MailMaestro

Kraken Digital Asset Exchange

Senior Software Engineer

Alpaca Technology, Inc. (acquired by Boom)

Alpaca is a startup that helps both renters and landlords throughout the tedious journey of renting apartments.

We own the largest global network of housing rental groups on Facebook. We manage our groups through proprietary algorithms, and smartly match renters with landlords through an AI assistant. Think of Alpaca as your personal AI agent for apartment rentals: you tell it a few things and it goes out into the world to look for what you want, so that you don’t sift through all that information. You make the decisions but save time not having to do any of the boring stuff.

Co-Founder and CIO

Freie Universität Berlin

- Head of technical development for INTEGRATION (subproject of the nationwide, government funded research initiative MiKlip - Decadal climate prediction)
- Conception and technical realization of a central evaluation system for decadal climate predictions ("FREVA"), that meanwhile has been implemented internationally at various research institutes
- Programming of the django based website to access the evaluation system running on a high performance computer through the web (https://www-miklip.dkrz.de)

Researcher and Tech Lead

Lightspeed Venture Partners

Lightspeed's annual Summer Fellowship gives entrepreneurs the resources, guidance, and connections they need to grow their ventures. This year, 8 teams were selected to participate out of nearly 300 applications.

Summer Fellow

WGcast GmbH

Co-Founder and CTO

Weather forecasting is starting to feel… solved

After 7 years as a climate scientist, I’ve spent enough time inside physics-based modelling to know when something fundamentally changes. 

This is one of those moments.

Google Deepmind just released a new AI weather model: 
𝗪𝗲𝗮𝘁𝗵𝗲𝗿𝗡𝗲𝘅𝘁 𝟮

𝗧𝗵𝗲 𝘀𝗵𝗼𝗿𝘁 𝘃𝗲𝗿𝘀𝗶𝗼𝗻:
 • Global forecasts in under a minute on a single TPU
 • Hundreds of scenarios from one initial state
 • Higher resolution and better skill on 99.9% of variables and lead times
 • Stronger performance on extremes and cyclone tracks (validated in the paper)

𝗪𝗵𝗮𝘁'𝘀 𝗺𝗼𝘀𝘁 𝗶𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝘁𝗼 𝗺𝗲:
The model is only trained on marginals but it still learns the joint structure of the atmosphere. That’s something we thought required full physics. 

It clearly doesn't anymore.

AI is competing with the methods we built entire institutions around.

For teams in energy, insurance, agriculture, logistics: this enables use cases that were impossible even 2–3 years ago.

Where do you see the biggest gap today between the forecasts we get… and the decisions we need to make?

Sebastian Illing

Best AI for Email and Meetings on MS365 | CTO @ MaestroLabs | Serial Entrepreneur 💡 | Building with AI 🤖 for over 6 years

Sam Altman sent a memo to OpenAI staff. The message was clear: "𝗥𝗼𝘂𝗴𝗵 𝘃𝗶𝗯𝗲𝘀" are coming. Why? 

Because Google just dropped 𝗚𝗲𝗺𝗶𝗻𝗶 𝟯 and 𝗡𝗮𝗻𝗼 𝗕𝗮𝗻𝗮𝗻𝗮 𝗣𝗿𝗼. And for the first time in 3 years, OpenAI doesn't have an answer.

They are scrambling with a new code-name project ("Shallotpeat") just to catch up.

If you built your entire roadmap assuming OpenAI would always be #1, you are in trouble today.

The gap between "Leader" and "Follower" in AI is not measured in years anymore. It is measured in weeks or days.

 • 𝗠𝗼𝗻𝗱𝗮𝘆: GPT-5.1 is the standard.

 • 𝗧𝘂𝗲𝘀𝗱𝗮𝘆: Google releases a model with better pre-training.

 • 𝗪𝗲𝗱𝗻𝗲𝘀𝗱𝗮𝘆: Your "OpenAI-only" strategy looks obsolete.

The industry is ping-ponging. One month it’s GPT. The next it’s Claude. Now it’s Gemini/Banana.

𝗧𝗵𝗲 𝗟𝗲𝘀𝘀𝗼𝗻: Stop marrying the model. The model is a commodity. It is interchangeable.

If your software stack breaks because you switched from GPT-5 to Gemini 3, you built it wrong.

At MaestroLabs, we don't care who is winning the "Model Wars." 
We care about the Workflow.

We built TeamsMaestro and MailMaestro to be model-agnostic. Whether the intelligence comes from a "Shallot" or a "Banana," the work still gets done.

I've been playing with Gemini 3 and what can I say?
It’s over. AGI is basically here. 😅

Looks like we have to shift all our workload to Gemini 3 now.

Actually, the model reasoning is quite convincing (see 🧵 below). 

Do you still see any reason not to use MailMaestro?

Werner Vogels (Amazon CTO) just dropped a reality check at AWS re:Invent. If you are worried about AI stealing your job, you need to read this.

His argument is simple: The era of the "I just write code" engineer is over. AI can write code faster, cheaper, and (soon) better than you.

But AI cannot be an Architect.

Vogels calls the survivor of this shift the "𝗥𝗲𝗻𝗮𝗶𝘀𝘀𝗮𝗻𝗰𝗲 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿." 
It's no longer about how many languages you know. The focus shifts to these five qualities.

𝟭. 𝗔𝗴𝗴𝗿𝗲𝘀𝘀𝗶𝘃𝗲 𝗖𝘂𝗿𝗶𝗼𝘀𝗶𝘁𝘆 
The framework you love today will be obsolete in 2026. If you aren't willing to unlearn your favorite tools every 18 months, you are already legacy.

𝟮. 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 𝗧𝗵𝗶𝗻𝗸𝗶𝗻𝗴 
Juniors think about functions. Seniors think about systems. Stop optimizing the line of code. Start optimizing the leverage point.

𝟯. 𝗖𝗼𝗺𝗺𝘂𝗻𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗮𝘀 𝗮 𝗣𝗿𝗼𝘁𝗼𝗰𝗼𝗹 
The best code is useless if you can't explain why it matters to a non-technical stakeholder. Clear writing is the most underrated programming language.

𝟰. 𝗘𝘅𝘁𝗿𝗲𝗺𝗲 𝗢𝘄𝗻𝗲𝗿𝘀𝗵𝗶𝗽 
"You build it, you run it." If you aren't ready to wake up at 3 AM when your system breaks, you aren't an owner. You're a tourist.

𝟱. 𝗧𝗵𝗲 𝗣𝗼𝗹𝘆𝗺𝗮𝘁𝗵 (Broaden the 'T') 
Deep specialization is risky. The engineers who will thrive are the ones who combine tech with psychology, economics, or design. Generalists win in a shifting landscape.

𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: 
AI won't fire you. It will just promote you from 'writer' to 'editor'."

The developers who hide behind tickets will disappear. The developers who step up to own the system will become invaluable.

Which of these 5 traits do you believe is the most neglected by current engineering leadership? Let me know in the comments.

I just wasted 30 minutes of my team's time.

We were deep in a technical huddle. Debating data models. Users, organizations, complex relationships. The whiteboard in our heads was full.

Then, at the end, someone asked the dreaded question: "So… who's going to write up the requirements?"

Silence.

I realized we had just filled a whiteboard with answers, and then erased it before capturing a single word.

I forgot to invite our TeamsMaestro notetaker.

It sounds ridiculous but I was so focused on the problem, I forgot the tool.

The lesson? The highest ROI for AI right now isn't generating code or writing marketing copy. It’s being the synthesizer.

It turns messy, nonlinear human conversation into structured, actionable documentation.

My new rule: 
Human brain = Thinking 
AI brain = Typing

Don't be like me. Invite TeamsMaestro!

Do you have an "AI notetaker" in every meeting yet, or are you still taking manual notes?

"I can't afford a ChatGPT license"
Here is a life-hack for you:

 1. Hop on a quick call with someone

 2. Ask them to bring their 𝗧𝗲𝗮𝗺𝘀𝗠𝗮𝗲𝘀𝘁𝗿𝗼 AI notetaker

 3. Have the call, then when they leave, just stay in the call

 4. Ask the notetaker questions

 5. Unlimited AI, zero subscription

𝗣𝗿𝗼 𝗧𝗶𝗽:
We are running a Black Friday campaign right now. 
Use 𝗕𝗟𝗔𝗖𝗞𝗙𝗥𝟰𝟬 to save 40% on all Maestro Labs products.

Off to Las Vegas for AWS re:Invent 2025! 

Excited to dive into the latest in Agentic AI, cloud infrastructure, and all things innovation this week.

Who else is on their way or already there? 

Let me know if you'd like to connect for a quick chat, coffee, or a drink to talk cloud strategy!

Turns out LLMs have strong opinions about Microsoft products.
I asked 4 of them which one they’d erase from history…
And it's not Microsoft Teams 😜

🧠 𝗚𝗣𝗧-𝟱:

“𝗪𝗶𝗻𝗱𝗼𝘄𝘀 𝗩𝗶𝘀𝘁𝗮. It damaged Microsoft’s reputation with poor performance, compatibility issues, and an overcomplicated security model that alienated users — slowing adoption of later, much better products like Windows 7.”

🤓 𝗖𝗹𝗮𝘂𝗱𝗲 𝗦𝗼𝗻𝗻𝗲𝘁 𝟰.𝟱:

“𝗪𝗶𝗻𝗱𝗼𝘄𝘀 𝗠𝗘. It was plagued by instability and crashes, offered minimal improvements over Windows 98, and hurt Microsoft’s reputation when they needed a strong consumer OS — all while Windows 2000 already existed as a stable alternative.”

📊 𝗚𝗲𝗺𝗶𝗻𝗶 𝟮.𝟱 𝗣𝗿𝗼:

“𝗠𝗶𝗰𝗿𝗼𝘀𝗼𝗳𝘁 𝗕𝗼𝗯 — the 1995 graphical interface that replaced the desktop with a cartoon house. It was resource-heavy, condescending, and more confusing than the thing it tried to simplify.”

📎 𝗞𝗶𝗺𝗶 𝟮:

“𝗖𝗹𝗶𝗽𝗽𝘆. Releasing a smug, interruptive paperclip as the face of ‘help’ set user trust in Microsoft back a decade — and became the enduring punchline for every bad UX decision.”

Interestingly, every model picked a different product but somehow… every answer feels right. 😅

I'm also surprised that Internet Explorer is not in the list. That would have been my first choice!

What would you erase from Microsoft history?

Most companies brag about their "AI strategy."
𝗦𝗮𝘁𝘆𝗮 𝗡𝗮𝗱𝗲𝗹𝗹𝗮 just explained why 90% of them are accidentally giving away their crown jewels.

The real question of this AI wave isn’t “Who has the biggest model?”
It's: 𝗪𝗵𝗼 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗸𝗲𝗲𝗽𝘀 𝘁𝗵𝗲 𝘃𝗮𝗹𝘂𝗲 𝘁𝗵𝗲𝘆 𝗰𝗿𝗲𝗮𝘁𝗲?

Satya's definition of a real platform is brutal and brilliant:
𝘈 𝘱𝘭𝘢𝘵𝘧𝘰𝘳𝘮 𝘸𝘪𝘯𝘴 𝘰𝘯𝘭𝘺 𝘸𝘩𝘦𝘯 𝘵𝘩𝘦 𝘦𝘤𝘰𝘯𝘰𝘮𝘪𝘤 𝘷𝘢𝘭𝘶𝘦 𝘤𝘳𝘦𝘢𝘵𝘦𝘥 𝘣𝘺 𝘦𝘷𝘦𝘳𝘺𝘰𝘯𝘦 𝘶𝘴𝘪𝘯𝘨 𝘪𝘵 𝘦𝘹𝘤𝘦𝘦𝘥𝘴 𝘵𝘩𝘦 𝘷𝘢𝘭𝘶𝘦 𝘰𝘧 𝘵𝘩𝘦 𝘤𝘰𝘮𝘱𝘢𝘯𝘺 𝘵𝘩𝘢𝘵 𝘣𝘶𝘪𝘭𝘵 𝘪𝘵.

By that standard, a lot of today’s "AI partnerships" are nothing more than value-extraction funnels.
Press-release theatre dressed up as innovation.

Microsoft + OpenAI is his counterexample:
 • Microsoft helped OpenAI scale
 • OpenAI accelerated Microsoft’s roadmap
 • Nvidia + AMD leveled up because the entire ecosystem expanded
 • And now everyone builds on that infrastructure

That’s positive-sum.

But the real test of AI isn’t another trillion-dollar valuation.
It’s when:
• Drug discovery takes 1 year instead of 12
• Supply chains reconfigure overnight
• Teachers personalize for every student
• Farmers prevent crop failures before they happen

The real risk isn’t missing AI.
It’s letting someone else capture the value you create with it.

The winners of this decade will be the companies that treat AI as capability-building, not capability outsourcing.

I’m curious, where do you see companies unintentionally giving away their value today?

Did you ever wonder how MailMaestro works in the backend?

#building_in_public

You're an exhausted employee trying to survive waves of fake "urgent" emails. Now it's an actual game you can play.

For months, I've had an idea for a stupidly simple browser game: 𝗢𝘂𝘁𝗹𝗼𝗼𝗸 𝗗𝗲𝗳𝗲𝗻𝗱𝗲𝗿. You play as an overworked employee defending your inbox from waves of "Urgent" emails and corporate spam.

I've thrown the same exact prompt at every major LLM release for the last couple months. They all failed. Until today.

Claude Opus 4.5 is the first one that managed to build a playable version in one-shot. It even added (quite annoying) sound effects.

𝗧𝗵𝗲 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲: I have paused the deployment pipeline. 🛑

You want to waste 10 minutes playing "Outlook Defender" instead of doing actual work? Prove it.

If this post gets enough engagement, I'll publish the game so you can play it yourself. 🕹️

taplio

Sebastian Illing's Linkedin Analytics

Sebastian Illing

Performance ScoreSebastian Illing's LinkedIn Performance

Sebastian Illing's Best Posts (last 30 days)

Ready to grow your LinkedIn presence?

taplio

.css-6ytdur{background-image:linear-gradient(284deg, rgba(39, 43, 72, 1) 0%, rgba(33, 47, 183, 1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Sebastian Illing's Linkedin Analytics

Sebastian Illing

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);}Sebastian Illing's LinkedIn Performance

Sebastian Illing's Best Posts .css-1emoo84{font-size:var(--chakra-fontSizes-md);font-weight:400;color:var(--chakra-colors-gray-600);}(last 30 days)

Ready to grow your LinkedIn presence?

Sebastian Illing's Linkedin Analytics

Performance ScoreSebastian Illing's LinkedIn Performance

Sebastian Illing's Best Posts (last 30 days)