definitions are not 100% clear.
→ people don’t use the data.

The result: another data projects fails.

We solve this by creating a Metrics Dictionary.

It’s a simple Google Sheets that makes sure we all know what each metric means.

(this makes working with data a lot easier) 

Here’s what we include:

1. unit → what it's measured in
2. filters → how to slice the data
3. calculation → how it's calculated
4. department → who cares / uses it
5. metric owner →who is accountable
6. refresh rate → how often it updates
7. description → what the metric means
8. data source →where the data comes from

I included the template we use in the comments.

Copy it, tweak it, use it.

Good luck!

I'm launching a Youtube channel.

In the last 3 years, I’ve helped 20+ B2B SaaS companies build their data systems and create content about it.

Linkedin is great, but it's hard to show how to actually fix a broken ARR calculation in a 60-second clip.

Youtube allows me to go much deeper.

The goal is the same: help B2B SaaS companies regain control by turning their data into direction.

Topics:

- how to use data for ARR growth.
- how to align your team using data.
- how to operationalize B2B SaaS metrics.
- how to analyze your B2B SaaS business.

Basically everything a B2B SaaS leader needs to build a data-driven business.

The first video is live: https://lnkd.in/egPNPSs5

I show the exact data intelligence system we set up for most of our clients.

The plan: upload every week.

Check it out and let me know what you think!

The past 12 months were pretty stressful.

I’ve been in Asia for 2 months now, thinking a lot about what I actually want from work and life.

My goal with business has always been simple: work when I want, on projects I care about. 

Being here made me realize I’ve more or less reached that point. 

The money isn’t endless, but it’s enough for now, and I don’t need much to enjoy my life. I know being able to say that is a privilege, and I’m really grateful for it.

That made me ask: if I could design a day I’d never get bored of, what would it look like?

(still working on the answer)

Back in the Netherlands I was always in my head about the future, next milestone, next project, skipping fun because “work had to be done.” 

Here, I’m more in the moment. 

Most days I feel relaxed, rested, and creative. I follow more of what I 'feel' like doing instead of what I think I 'must' do.

It’s freeing, and it’s brought up a lot of emotions I’d pushed away. I keep having these random moments of: “Wow, life is actually pretty nice,” lol.

Not sure why I’m sharing this, just felt right.

Maybe it’ll resonate with someone.

Enjoy your weekend.

10 dashboards we’ve built for GTM teams to fix deal pipelines, missed targets, and slow leads.

Over the past few years, we’ve built many custom dashboards for B2B SaaS companies in the $2M–$10M ARR range. 

Below are 10 dashboards we have built for marketing & sales teams to make better decisions:

1. Deals Pipeline
2. Lead Lifecycle
3. Sales Rep Overview
4. Sales Funnel
5. Campaign Performance
6. Sales Overview
7. Target Account Overview
8. Advertisement Performance
9. Website Insights
10. Lead Journey

Scroll to see what each dashboard looks like ↓

Marketing__Sales_Dashboard_Showcase

I fixed a $3.2M ARR B2B SaaS's broken data system

(here’s how)

There is a dangerous transition for B2B SaaS companies around $2M ARR.

In the early days, data is a "favor."

You ask an engineer to pull a quick CSV. They write some code. You get the data.

But as the business grows, that method breaks:

- volume gets bigger.
- questions get harder.
- data is more scattered.

You move from needing a quick export to needing insights on daily basis.

And the engineer who used to run those scripts?

- they are not specialized in analytics.
- they are busy building the product.
- or worse: they left the company.

This creates a massive bottleneck.

And was exactly the case for a $3.2M ARR B2B SaaS I worked with last year. 

The CEO was waiting days for basic insights, and the data pipelines were impossible to maintain.

You can’t lead a company if your data is outdated or incorrect.

I just uploaded a video of how I completely overhauled the data infrastructure for a $3.2M ARR B2B SaaS.

In my new video on Youtube, I break down:

- the specific 15 dashboards I built.
- how I reduced data processing time by 98%. 
- how I aligned the teams on the right metrics. 
- how to drill from ARR to customer details in 2 clicks.

Check it here: https://lnkd.in/eAEps9T7

Our 5-dashboard system that gives B2B SaaS teams 80% of the insights they need.

Many B2B SaaS use data in:

1. spreadsheets
2. some pdf slides
3. a handful of dashboards

But this spread-out approach makes it hard to pinpoint exactly what’s driving growth, or what’s causing churn.

We fix this by building a 5-dashboard system.

It's designed for one purpose: give you complete visibility across everything that leads to ARR:

1. The ‘What’ (Management Overview)

Your high-level Command Center. It's the executive summary of all your key subscription metrics

2. The ‘Why’ (Revenue Overview)

The deep dive on your revenue. The ARR waterfall chart is the most important part. It shows you precisely what's driving ARR change.

3. & 4. The ‘Where’ (Country and Industry Overview)

This is how you validate your GTM strategy. You drill down into a specific segment and see its exact performance.

5. The ‘Who’ (Account View)

The 360-degree view of a single customer. You see their entire MRR history , sales activity, health, and user activity, all in one place.

You can drill down from your top-level ARR number... to the 'Agriculture' industry that's underperforming... all the way down to the specific customer who left.

Leading to:

→ more confident decisions. 
→ better team collaboration.
→ fewer data discussions. 

I put these 5 dashboards in the document below.

Check 'em ↓

The 5 dashboards that cover 80% of the insights (B2B SaaS)

One thing I’ve loved about Bangkok is how easy it is for founders and remote workers to meet people.

There are events almost every single day.

You can literally go to an event on a Tuesday afternoon and meet 10+ people all working on interesting projects.

My favorite example: there’s a weekly co-working meetup hosted at IKEA.

Sounds random, but it’s pretty cool. About 10 people of show up every time. 

We work together, share ideas, and just get to know each other.

I’ve joined twice since arriving, and every time I leave with a couple of new connections and fresh perspectives.

If you are a remote founder / worker, you can find a ton of these events on the Meetup app. 

Highly recommend!

At some point, data starts living in too many places.

That’s when even simple questions like “What’s our churn?...” become surprisingly hard to answer.

Teams pull numbers from different systems, rebuild metrics in spreadsheets, and show up to meetings with dashboards that don’t match.

The result? 

Decisions made on bad data.

At this stage, aligning your data stack becomes essential. But knowing where to start isn’t always obvious.

So here’s the 3-layer data stack we frequently work with:

𝟭. 𝗗𝗮𝘁𝗮 𝗦𝗼𝘂𝗿𝗰𝗲𝘀
Where raw data lives across your business.

→ PostgreSQL:  product usage and event data.
→ Stripe: tracks subscriptions and billing activity.
→ HubSpot: contains commercial customer data.
→ Intercom: logs support tickets to capture feedback.
→ Google Analytics: measures website engagement.

𝟮. 𝗗𝗮𝘁𝗮 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻
Moves, syncs, and activates data across tools.

→ Weld: centralizes data from all sources.
→ Twilio Segment: sends events across apps.
→ Hightouch: pushes insights back into business tools.

𝟯. 𝗗𝗮𝘁𝗮 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀
Where raw data becomes actionable insight.

→ BigQuery: storage for all company data.
→ Dataform: models data into analysis-ready tables.
→ CompanyEnrich: data enrichment for segmentation.
→ Looker Studio: visualizes metrics for the team.
→ Python / Sheets: enables deep dives and prototyping.

If you're feeling the pain of inconsistent metrics or messy reporting, this structure might show you exactly what's missing.

And possibly solve the problem you've been trying to fix with spreadsheets.

It took levelworks 12 hours every month just to get accurate insights.

(this is Tim from levelworks, check out the video 👇)

This is a classic scaling problem: the most important data lives in manually updated spreadsheets. 

It’s hard to grow when you’re spending all your time just trying to get basic numbers right.

We stepped in and built them a real data foundation.

The result? 

In Tim’s words: “It basically saves us 2 to 3 hours a week.”

More importantly, the team’s conversations are now deeper, more strategic, and lead to higher-quality decisions.

And that’s the point: not pretty charts, but time saved and better decision-making.

Better data → better decisions → better business.

What got you to $2M ARR won't get you to $10M ARR.

Scaling from $0 → $2M is mainly about gut feel.

You're a small team. You're all in the same room. You're 100% focused on finding product-market fit. 

Making decisions on intuition works.

Scaling from $2M → $10M is about systems.

It's not that your gut feel is wrong. 

The problem is that it's not scalable.

You can't be in every sales call, every marketing meeting, and every product debate anymore. 

To scale, your team needs to make their own decisions.

That requires a better data system. 

One that 𝗮𝗹𝗶𝗴𝗻𝘀 everyone on what's important (dashboards) and 𝗲𝗻𝗮𝗯𝗹𝗲𝘀 them to find what's working (analytics).

We identified 𝟲 𝘀𝗽𝗲𝗰𝗶𝗳𝗶𝗰 𝘀𝗶𝗴𝗻𝘀 that tell you your current data stack is breaking.

Check the carousel below to see if you recognize them ↓

6 Signs It's Time to Upgrade Your Data System

6 Dashboards we’ve built for a B2B SaaS to give them end-to-end visibility. 

At Drilldown, we help B2B SaaS leaders regain control by turning data into direction. 

Below are 6 dashboards we built to help their Management, RevOps, Sales, and Customer teams make better decisions:

1. Management Overview
Single overview of the business.

2. RevOps 
Understand how revenue is generated.

3. Account Overview
All account-level info in one spot.

4. Self-Serve Funnel
How users move through the self-serve funnel.

5. Sales Deals Pipeline
Complete view of all deals.

6. Account Health Overview
Spot churn risks & growth opportunities.

Swipe to see them ↓

6 B2B SaaS Dashboards for End-to-End Visibility

Ecochain was struggling with a loss of control.

This is a classic growth problem: their most important data was not readily available.

It’s hard to lead a company when you don’t have a clear picture of what's happening.

We stepped in and built the dashboards + data infrastructure they needed.

The result?

As the CEO Frans-Willem put it:

"The moment I realized it was working was when we could compare the previous results with the current results and see the trends."

They now have the ability to show the results in a better way to the board and can see where they really have the challenges... and to get it resolved.

And that’s the point of data: not dashboards or nice charts. 

But improved understanding, control, and ultimately, better decisions.

The reason you have 20+ unused dashboards? Big chance you skipped some steps.

Here’s what happens:

- you needs insights. 
- you buy a BI tool. 
- you tell an engineer to "build dashboards"

A month later, there are 20+ outdated reports nobody uses.

If this is the case, you probably skipped some foundational work.

Here's the full 5-step system we use to build useful & accurate dashboards 👇

1. Choose metrics

Answer:

- what are your goals? 
- what KPIs track it?
- what are the metrics driving the KPIs? 

This helps to focus on the essentials. 

2. Create a metrics dictionary 

- write down what "ARR" means. 
- what's the calculation? 
- what's the data source? 

This document ends discussions about data. 

3. Extract, Transform & Load (ETL)  

This is the "lean" data stack we recommend for almost everyone:

- extraction: Fivetran / Airbyte / Weld
- warehouse: BigQuery
- transformation: Dataform
- visualization: Looker Studio

You don't need big, complex stack. This setup is cheap, scalable, and powerful.

4. Build dashboards 

- take your metrics dictionary
- start building charts for the metrics
- ask feedback from users

5. Integrate in daily work 

Data means nothing if no trust it.

- give every metric an owner.
- use the dashboard in your meetings.
- review your metrics definitions every quarter.

This is how you go from random dashboards to a data intelligence system that helps everyone.

'How seriously should we take our data?'

Usually, founders know they have a problem. They spend too much time in spreadsheets. They know it's messy.

But they don't know if it's the right time to fix it.

Here is my rule of thumb:

𝗣𝗵𝗮𝘀𝗲 𝟭: $𝟬 - $𝟮𝗠 𝗔𝗥𝗥
Don't invest too much in analytics.

Your Stripe dashboard + spreadsheet is fine.
You are just trying stuff out. 
No need to optimize yet.

𝗣𝗵𝗮𝘀𝗲 𝟮: $𝟮𝗠+ 𝗔𝗥𝗥
This is the tipping point.

You have product-market fit. 
You probably have +30 team members.
You aren't running the day-to-day anymore.

But most importantly: you have a new management team.

These people don't have the historical founders knowledge you have.

When they join, the first thing they usually ask is: "Where do we stand? What are the metrics?"

That's when it's time to invest in better data.

If you want them to lead, you have to give them the data to run the business.

Here's how Ecochain finally got insights they can trust.

Ecochain helps businesses measure their environmental impact.

After setting ambitious growth targets, they faced challenges with data:

→ data scattered across different systems.
→ arguing about which numbers were right.
→ manually exporting data for every report.

You can't grow efficiently when you don’t have the right insights.

So we build their data foundation:

1. Start with a workshop (not a dashboard)

We aligned their teams on the right metrics to track (Trial Users, Conversion Rate, Churn).

2. Automated the data flow

We connected all their key tools (Stripe, Mixpanel, HubSpot) into a central data warehouse. 

This stopped manual exports.

3. Built dashboards

Once the foundation was set up, we built the dashboards. 

Now the data is updated every day and easy to work with.

The result?

→ teams aligned on one set of numbers.
→ faster access to custom data insights.
→ hours saved every week.

Check out the full case study 👇

Case Study Ecochain

A dashboard that takes 10 minutes to load is pretty much worthless.

Nobody will use it. 

This was the reality for Rentman after raising $22M.

They were ready to scale, but their data setup was holding them back:

→ dashboards took 10+ min to load.
→ data was scattered across many tools.
→ teams struggled to use the messy data.

So we came in to rebuild their infrastructure from the ground up.

Here is the 3-step process we used:

1. The deep clean

We worked with their CTO to overhaul BigQuery. We split "Raw Data" from "Production Data" to ensure stability and speed.

2. Smart data orchestration

We set up tools to transform the data automatically. No more manually typing numbers in spreadsheets.

3. Actionable dashboards

We connected all their data points (like NPS & Subscriptions) into single views. Then built fast-loading dashboards for the teams to use.

Now the data is live, reliable, and accessible every day.

The results:

→ instant data visibility.
→ marketing, Sales, and CS aligned.
→ No more waiting 10 minutes for insights.

Check out the full breakdown in the slides below ↓

Case Study Rentman

We recently built a data intelligence system for Expandi, a LinkedIn automation platform doing $10M ARR.

The challenge: 

They had BigQuery and dashboards, but teams were misaligned, with no central place to find all data insights.

We started with a workshop to agree on the top priorities.

Then we rebuilt their data foundations from the ground up:

→ enhanced customer data (30% to 90% coverage).
→ connected user behavior to company metrics.
→ implemented data checks with Slack alerts.
→ cleaned and centralized all data sources.
→ built dedicated datasets for the teams.
→ deployed 10 production dashboards.

The result: 

1. hours saved to focus on strategy instead of data.
2. better alignment between teams.
3. more efficient meetings.

Very happy how this one turned out.

Check the full case study below ↓

Case Study Expandi ($10M ARR)

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

The best AI engineers master PyTorch. The rest just import it.

Here's how to join the first group:

1️⃣ Master the Foundation
🔸 NumPy arrays → PyTorch tensors: https://lnkd.in/grmapDsK
🔸 Tensor operations that matter: https://lnkd.in/gjk3ErXu

2️⃣ Learn PyTorch Basics (Skip the Fluff)
🔸 60-minute blitz (actually 60 mins): https://lnkd.in/gU86ZGqt
🔸 What torch.nn really does: https://lnkd.in/gEmCGg6R
🔸 Daniel Bourke's 1-day crash course: https://lnkd.in/gCdBG_jc

3️⃣ Understand Autograd (The Magic Behind It All)
🔸 Karpathy's micrograd repo: https://lnkd.in/g_hTUmji
🔸 Karpathy's micrograd video: https://lnkd.in/gxM7h9E9
🔸 See gradients flow in real-time: https://lnkd.in/gKdwgAzz

4️⃣ Build Real Networks
🔸 Aladdin's playlist (underrated gem): https://lnkd.in/gP3k_rNx
🔸 NLP from scratch (no shortcuts): https://lnkd.in/g77inpPa
🔸 Adversarial examples (break things): https://lnkd.in/gFDzkep8

5️⃣ Production PyTorch (Where 90% Fail)
🔸 Distributed training without tears: https://lnkd.in/gdPZDnxB
🔸 Profile before you scale: https://lnkd.in/gfs7JQ6R
🔸 Mixed precision (2x speedup): https://lnkd.in/gikyETYK
🔸 Performance tuning secrets: https://lnkd.in/g8BE7J28

The uncomfortable truth:

Most "PyTorch experts" can't debug a CUDA out-of-memory error.
They copy tutorials. They call high-level APIs.

But when production breaks? Crickets.

Start with micrograd. Build intuition.
Then scale up. One tensor at a time.

What PyTorch concept finally clicked for you?

♻️ Repost to help someone skip the 3am debugging sessions

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.

Your AI agent has dementia. LangChain just dropped the cure: 

DeepAgents CLI with actual persistent memory.

This project solves the most infuriating problem in AI development, agents that forget everything the moment you close your terminal. 

It gives your AI assistant an actual memory that persists across sessions. 

Key Features:

✅ True persistent memory
Your agent writes markdown notes to ~/.deepagents/memories/ and reads them before every response. Like a human taking notes.

✅ File operations that work: 
Read, write, and edit files in your project with diff previews and human approval. No more "I can't access your filesystem."

✅ Shell command execution
Run tests, install packages, deploy code - all with your permission. Because we're not insane.

✅ Multi-agent personalities
Create specialized agents (backend-dev, frontend-wizard, devops-guru) each with their own memory and expertise.

Getting started is stupidly simple. 
Just install and run: pip install deepagents-cli && deepagents

Harrison said they built this because copy-pasting context 47 times a day is insanity (we agree).

Check out the code, see the full toolset, and contribute on GitHub!

♻️ Repost if you're done being your AI's memory bank

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

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

Plexe is a Python library that turns "I need to predict customer churn" into a working ML model.

No feature engineering. No pipeline code. Just natural language intent and your data.

It integrates with OpenAI, Anthropic, and Ray, making it easier to build sentiment analysis, fraud detection, and forecasting models in minutes instead of weeks.

Plexe doesn't just build the model - it generates a thorough evaluation report showing exactly where and why it performs well (or doesn't).

A powerful tool for anyone working at the intersection of product development and machine learning.

♻️ Repost if you've ever deployed a model and just hoped it worked.

🎁 Launch special: Get $20 free credits → https://console.plexe.ai
Code: LAUNCHDAY20

The most underrated AI agents video on YouTube has 155K views.

It should have 15 millions.

Dexter breaks down why coding agents actually fail in production.
Not theory. What broke when they shipped real agent systems to real software teams.

The line that changed how I think about agents:
"Everything is context engineering."

He shows:

→ Why naive back-and-forth prompting doesn't scale (and what does)
→ Spec-first development that keeps agents on track
→ Context compaction strategies for long-running workflows
→ How subagents handle complexity without exploding

No demo theater. Just hard-won lessons from production.

If you're building agents that need to work beyond the prototype, this is required watching.

Watch it: https://lnkd.in/e99smz5g

♻️ Repost if you've hit the "context limit exceeded" wall

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

I just watched 4 AI agents compete on the same coding task.

Then an AI judge picked the winner.

The best code became the pull request.

This is how coding agents should work.

BLACKBOX.AI AI just shipped Multi-Agent Execution.

Here's what happens:

You give one prompt: "Implement OAuth authentication"

4 agents run in parallel: 
→ Claude Code 
→ Blackbox CLI 
→ Gemini
→ Codex 

Each builds their own solution.
You watch all 4 logs side by side in real-time.

Then an AI judge evaluates: 
✅ Code quality 
✅ Error-proneness 
✅ Efficiency 
✅ Edge case 

The best one wins. That code becomes your PR.

Why this matters:
Every AI coding agent has blind spots.

🔹 Gemini: Ships fast. Weak debugger
🔹 Codex: Blazing speed. Loses context
🔹 Claude: Nails architecture. Struggles with debugging.

Running one agent = hoping you picked the right one.
Running all of them = letting the best approach win.

The insight:
We've been using AI agents like single players.

The future is ensemble.
Run multiple approaches. Compare. Select the best.

Still guessing which AI to use?
♻️ Repost to save someone from picking the wrong agent again.

taplio

The Battle

Thomas Varekamp

Paolo Perrone

Followers

Posts

Engagements

Likes

Performance ScoreThomas Varekamp's LinkedIn Performance

Performance ScorePaolo Perrone's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Thomas Varekamp

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);}Thomas Varekamp's LinkedIn Performance

Performance ScorePaolo Perrone's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreThomas Varekamp's LinkedIn Performance