Philipp Müller

Verified account

@philipp_muller

building a future for physical retailers with computer vision and ai. i like guitars, sailboats and computers. co-founder@signatrix. on the side

Berlin, Germany

Joined June 2009

1.5K Following

166 Followers

226 Posts

Philipp Müller

@philipp_muller

4 months ago

The convergence of AI with web, mobile, and enterprise tools is creating incredible opportunities right now. Feels like discovering a massive workshop full of parts that actually fit together. Shift from passive user to active builder and you'll be plenty busy for years to come.

0

0

0

0

14

Philipp Müller

@philipp_muller

4 months ago

Agentic AI is “coming soon” to an e-comm store near you. Shopify ramped up agent-driven shopping (personalized discovery + purchase via Sidekick protocols), Visa started testing AI agents that can initiate real payments (with proper safeguards), and NVIDIA open-sourced its Agent Toolkit for enterprise workflows. We’re seeing agents move into fulfillment orchestration, dynamic merchant interactions, and automated procurement where margins and speed matter most. If you’re in retail, e-comm, or CPG and piloting agentic systems for operations or customer journeys, what’s one workflow you’d hand off to an agent first? #RetailTech #AgenticAI #SupplyChain #AIinRetail #Commerce

1

0

0

0

47

Philipp Müller

@philipp_muller

4 months ago

NVIDIA GTC 2026 has been a blast so far. The scale and velocity of this buildout and usage explosion in AI is astonishing, it’s great to see up close. I really enjoyed to give a talk on our Signatrix story scaling AI within retail yesterday. Thanks to NVIDIA Inception and Chris Brown for the invite! Some things I’ve found noteworthy: - @steipete and his Claws is universally adored. Jensen seems to be a big fan and had NVIDIA cooperate with Peter to develop their own NemoClaw which is an adopted OC version with a focus on enterprise and security. - With Mistral AI and Black Forest Labs, there are two European AI labs as founding members of NVIDIAs Nemotron Coalition together with Cursor, LangChain, Perplexity, Reflection AI, Sarvam and Thinking Machines Lab. Notably, none of the large frontier labs like OpenAI, Anthropic or Gemini have opted to join this party. This could be a brilliant move by NVIDIA, basically forming an „anti-large-frontier-lab“ coalition. Model builders in the coalition could save 10s of millions in compute that they’d otherwise have to raise for, application layer companies could use these open models to presumably achieve much better per token economics and expand their margins compared to paying Anthropic or OpenAI markup. There is a risk I guess for the model companies in this coalition that its models compete with their own for margin. Some people at Mistral told me their business model isn’t really in token monetization but enterprise projects rather so this may not be a deal breaker. - Compute demand is so large that basically every datacenter built by neoclouds that gets finished is immediately bought for years by the hyperscalers before this capacity organically makes it to their on-demand consumer pool. This is probably good for quick capex turnaround but also caps their margin potential. - NVIDIA announced their next Vera Rubin generation to power the next wave of AI training and inference. The numbers are insane: 40,000,000x increase in compute per envelope of power/budget within one decade. They also doubled their revenue outlook for 2025-2027 to $1 trillion. Also it looks like alien tech out of a sci-fi movie. - Largest chunk of GTC attendees this year was from financial services, make of that what you will. - Jensen talked for over 2,5h in his keynote. Data centers in space was 1 minute. The challenge is heat dissipation in space: „No conduction, no convection, only radiation.“ „We have many engineers working on that“. - „Every company will be token consumers for their engineers and token producers for their customers.“ - The mantra is simple: tokens=cash and more is better.

philipp_muller's tweet photo. NVIDIA GTC 2026 has been a blast so far. The scale and velocity of this buildout and usage explosion in AI is astonishing, it’s great to see up close.

I really enjoyed to give a talk on our Signatrix story scaling AI within retail yesterday. Thanks to NVIDIA Inception and Chris Brown for the invite!

Some things I’ve found noteworthy:

- @steipete and his Claws is universally adored. Jensen seems to be a big fan and had NVIDIA cooperate with Peter to develop their own NemoClaw which is an adopted OC version with a focus on enterprise and security.

- With Mistral AI and Black Forest Labs, there are two European AI labs as founding members of NVIDIAs Nemotron Coalition together with Cursor, LangChain, Perplexity, Reflection AI, Sarvam and Thinking Machines Lab. Notably, none of the large frontier labs like OpenAI, Anthropic or Gemini have opted to join this party.

This could be a brilliant move by NVIDIA, basically forming an „anti-large-frontier-lab“ coalition. Model builders in the coalition could save 10s of millions in compute that they’d otherwise have to raise for, application layer companies could use these open models to presumably achieve much better per token economics and expand their margins compared to paying Anthropic or OpenAI markup. There is a risk I guess for the model companies in this coalition that its models compete with their own for margin. Some people at Mistral told me their business model isn’t really in token monetization but enterprise projects rather so this may not be a deal breaker.

- Compute demand is so large that basically every datacenter built by neoclouds that gets finished is immediately bought for years by the hyperscalers before this capacity organically makes it to their on-demand consumer pool. This is probably good for quick capex turnaround but also caps their margin potential.

- NVIDIA announced their next Vera Rubin generation to power the next wave of AI training and inference. The numbers are insane: 40,000,000x increase in compute per envelope of power/budget within one decade. They also doubled their revenue outlook for 2025-2027 to $1 trillion. Also it looks like alien tech out of a sci-fi movie.

- Largest chunk of GTC attendees this year was from financial services, make of that what you will.

- Jensen talked for over 2,5h in his keynote. Data centers in space was 1 minute. The challenge is heat dissipation in space: „No conduction, no convection, only radiation.“ „We have many engineers working on that“.

- „Every company will be token consumers for their engineers and token producers for their customers.“

- The mantra is simple: tokens=cash and more is better.

philipp_muller's tweet photo. NVIDIA GTC 2026 has been a blast so far. The scale and velocity of this buildout and usage explosion in AI is astonishing, it’s great to see up close.

I really enjoyed to give a talk on our Signatrix story scaling AI within retail yesterday. Thanks to NVIDIA Inception and Chris Brown for the invite!

Some things I’ve found noteworthy:

- @steipete and his Claws is universally adored. Jensen seems to be a big fan and had NVIDIA cooperate with Peter to develop their own NemoClaw which is an adopted OC version with a focus on enterprise and security.

- With Mistral AI and Black Forest Labs, there are two European AI labs as founding members of NVIDIAs Nemotron Coalition together with Cursor, LangChain, Perplexity, Reflection AI, Sarvam and Thinking Machines Lab. Notably, none of the large frontier labs like OpenAI, Anthropic or Gemini have opted to join this party.

This could be a brilliant move by NVIDIA, basically forming an „anti-large-frontier-lab“ coalition. Model builders in the coalition could save 10s of millions in compute that they’d otherwise have to raise for, application layer companies could use these open models to presumably achieve much better per token economics and expand their margins compared to paying Anthropic or OpenAI markup. There is a risk I guess for the model companies in this coalition that its models compete with their own for margin. Some people at Mistral told me their business model isn’t really in token monetization but enterprise projects rather so this may not be a deal breaker.

- Compute demand is so large that basically every datacenter built by neoclouds that gets finished is immediately bought for years by the hyperscalers before this capacity organically makes it to their on-demand consumer pool. This is probably good for quick capex turnaround but also caps their margin potential.

- NVIDIA announced their next Vera Rubin generation to power the next wave of AI training and inference. The numbers are insane: 40,000,000x increase in compute per envelope of power/budget within one decade. They also doubled their revenue outlook for 2025-2027 to $1 trillion. Also it looks like alien tech out of a sci-fi movie.

- Largest chunk of GTC attendees this year was from financial services, make of that what you will.

- Jensen talked for over 2,5h in his keynote. Data centers in space was 1 minute. The challenge is heat dissipation in space: „No conduction, no convection, only radiation.“ „We have many engineers working on that“.

- „Every company will be token consumers for their engineers and token producers for their customers.“

- The mantra is simple: tokens=cash and more is better.

philipp_muller's tweet photo. NVIDIA GTC 2026 has been a blast so far. The scale and velocity of this buildout and usage explosion in AI is astonishing, it’s great to see up close.

I really enjoyed to give a talk on our Signatrix story scaling AI within retail yesterday. Thanks to NVIDIA Inception and Chris Brown for the invite!

Some things I’ve found noteworthy:

- @steipete and his Claws is universally adored. Jensen seems to be a big fan and had NVIDIA cooperate with Peter to develop their own NemoClaw which is an adopted OC version with a focus on enterprise and security.

- With Mistral AI and Black Forest Labs, there are two European AI labs as founding members of NVIDIAs Nemotron Coalition together with Cursor, LangChain, Perplexity, Reflection AI, Sarvam and Thinking Machines Lab. Notably, none of the large frontier labs like OpenAI, Anthropic or Gemini have opted to join this party.

This could be a brilliant move by NVIDIA, basically forming an „anti-large-frontier-lab“ coalition. Model builders in the coalition could save 10s of millions in compute that they’d otherwise have to raise for, application layer companies could use these open models to presumably achieve much better per token economics and expand their margins compared to paying Anthropic or OpenAI markup. There is a risk I guess for the model companies in this coalition that its models compete with their own for margin. Some people at Mistral told me their business model isn’t really in token monetization but enterprise projects rather so this may not be a deal breaker.

- Compute demand is so large that basically every datacenter built by neoclouds that gets finished is immediately bought for years by the hyperscalers before this capacity organically makes it to their on-demand consumer pool. This is probably good for quick capex turnaround but also caps their margin potential.

- NVIDIA announced their next Vera Rubin generation to power the next wave of AI training and inference. The numbers are insane: 40,000,000x increase in compute per envelope of power/budget within one decade. They also doubled their revenue outlook for 2025-2027 to $1 trillion. Also it looks like alien tech out of a sci-fi movie.

- Largest chunk of GTC attendees this year was from financial services, make of that what you will.

- Jensen talked for over 2,5h in his keynote. Data centers in space was 1 minute. The challenge is heat dissipation in space: „No conduction, no convection, only radiation.“ „We have many engineers working on that“.

- „Every company will be token consumers for their engineers and token producers for their customers.“

- The mantra is simple: tokens=cash and more is better.

philipp_muller's tweet photo. NVIDIA GTC 2026 has been a blast so far. The scale and velocity of this buildout and usage explosion in AI is astonishing, it’s great to see up close.

I really enjoyed to give a talk on our Signatrix story scaling AI within retail yesterday. Thanks to NVIDIA Inception and Chris Brown for the invite!

Some things I’ve found noteworthy:

- @steipete and his Claws is universally adored. Jensen seems to be a big fan and had NVIDIA cooperate with Peter to develop their own NemoClaw which is an adopted OC version with a focus on enterprise and security.

- With Mistral AI and Black Forest Labs, there are two European AI labs as founding members of NVIDIAs Nemotron Coalition together with Cursor, LangChain, Perplexity, Reflection AI, Sarvam and Thinking Machines Lab. Notably, none of the large frontier labs like OpenAI, Anthropic or Gemini have opted to join this party.

This could be a brilliant move by NVIDIA, basically forming an „anti-large-frontier-lab“ coalition. Model builders in the coalition could save 10s of millions in compute that they’d otherwise have to raise for, application layer companies could use these open models to presumably achieve much better per token economics and expand their margins compared to paying Anthropic or OpenAI markup. There is a risk I guess for the model companies in this coalition that its models compete with their own for margin. Some people at Mistral told me their business model isn’t really in token monetization but enterprise projects rather so this may not be a deal breaker.

- Compute demand is so large that basically every datacenter built by neoclouds that gets finished is immediately bought for years by the hyperscalers before this capacity organically makes it to their on-demand consumer pool. This is probably good for quick capex turnaround but also caps their margin potential.

- NVIDIA announced their next Vera Rubin generation to power the next wave of AI training and inference. The numbers are insane: 40,000,000x increase in compute per envelope of power/budget within one decade. They also doubled their revenue outlook for 2025-2027 to $1 trillion. Also it looks like alien tech out of a sci-fi movie.

- Largest chunk of GTC attendees this year was from financial services, make of that what you will.

- Jensen talked for over 2,5h in his keynote. Data centers in space was 1 minute. The challenge is heat dissipation in space: „No conduction, no convection, only radiation.“ „We have many engineers working on that“.

- „Every company will be token consumers for their engineers and token producers for their customers.“

- The mantra is simple: tokens=cash and more is better.

0

0

0

0

60

Philipp Müller

@philipp_muller

4 months ago

0

0

0

0

41

Who to follow

Verified account

building @cardless & trustee @haverfordschool | fitz.eth

daniel stadelmann

Verified account

building @athletedataHQ 🎽 ex growth @traderepublic 📈 sub-3 marathoner & ironman 70.3 world champs qualifier 🏃‍♂️ powered by pizza & pasta 🍝

Verified account

Part Investor (AI Buildout), Part Entrepreneur on hunt for next venture. Founded @FreshWorksApps - sold to EY '22. Interested in 🇺🇸 economy and 🇨🇦 politics

Philipp Müller

@philipp_muller

5 months ago

https://t.co/0xqFawyrJI

0

1

0

0

16

Philipp Müller

@philipp_muller

5 months ago

@heynavtoor Cool! How does it compare to the browser use abilities and efficiency of openclaw?

0

0

0

0

51

Philipp Müller

@philipp_muller

5 months ago

@christophersaum Would be happy to hear your thoughts: https://t.co/4iE9Lq0YH2

0

0

0

0

23

Philipp Müller

@philipp_muller

5 months ago

@thsottiaux Congratulations @steipete go build more awesome things

0

0

0

0

441

Philipp Müller

@philipp_muller

5 months ago

Good read, interesting move by Kimi

5 months ago

https://t.co/29rGYtxrIG

120

2K

157

3K

351K

1

1

0

0

39

philipp_muller retweeted

5 months ago

https://t.co/29rGYtxrIG

120

2K

157

3K

351K

Philipp Müller

@philipp_muller

5 months ago

@openclaw is going to be a massive AI success story in Europe.

0

0

0

0

14

Philipp Müller

@philipp_muller

5 months ago

Most OpenClaw hosting = shared infra + exposed gateways. Security nightmare. @getblitzclaw : isolated VM per user. Gateway never touches the internet. Auth proxy handles everything. Simple rule: if your AI has access to your stuff, it shouldn't be on shared hosting.

2

1

0

0

22

Philipp Müller

@philipp_muller

5 months ago

@anshnanda We give each user their own isolated VM. Gateway runs locally (not exposed), auth through our API proxy. No shared hosting = no cross-user discovery risks. @getblitzclaw

0

0

0

0

89

Philipp Müller

@philipp_muller

5 months ago

@ycombinator Won't be long until YCombinator will only accept (and VC will only fund) AI-native companies. Start becoming one today at https://t.co/iAXJqcFiTz ;)

0

0

0

0

6

Philipp Müller

@philipp_muller

5 months ago

The convergence of AI with web, mobile, and enterprise tools is creating incredible opportunities right now. Feels like discovering a massive workshop full of parts that actually fit together. Shift from passive user to active builder and you'll be plenty busy for years to come.

0

0

0

0

9

Philipp Müller

@philipp_muller

5 months ago

Built on @OpenClaw — open source AI agent framework. BlitzClaw handles the infrastructure so you don't have to. Would love feedback: https://t.co/zmObcdPubo / [email protected]

0

0

0

0

19

Philipp Müller

@philipp_muller

5 months ago

📷 Just launched BlitzClaw — managed OpenClaw hosting I wanted my own always-on AI assistant without the DevOps overhead. So I built it. Now you can have one too. https://t.co/zmObcdPubo

1

0

0

0

31

Philipp Müller

@philipp_muller

5 months ago

What can it do? • Research anything (web search, deep dives) • Browse the web (screenshots, form filling) • Write and run code • Manage files, set reminders • Whatever you teach it

1

0

0

0

7

Last Seen Users on Sotwe

Trends for you

Most Popular Users