With more than 25 years of experience in go-to-market and growth strategies, I have developed the "Spark Strategy" that has led to follow-up projects 100% of the time since we started using it. I love helping clients and partners develop fact-based, revenue-driven and innovative solutions to their problems.

My focus areas include the forward-looking fields of Trustworthy AI, Augmented Work, and Knowledge Management. These themes are integrated into my work to create sustainable and ethical solutions that optimize workflows and enhance knowledge management.

I have a proven track record in orchestrating and leading cross-functional, cross-location, multicultural teams in matrix organizations for developing transformative, cost-saving solutions and strategies in highly technical, complex and ambiguous situations for brands such as Microsoft, T-Mobile, General Electric, Siemens, BMW, Nestle, Loreal, Bayer, Saint Gobain and many others I cannot mention here (non-disclosure agreements!).

The greatest satisfaction for me comes from pioneering unique approaches that enable multidisciplinary and multicultural teams in flexible environments, in order to provide a user-centric, prototype-driven environment where they can find self-confidence in their creativity and be innovative like crazy.

The greatest satisfaction for me comes from pioneering unique approaches that enable multidisciplinary and multicultural teams in flexible environments, in order to provide a user-centric, prototype-driven environment where they can find self-confidence in their creativity and be innovative like crazy.

Christian Hubmann

Most people using Claude Code skip 80% of its power.

They type prompts. They get code. Done.

But there are 50+ slash commands that transform Claude Code from a coding assistant into a full AI development system.

The ones that changed how I work:

⚡ /init — Run this before anything else. Creates your CLAUDE.md file with persistent project memory. Tell it your stack, your rules, your preferences. Claude remembers everything from session one.

🧠 /plan — Forces Claude to think before it codes. Back-and-forth planning saves 10x more tokens than fixing wrong code later.

🍴 /fork — Branch the conversation. Try two solutions simultaneously. Nothing gets lost. Merge the best parts.

💬 /btw — Ask a side question while Claude is coding without interrupting its flow. Underrated.

✨ /simplify — Spawns 3 parallel review agents to clean up your code. One command, instant quality.

🔒 /security-review — Scans for vulnerabilities before you ship. Built into Claude Code. Most people never touch it.

🔁 /loop — Runs a prompt every X minutes. Watch a deployment, retry a failing test, monitor an API automatically.

The pattern I see: developers use Claude Code like a smarter autocomplete.

The people building real systems use the full command set.

Which commands are you actually using? 👇

Building your Corporate Digital Brain with Trustworthy AI and Knowledge Graphs, acting as a catalyst for augmented working and driving AI innovation. #stayAugmented #VibeCoding

No alternative text description for this image

Stop explaining your coding standards to Claude. Again. And again.
Claude Code gives you 5 features. Most developers use 1.

Here's the decision matrix that changes everything:

1️⃣ CLAUDE.md
→ Should Claude ALWAYS know it? → Put it here
Non-negotiable rules. Loads every session. Always on.

2️⃣ Skills
→ Should Claude know it SOMETIMES? → Make it a skill
PR checklists, deploy guides. Load only when needed. Zero context waste.

3️⃣ Sub-agents
→ Should this run in ISOLATION? → Delegate it
Their own context window. Research 100 files. Zero clutter in main.

4️⃣ Hooks
→ Should this happen AUTOMATICALLY on events? → Hook it
Auto-format. Block rm -rf. Run tests. Every time. Guaranteed.

5️⃣ MCP Servers
→ Does Claude need EXTERNAL tools? → Connect it
GitHub, Sentry, Postgres, Jira. One command. Done.

The #1 mistake: cramming everything into CLAUDE.md.
Your deploy checklist loads while you're fixing a typo. That's wasted tokens.

Context window is currency. Spend it wisely.



#ClaudeCode #AI #ProductManagement #DevTools #AIAgent #Productivity

Using AI in Europe without understanding GDPR is like driving blindfolded.

Eventually, you WILL crash. And fines go up to 20M EUR or 4% of global revenue.

Here are 6 things your legal team wishes you knew:

⚖️ 1. AI training data needs a legal basis
You can't just feed customer emails into ChatGPT. Every piece of personal data needs consent, legitimate interest, or another GDPR basis.

🔒 2. "Anonymous" data often isn't
If you can re-identify someone by combining data points, it's still personal data. Pseudonymization ≠ anonymization.

📋 3. You need a Data Protection Impact Assessment for AI
DPIA is mandatory for high-risk processing. Most AI use cases with personal data qualify. Don't skip this.

🌍 4. Data transfers outside the EU are a minefield
Using US-based AI tools? You need Standard Contractual Clauses, adequacy decisions, or equivalent safeguards. "We use AWS" isn't enough.

🤖 5. Automated decisions require human oversight
If AI makes decisions that significantly affect people — hiring, lending, pricing — you MUST offer human review. Article 22 is non-negotiable.

📊 6. The AI Act adds another layer
Since 2025, the EU AI Act classifies AI systems by risk level. High-risk systems need conformity assessments, documentation, and ongoing monitoring.

The good news: GDPR-compliant AI is absolutely possible. It just requires intentional design, not afterthought compliance.

Start with your DPO. They're probably already worried about this.

What's your biggest GDPR + AI challenge?

#GDPR #AI #DataPrivacy #Compliance #Enterprise

The skills that made you a great manager in 2020 won't be enough in 2027.

AI is changing what leadership looks like. Here are 7 skills every manager needs now:

🧠 1. AI literacy — not expertise
You don't need to code. You need to understand what AI can and can't do, so you make better decisions about when to use it.

🎯 2. Human-AI task allocation
The art of knowing which tasks to delegate to AI and which need human judgment. This is the new core management skill.

💡 3. Prompt design thinking
Not just writing prompts — designing workflows where AI and humans collaborate. Thinking in systems, not tools.

🤝 4. Psychological safety in uncertain times
Your team is scared AI will replace them. Your job: create safety, define the human value, and lead through the transition.

📊 5. Data-informed intuition
Use AI insights to inform decisions, not make them. The best leaders combine data with experience and empathy.

🔄 6. Rapid experimentation mindset
"Let's test it for 2 weeks"  "Let's have 5 meetings about whether to try it." Speed of learning is the new competitive advantage.

🌱 7. Continuous upskilling culture
If your team isn't learning, they're falling behind. Build learning into the workflow — not as a quarterly training event.

The leaders who thrive won't be the most technical.

They'll be the most adaptable.

Which skill are you working on? 👇

#Leadership #AI #Management #FutureOfWork #Skills

I've analyzed 500+ viral LinkedIn posts.

Here are 10 formats that consistently generate 10x engagement:

📝 1. The Contrarian Take
"Unpopular opinion: [thing everyone agrees on] is actually wrong." Triggers debate = comments = reach.

📊 2. The Number Cruncher
"I analyzed 10,000 [things]. Here's what I found." Data-driven posts get shared because they feel credible.

🎯 3. The Step-by-Step Guide
"How to [achieve thing] in [number] steps." Actionable = saveable = the algorithm loves saves.

💡 4. The Mistake List
"5 mistakes I made when [doing thing]." Vulnerability + lessons = trust + engagement.

🔄 5. The Before/After
"6 months ago I [situation]. Today I [result]. Here's what changed." Transformation stories are irresistible.

📋 6. The Checklist
"Use this checklist before [important thing]." People screenshot checklists. Screenshots = saves = reach.

🎭 7. The Hot Take + Nuance
Start with a bold statement. Then add "But here's what people miss..." Shows depth.

📈 8. The Case Study
"How [company] achieved [result] by [doing thing]." Specific > generic. Always.

🤔 9. The Question Post
"I'm curious: [genuine question about your industry]." Simple but powerful. People love sharing opinions.

🏆 10. The Resource List
"7 tools/books/frameworks that changed how I [thing]." Practical value = bookmarks = long-tail reach.

The secret? Pick 3 formats. Rotate them. Be consistent.

Which format works best for you? 👇

#LinkedIn #ContentStrategy #Marketing #SocialMedia #B2B

They copied a fly's brain into a computer.

The virtual fly just... started walking.

No training data. No reinforcement learning. No instructions.

Just a brain — doing what brains do. On new hardware.

Here's why this is one of the biggest AI moments in years:

👉 Eon Systems took the complete connectome of a fruit fly brain
(125,000 neurons, 50 million connections — published in Nature 2024)

👉 They didn't study it. They ran it.
Connected to a virtual body via physics simulation — and let it go.

👉 The result: walking, grooming, egg-laying behaviors.
Emerged naturally. From the wiring alone.

---

This is fundamentally different from everything we've built before.

Current AI: We feed models billions of words → they learn patterns.
This: We copy the actual wiring of a biological brain → it just works.

One is imitation. The other is the real thing — running on different hardware.

---

 The implications for AGI are wild:

Instead of training intelligence from scratch...
What if we just map it and copy it?

The fruit fly has 125K neurons. The human brain has 86 billion.

We're not there yet. But the blueprint exists. And now we know it can run.

Are you watching this? 🧠

\#AI \#AGI \#Neuroscience \#ArtificialIntelligence \#MachineLearning \#CDBrain
www.cdbrain.de

A new AI agent just dropped. And it solves the #1 problem nobody talks about.

Most AI agents get blocked.

When a bot tries to access a website from a data center IP — LinkedIn, booking sites, enterprise tools — the site blocks it. No browsing, no form filling, no automation.

Perplexity Computer figured out a different architecture.

🔵 HOW IT WORKS
Your prompt → Cloud Orchestrator (Claude Opus 4.6) → two separate VMs:

⚡ MicroVM: code, files, terminal. Boots in 125ms. Dies after session. Isolated IP.
🌐 Cloud Browser: residential IPs, browses like a human, never blocked.

Two machines. Different IPs. Neither looks like a bot.

🟠 SMART MODEL ROUTING
Perplexity Computer isn't locked to one provider.
→ Orchestration: Claude Opus
→ Research: Claude Sonnet
→ Coding: GPT-5.4
→ Multi-answer: Model Council (all top models at once, synthesized)

When a better model ships, they update the router. You always get the best tool for the job.

✅ WHAT IT'S GOOD AT
Deep web research. Browser automation. Clicking, form filling, navigating dynamic pages. 400+ integrations. Zero setup — no VPS, no API keys, no config.

❌ WHERE IT FALLS SHORT
$200/month with only 10,000 default credits — burns fast.
File system: multiple sessions struggle to find the same file.
Coding: Claude Code still wins, clearly.
Verdict from the video: "Still early. Wait 2 months."

─────────────────────────────────

Interesting parallel: open-source agents like OpenClaw or Agent Zero solve the same blocking problem differently — they run locally on your machine, using your residential IP. Free. Full control.

The question isn't "which AI agent is best."
It's "which architecture matches your use case and budget."

For research-heavy browser automation → Perplexity Computer.
For coding → Claude Code.
For full control → self-hosted.

The agent market is moving fast.
What matters is understanding the architecture, not just the brand.

→ Full breakdown: https://lnkd.in/dBBtVVBk

#AI #AIAgents #PerplexityAI #ClaudeCode #Automation #LLM #ProductivityTools #TechLeadership

Follow me to get insights on how to use AI at work and beyond. Join the world's biggest AI newsletter with 1M+ readers ↓

🚨 How to setup Claude Code for coding.

Most developers install it and wing it. That's why their results are average.

P.S. My team shares more engineering resources with 200K+ AI/ML Engineers everyday here: https://lnkd.in/dMGZuZAj

Here's the exact setup that actually works:

Step 1. Install Claude Code globally
→ Open your terminal right now.
→ Run: npm install -g @anthropic/claude-code
→ Needs Node.js 18 or higher.
→ Get your API key at console. anthropic .com.
→ Run claude to authenticate and start.

Step 2. Always start inside your project
→ Never run Claude from home directory.
→ cd your-project first, then claude.
→ Wrong folder means wrong context.
→ Context is everything here.

Step 3. Create your CLAUDE. md file
→ This is Claude Code's permanent memory.
→ Create it in your project root folder.
→ It reads this before every single session.
→ No CLAUDE.md = starting blind every time.

Step 4. Structure CLAUDE. md into four sections
→ Project overview in three sentences max.
→ Full tech stack with exact versions listed.
→ Naming and coding conventions you follow.
→ Files Claude should never touch.

Step 5. Run /init on every new project
→ Type /init on your very first run.
→ Claude scans your entire repo automatically.
→ It drafts a CLAUDE.  md for you.
→ Review it, edit it, make it yours.

Step 6. Give Claude the full execution loop
→ Don't just ask it to write code.
→ Say: write, run tests, fix failures.
→ One instruction, complete loop.
→ That's where real speed comes from.


Step 8. Connect MCP servers to your stack
→ Claude Code supports external MCP integrations.
→ Connect your database, docs, internal APIs.
→ Now Claude understands your live environment.
→ This turns useful into genuinely powerful.

Step 9. Break every task into checkpoints
→ "Build auth module" is too vague.
→ Try: "Create login route, then session handling."
→ Smaller scope means much better output.
→ You can iterate this in minutes.

Step 10. Always review the plan first
→ Ask: "show me your plan first."
→ Especially before any destructive changes.
→ Claude lays it out, you approve.
→ This one habit prevents 90% of mistakes.

Before your next session, check three things:
→ Are you in the right directory?
→ Does your CLAUDE. md exist, and is it updated?
→ Is your task scoped small enough to iterate?

Set these up first. Then code.

💡 The best product teams won't win because of instincts.

They'll win because of feedback loops.

Andrej Karpathy just released autoResearch — AI that experiments autonomously, logs its own learnings, and self-improves overnight.

100 experiments while you sleep. 0 humans in the loop.

In this week's newsletter I break down what this means for PMs, CTOs and everyone running A/B tests manually in 2026.

👇 Read it here:


#ProductManagement #AI #Innovation #autoResearch #cdbrain #cdb

View newsletter: Digital Product Management

There's a new repo on GitHub that made me rethink how I approach product optimization. And it's not from a startup. Andrej Karpathy — co-founder of Op

Open article: The End of Manual Iteration: Andrej Karpathy Just Changed How We Build by Christian Hubmann • 4 min read

The End of Manual Iteration: Andrej Karpathy Just Changed How We Build

I'm not a doctor. This is not medical advice.

But I stumbled across this video a few weeks ago, got curious, tried it — and honestly? Something felt different.

Saffron extract. 30mg. One dose before a stressful day.

The study in the video doesn't claim it lowers cortisol. It just delays the spike. Makes the stress response feel more like a slope than a cliff.

Same pressure. Just... more manageable.

Might be placebo. Might not be. Either way — worth knowing this research exists.

🎬 Link in the comments — judge for yourself.

What's your stress management go-to? Drop it below 👇

Your AI writes garbage code.

Not because of the model. Because you have no process.

Here are the 5 skills that changed everything for me:

1️⃣ Grill Me — AI interviews you relentlessly until real shared understanding
2️⃣ Write PRD — Document the destination, not the journey
3️⃣ PRD → Issues — Vertical slices that flush out unknowns first
4️⃣ TDD — Red. Green. Refactor. One test at a time.
5️⃣ Improve Architecture — Spawn parallel agents, pick the best interface

The pattern is simple:
Garbage codebase → garbage AI output.
Clean process → exponentially better results.

The engineers winning with AI aren't better at prompting.
They're better at engineering.

→ What does your AI dev process look like? 👇

👇 Full video in the first comment.

#AIEngineering #ClaudeCode #SoftwareEngineering #AIProductivity

AI created 1.3M new jobs last year. Great news, right?

Here's the part nobody talks about:

90% of meeting knowledge vanishes within 7 days. Employees waste 20-30% of their workday just *searching* for information. And Gen Z stays ~2.8 years before moving on — taking everything they learned with them.

So we're building an AI-powered workforce at record speed... on top of a knowledge foundation made of sand.

The real bottleneck isn't talent. It's that companies invest millions in AI infrastructure while the expertise that makes AI actually useful bleeds out quietly — through turnover, retirement, undocumented decisions.

Every departure compounds the debt. Every unrecorded meeting is institutional amnesia.

That's exactly why we built CDBrain's Knowledge Collector — to capture expertise *before* it walks out the door. Not as static docs collecting dust, but as structured, AI-ready knowledge with clear trust levels.

Because the companies that win the AI era won't just hire the smartest people. They'll be the ones who actually remember what those people knew.


#KnowledgeManagement #AI #FutureOfWork #CDBrain #EnterpriseAI

Your AI agent just learned to work while you sleep.

Claude Code shipped 4 updates this week that change what agentic workflows actually look like in practice. Here's what they do — and when to use each one:

─────────────────────────────────

🔵 /loop — Short-Term Automation
→ Natural language recurring prompts inside your session
→ "Every 10 min, check my inbox for urgent emails"
→ Runs silently. You don't touch it.
Limitation: Expires after 3 days + dies with the session. Think sprints, not habits.
📖 https://lnkd.in/dsrQ4-C4

🟢 Scheduled Tasks — Long-Term Workflows
→ Daily/weekly automations that run even when you forget
→ Fresh session every run, reads your project files, executes your skills
→ Catches up missed runs if your computer was off
This is the "set it and it runs every Monday morning" version.
📖 https://lnkd.in/dFGj-H75

🟠 Google Workspace CLI
→ Google just released an open-source CLI for Drive, Gmail, Calendar, Docs, Sheets, Slides
→ 100+ built-in recipes ready to use
→ One tool = Claude Code has your full Google ecosystem
Setup: just ask Claude to install it. Done in 5 min.
📖 https://lnkd.in/d3J8rcgj

🟣 Skills 2.0 — Eval & Improve
→ Stop guessing whether your skills work
→ Run automated tests, get HTML reports with pass/fail scores
→ A/B test: with skill vs without skill, side by side
→ Iterate until you hit 9/10. Then move on.
📖 https://lnkd.in/drDPYmGC

─────────────────────────────────

The pattern here:
/loop = help me right now (hours)
Scheduled Tasks = do this every week (forever)
Google CLI = use my tools
Skills Eval = make my tools better

Claude Code now runs on a schedule, accesses your workspace, and improves itself.

What are you automating first?

────
📌 Full video breakdown by @ShortcutChris:
youtu.be/F4zSxfBe5R0

#ClaudeCode #AI #Automation #AIAgents #Productivity #LLM #AnthropicAI

Most AI agents fail in production.

Not because the model is bad. Because you gave one agent too many jobs.

Here's what changes when you build an AI Agency instead 👇

The problem with solo agents:
→ Context window overflows on complex tasks
→ No specialization = mediocre at everything
→ One failure = full system down
→ Impossible to debug at scale

The Agency Swarm approach:
You structure AI like a real company.

↳ CEO Agent — the only one talking to the user. Plans, delegates, coordinates.
↳ Developer Agent — gets tasks from CEO, handles technical execution.
↳ Virtual Assistant — handles communication, research, admin.

Each agent has its own instructions, tools, and scope. Communication is strictly directional.

⁠ CEO → Developer → VA ⁠

No chaos. No circular calls. Just clean, auditable workflows.

What makes agency-swarm different:
✅ Built on OpenAI Agents SDK (production-grade foundation)
✅ Type-safe tools via Pydantic (no more silent failures)
✅ Explicit ⁠ communication_flows ⁠ — agents can ONLY talk to who you allow
✅ Works with Claude, Gemini, Grok via LiteLLM
✅ Persistent state across sessions (plug in your own DB)

This is the mental shift that separates toy demos from real AI systems:

Stop thinking in agents. Start thinking in organizations.

→ https://lnkd.in/dyCyRSib (open source, 8k+ ⭐)

---

What's your current multi-agent setup? Drop it below 👇

---

#AIAgents #MultiAgent #LLM #AgentSwarm #OpenAI #AIEngineering #Python

AI Automation Expert | Helping Businesses Eliminate Manual Work with Automations | Saved Clients $2000+/Month | 200+ Workflows Deployed | Building Smarter Businesses

The creator of Claude Code just shared how his team trains AI agents at Anthropic.

One file. CLAUDE.md. That's it.

Here's what goes inside it and why it works:

Boris Cherny places it at the root of every project. 

Claude reads it at the start of every session. No setup. 

No prompting. Just rules that stick.

The template he shared breaks down into 6 patterns:


1. Plan Mode Default: Claude enters plan mode for any task with 3 or more steps. If something goes wrong, it stops and replans instead of pushing forward.

2. Subagent Strategy: offload research and parallel work to subagents. Keep the main context window clean. More compute, better output.

3. Self-Improvement Loop: after every correction, Claude updates a lessons.md file. It writes rules for itself. Mistake rate drops over time automatically.

4. Verification Before Done: nothing gets marked complete without proof. Run tests. Check logs. Ask: would a staff engineer approve this?

5. Demand Elegance: for non-trivial changes, Claude asks itself if there's a cleaner solution. No over-engineering on simple fixes.

6. Autonomous Bug Fixing: give Claude a bug report and walk away. It uses logs, errors, and failing tests to diagnose and fix without you switching context.

The task system is just as clean. Plan in todo.md. Track progress. Explain changes at each step. Capture lessons after every correction.

The result? An agent that improves with every session because your past mistakes become its permanent rules.

This is not prompt engineering. This is agent memory done right.
If you're building with Claude Code and not using CLAUDE.md properly, you're leaving 80% of its power on the table.

What rules would you put in yours?

#ClaudeCode #AIAutomation #AgenticAI #AIFLOXIUM #BuildWithAI

HR leaders are losing critical capacity to repetitive enquiries that stifle strategic growth.

Left unmanaged, operational loads involving policy clarifications, leave entitlements, and payroll questions consume time better spent on workforce planning and capability uplift.

Knowledge management software acts as a strategic buffer. By implementing self-service solutions, organizations are achieving significant operational efficiency gains.

📊 Key Performance Metrics:

* A 40% reduction in HR administration through self-service ROI.
* Automated responses to repetitive policy and benefit queries.
* Real-time access to compliance and workplace strategy documentation.

Transitioning from reactive support to proactive strategy requires a centralized, accessible knowledge base. Reducing the administrative burden is no longer a luxury—it is a requirement for modern workplace compliance and effective strategy.

How is your HR team balancing the tension between operational administration and long-term workforce planning?

https://lnkd.in/djAUHyzX

#KnowledgeManagement #HRTech #OperationalEfficiency #WorkforcePlanning

I gave Perplexity Computer access to my portfolio & told my AI trader:

“Turn this into $10M. Make no mistakes.”

It wrote 69 strategies.
Generated 1,500+ research reports.
Spun up 19 new algos.

It scraped X, Reddit, Instagram, and LinkedIn.
Mapped sentiment.
Read every chart pattern it could find.

It traded 24/7.

It lost everything.

But boy was it beautiful.

P.S. on a serious note, check out how I Turned Claude in Excel Into My Senior Financial Analyst 📈: https://lnkd.in/gAD58P2V

Agree or disagree? Let me know in the comments :)

Nvidia CEO just said Agentic OS is the future 😳

“Every software company needs an OpenClaw strategy” - Jensen Huang

At GTC 2026, Jensen compared OpenClaw to Linux, HTTP, Kubernetes.

Not a tool - a new layer of computing.

And that’s the shift most people are missing:

We’re not upgrading software.
We’re replacing who uses it.

From humans clicking dashboards → agents running workflows.

Which means that SaaS isn’t becoming AI-powered.
It’s becoming agent-consumable.

That flips the entire model.

→ OpenClaw/AI agents don’t just answer - they act
(read/write files, run code, book meetings, open PRs)

→ They run 24/7 with memory + scheduled tasks
(not prompt → response, but continuous execution)

→ NVIDIA’s NemoClaw adds sandboxing + policy engines
(agents with guardrails = enterprise-ready)

→ Real workflows already chain agents together
(sales → scheduling → fulfillment, no human in the loop)

→ 22% of employees are already running agents at work
(most without IT approval)

This means that:

Your product is no longer the interface.
It’s a tool an agent calls.

Your moat is no longer UX.
It’s orchestration + security.

And your competition is no longer just another SaaS.
It’s someone else’s AI agent doing your job faster.

In the agent & machine economy, the interface will become optional.

And the winners won’t build better tools - they’ll build better agents.

P.S. check out how I built an AI operating system to run a startup with Claude 🤖: https://lnkd.in/dv7-ZRVc

At Product Masterclass, our team empowers CEOs and product leaders to adopt a product mindset, accelerate market delivery, and drive remarkable outcomes. We've pioneered a shift from project-focused to product-oriented strategies, significantly enhancing team performance and fostering a culture of outcome-driven innovation. With a dedication to customer-centric methodologies, we've coached numerous product managers and owners to elevate their careers and impact.

Previously, with str84wd Products, the focus was on fostering intrapreneurial talent within corporates, guiding them through the discovery and validation of innovative business models. Our 12-week program was instrumental in instilling a customer-centric approach, which remains a cornerstone of my professional ethos. My commitment to product excellence and innovation continues to influence the growth and success of emerging digital product managers and entrepreneurs.

Thomas Hartmann

🏗️ Vibe Coding vs. Vibe Engineering
Vibe coding is fun, but using easy prompts won't create great results. 

Here is my learning from the last weeks of vibe engineering:

⚡ The Vibe Coding feels great, but...
LLMs make development feel magic. 
Shiny UI. 
First impression: "WOW! I am a hero, I can build anything..." 
But then... the button does not work. Fixes take forever. You need to take out a lot of stuff. etc. 

Anybody can write a prompt and get a shiny object. 
The problem: all undefined areas, the AI just builds something random you need to clean up later. 

👉 Imagine to communicate with your engineering team with a single prompt. 
They are not allowed to ask questions. 
Just build what they think is needed.

It would actually be a funny experiment, isn't it 😂😂😂

✅ Vibe Engineering closes the gap

AI is the accelerator. 
1️⃣ You first use it to generate all required input:
- clear definition what to build
- clarify question on edge cases
- define architecture integration
- define tests
- etc
Now it hast the required context and boundaries.

THEN 2️⃣: You use AI it to execute.

Without direction, without the boundaries, AI accelerates technical debt just as efficiently as it accelerates delivery.

And by the way, I am not a technical guy. Sebastian Borggrewe still needs to explain me GitHub. But the thing is, AI is doing everything for me, and I learn along the way. 

What’s your experience? 

Thanks again Mikhail Shcheglov and Mathias Maisberger for sharing your experiences! Helped a lot :)

#ProductManagment #buiilder #AI #VibeCoding #FutureOfWork #Programming #vibeengineering

Become an AI PM in 8 Weeks at Product Masterclass | Just Product Host | Author “From Project to Product Mode”

Today: PMs write user stories and hand them to developers.
Tomorrow: PMs prompt AI directly.

But only if they can think.

The role of the product manager is shifting fundamentally.

Until now, technical execution was the bottleneck.
Soon, it'll be product thinking.

AI can build anything.
But it needs clear input:
Requirements. Edge cases. Acceptance criteria. Constraints.

The PMs who can define WHAT to build with precision
will have an unfair advantage.

Not because they can code.
Because AI turns their thinking into reality
at unprecedented speed.

The PM superpower of the future isn't code.
It's clarity.

This is exactly what we work on in "Week 3" of the
Product Masterclass: Vibe Coding for Product Managers.
Today I would rather call it Vibe Engineering though. 

How is AI already changing how you work?

I can see me in this video 🤣
Happy weekend!



Share with the fellow PMs you see in this video 😅
@carlthepm on instagram

Apple winning the AI race?!?
Are you kidding me, Thomas? Siri is still the dumbest AI (sorry Siri 😅)
But wait, everybody is buying Mac Minis and showing off on LinkedIn 🤔

Turns out, Apple has a fundamentally different strategy to AI.

While OpenAI, Google, and Microsoft are spending $690 billion on centralized GPU clusters this year, Apple is spending $13 billion on something else entirely:

Chips that run AI on your device. No cloud. No latency. No data leaving your pocket.

The M5 is expected to do approximately 133 TOPS (Trillion Operations Per Second). That's approaching a server-grade Nvidia GPU. In a fanless laptop. On battery. 

That's why everybody is buying the Mac Minis currently to run their local models and show of on LinkedIn 😅

And Apple has 2.5 billion devices with a Neural Engine.

I kept thinking about this and realized: we've seen this exact pattern before.

In the 1960s, IBM dominated with centralized mainframes. Then small, cheap PCs showed up on desks. The mainframe didn't die. It just became irrelevant for 95% of what people needed.

Cloud AI is the mainframe.
Apple Silicon is the personal computer.

I wrote a deep dive on the full analogy, where it holds, where it breaks (Apple's execution is genuinely bad right now), and what it means for product leaders building AI-powered products.

What's your take, is Apple winning or losing the AI race? 
Link in the first comment.

I post about #AI #ProductManagement #ProductMasterclass 
You are welcome to connect & follow if it fits your interests.

For 30 years, engineering was the bottleneck in every tech org.
Everything was built around it.
Product managers. Discovery. Roadmaps. 
All of it tried to optimize the bottleneck: engineering.

AI is making that constraint disappear. 
Cost to produce software is falling dramatically. 

But removing one bottleneck reveals the next bottleneck.

Easy example:
If you replace the weakest link in a chain, another link becomes the weakest. 

I think, product thinking could be the next bottleneck. 

I wrote about this. Link in the comment.
Would love to hear what you think about it!

Haha... I think Claude Code had too many beers yesterday 🤣🤣🤣

Claude Code is the best SW developer for product manager because:
- it's infinite patience to answer your questions 🤖

- ability to explain everything in any level of depth, from kindergarten level to senior scientist 🧑🏫

- it's ability to understand large projects and translate it in simple language and pseudocode 💡

- it's lack of feelings. Yes. Big advantage. I never heard: Oh, that's a lot of work, or, WHAT?!? It took me. xxx and you want to scrap it! 😅

Try it yourself. 
Ask your developer to grant you reading access to the git repository. 

Then start to prompt something like:
- Tell me what happens if I press button XYZ in the frontend. Explain it in pseudocode. 

- Look at my website / service / ... and give me 5 ideas for improvement. Use a UX expert, developer, architect, and conversion expert to analyse the page. 

- Give me a high-level view of the architecture. 

From there on, you will have interesting conversations with your AI developer! 🚀

Awesome for everybody to learn new stuff. Does not matter which level currently you are, junior, intermediate, senior... from the world knowledge everybody can learn. 🤯

Most product managers I talk to are anxious about AI.
There's too much noise.
New tools every week. New use cases. More hype.

Most are still just copy-pasting into ChatGPT.
That's not a workflow. That's survival mode.

So we tore up the old Product Masterclass and built a new one from scratch.

This time, it's AI-native. Built for product managers.

We walk you through a real workflow, not just prompts.
Research. Prioritization. Strategy. All covered.

You get hands-on guidance at every step.
You see exactly where AI fits in, and where it doesn't.

Check out https://lnkd.in/dxQGu5up for more information

Open article: Product Masterclass - ✅ Master Product Management in the AI Era by product-masterclass.com

Product Masterclass - ✅ Master Product Management in the AI Era

Phu... lay off 40%... stock up 25% 😳 
Right or wrong?
Will others follow?

I made a bet with Sebastian Borggrewe that we will have >10% unemployment in 2030 in Germany. 
But I did not expect such bold moves that fast. 

What do you think? Scary? Opportunity?

taplio

The Battle

Christian Hubmann

Thomas Hartmann

Followers

Posts

Engagements

Likes

Performance ScoreChristian Hubmann's LinkedIn Performance

Performance ScoreThomas Hartmann's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Christian Hubmann

Thomas Hartmann

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);}Christian Hubmann's LinkedIn Performance

Performance ScoreThomas Hartmann's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreChristian Hubmann's LinkedIn Performance