Karl Weinmeister
@kweinmeister.bsky.social
Cloud Developer Advocacy @ Google. AI/ML/Data, Blue Devil & Longhorn, wanna-be at home improvement. Opinions are my own.
Nano Banana Pro is here and ready for business.
🌍 Deploy localized global campaigns faster
🎨 Create more accurate, context-rich visual assets
🎯 Maintain stronger creative control and brand fidelity
🌍 Deploy localized global campaigns faster
🎨 Create more accurate, context-rich visual assets
🎯 Maintain stronger creative control and brand fidelity
Nano Banana Pro available for enterprise | Google Cloud Blog
Today, we announced Nano Banana Pro (Gemini 3 Pro Image). Nano Banana Pro excels in visual design, world knowledge, and text generation. It’s available on Vertex AI, Google Workspace, and Gemini…
cloud.google.com
November 20, 2025 at 4:23 PM
Nano Banana Pro is here and ready for business.
🌍 Deploy localized global campaigns faster
🎨 Create more accurate, context-rich visual assets
🎯 Maintain stronger creative control and brand fidelity
🌍 Deploy localized global campaigns faster
🎨 Create more accurate, context-rich visual assets
🎯 Maintain stronger creative control and brand fidelity
Go from "I've heard of Antigravity" to "I'm comfortable using it" with the getting started guide. It gets right to the point with each component and use case.
codelabs.developers.google.com/getting-star...
codelabs.developers.google.com/getting-star...
Getting Started with Google Antigravity | Google Codelabs
This codelab guides you through the process of installing and experiencing the features of Google Antigravity, an Agent-first development platform. The codelab covers multiple use cases to get the…
codelabs.developers.google.com
November 20, 2025 at 1:28 PM
Go from "I've heard of Antigravity" to "I'm comfortable using it" with the getting started guide. It gets right to the point with each component and use case.
codelabs.developers.google.com/getting-star...
codelabs.developers.google.com/getting-star...
Extra formatting costs tokens and bloats your context window. TOON is a format designed to increase information density for LLMs.
6 lines of JSON becomes: items[2]{sku,qty,price}: A1,2,9.99 B2,1,14.50
Learn how to include TOON in your development workflow: medium.com/google-cloud...
6 lines of JSON becomes: items[2]{sku,qty,price}: A1,2,9.99 B2,1,14.50
Learn how to include TOON in your development workflow: medium.com/google-cloud...
Save Tokens with TOON using Google Antigravity and the Gemini CLI
When working with Large Language Models, token count is a constant concern. Every token affects performance, adding to cost and latency…
medium.com
November 19, 2025 at 8:26 PM
Extra formatting costs tokens and bloats your context window. TOON is a format designed to increase information density for LLMs.
6 lines of JSON becomes: items[2]{sku,qty,price}: A1,2,9.99 B2,1,14.50
Learn how to include TOON in your development workflow: medium.com/google-cloud...
6 lines of JSON becomes: items[2]{sku,qty,price}: A1,2,9.99 B2,1,14.50
Learn how to include TOON in your development workflow: medium.com/google-cloud...
Google's most advanced model, Gemini 3, is now live!
📈 1487 Elo on WebDev Arena, 76.2% on SWE-bench Verified
🛠️ Try out Google Antigravity: A new agentic IDE with direct access to the terminal, editor, and browser to build and validate code.
blog.google/products/gem...
📈 1487 Elo on WebDev Arena, 76.2% on SWE-bench Verified
🛠️ Try out Google Antigravity: A new agentic IDE with direct access to the terminal, editor, and browser to build and validate code.
blog.google/products/gem...
A new era of intelligence with Gemini 3
Today we’re releasing Gemini 3 – our most intelligent model that helps you bring any idea to life.
blog.google
November 18, 2025 at 4:16 PM
Google's most advanced model, Gemini 3, is now live!
📈 1487 Elo on WebDev Arena, 76.2% on SWE-bench Verified
🛠️ Try out Google Antigravity: A new agentic IDE with direct access to the terminal, editor, and browser to build and validate code.
blog.google/products/gem...
📈 1487 Elo on WebDev Arena, 76.2% on SWE-bench Verified
🛠️ Try out Google Antigravity: A new agentic IDE with direct access to the terminal, editor, and browser to build and validate code.
blog.google/products/gem...
GEPA can improve your AI agent performance simply by evolving the instruction or prompt.
www.youtube.com/shorts/QGLbY...
www.youtube.com/shorts/QGLbY...
GEPA Introduction (Genetic-Pareto Prompt Optimizer)
YouTube video by Cloud with Karl
www.youtube.com
November 14, 2025 at 10:32 PM
GEPA can improve your AI agent performance simply by evolving the instruction or prompt.
www.youtube.com/shorts/QGLbY...
www.youtube.com/shorts/QGLbY...
“We adopted Rust for its security and are seeing a 1000x reduction in memory safety vulnerability density… With Rust changes having a 4x lower rollback rate and spending 25% less time in code review, the safer path is now also the faster one.”
security.googleblog.com/2025/11/rust...
security.googleblog.com/2025/11/rust...
Rust in Android: move fast and fix things
Posted by Jeff Vander Stoep, Android Last year, we wrote about why a memory safety strategy that focuses on vulnerability prevention in ...
security.googleblog.com
November 14, 2025 at 2:41 PM
“We adopted Rust for its security and are seeing a 1000x reduction in memory safety vulnerability density… With Rust changes having a 4x lower rollback rate and spending 25% less time in code review, the safer path is now also the faster one.”
security.googleblog.com/2025/11/rust...
security.googleblog.com/2025/11/rust...
Hugging Face and Google Cloud have announced a new and deeper partnership today!
⚡️ New CDN Gateway will cache Hugging Face models on Google Cloud
💻 Smooth experience across HF and Google surfaces
🔒 Google's security tech to provide stronger protection for HF Hub
huggingface.co/blog/google-...
⚡️ New CDN Gateway will cache Hugging Face models on Google Cloud
💻 Smooth experience across HF and Google surfaces
🔒 Google's security tech to provide stronger protection for HF Hub
huggingface.co/blog/google-...
Building for an Open Future - our new partnership with Google Cloud
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
November 13, 2025 at 3:55 PM
Hugging Face and Google Cloud have announced a new and deeper partnership today!
⚡️ New CDN Gateway will cache Hugging Face models on Google Cloud
💻 Smooth experience across HF and Google surfaces
🔒 Google's security tech to provide stronger protection for HF Hub
huggingface.co/blog/google-...
⚡️ New CDN Gateway will cache Hugging Face models on Google Cloud
💻 Smooth experience across HF and Google surfaces
🔒 Google's security tech to provide stronger protection for HF Hub
huggingface.co/blog/google-...
Agent Sandbox debuted at #Kubecon, unlocking a critical capability for agents: running untrusted actions. Safely validate generated code, test user input, and more.
Blog post: cloud.google.com/blog/product...
Documentation: docs.cloud.google.com/kubernetes-e...
Code: github.com/kubernetes-s...
Blog post: cloud.google.com/blog/product...
Documentation: docs.cloud.google.com/kubernetes-e...
Code: github.com/kubernetes-s...
November 11, 2025 at 9:08 PM
Agent Sandbox debuted at #Kubecon, unlocking a critical capability for agents: running untrusted actions. Safely validate generated code, test user input, and more.
Blog post: cloud.google.com/blog/product...
Documentation: docs.cloud.google.com/kubernetes-e...
Code: github.com/kubernetes-s...
Blog post: cloud.google.com/blog/product...
Documentation: docs.cloud.google.com/kubernetes-e...
Code: github.com/kubernetes-s...
Natural Language to SQL is the quintessential case study for multi-agent systems, with these challenges:
👻 Schema hallucinations
🔀 Order issues
🧩 Logic Errors
🚨 Dangerous execution
🐢 Poor performance
🧹 Messy output
www.youtube.com/shorts/-Vwd_...
👻 Schema hallucinations
🔀 Order issues
🧩 Logic Errors
🚨 Dangerous execution
🐢 Poor performance
🧹 Messy output
www.youtube.com/shorts/-Vwd_...
Natural Language to SQL: 6 Biggest Challenges for Multi-Agent Systems
Many models are great at writing SQL, but they often fail in subtle ways. In this video, we'll break down the six most common failures of NL-to-SQL and show you how to build a robust,…
www.youtube.com
November 11, 2025 at 2:54 PM
Natural Language to SQL is the quintessential case study for multi-agent systems, with these challenges:
👻 Schema hallucinations
🔀 Order issues
🧩 Logic Errors
🚨 Dangerous execution
🐢 Poor performance
🧹 Messy output
www.youtube.com/shorts/-Vwd_...
👻 Schema hallucinations
🔀 Order issues
🧩 Logic Errors
🚨 Dangerous execution
🐢 Poor performance
🧹 Messy output
www.youtube.com/shorts/-Vwd_...
Now you can see two websites in a single tab, with Chrome's new Split View.
Enable with "Split View" flag at chrome://flags.
⚠️ As an experimental feature, it could change or be removed at any time.
#browser #googlechrome #techtips
Enable with "Split View" flag at chrome://flags.
⚠️ As an experimental feature, it could change or be removed at any time.
#browser #googlechrome #techtips
November 2, 2025 at 10:22 AM
Now you can see two websites in a single tab, with Chrome's new Split View.
Enable with "Split View" flag at chrome://flags.
⚠️ As an experimental feature, it could change or be removed at any time.
#browser #googlechrome #techtips
Enable with "Split View" flag at chrome://flags.
⚠️ As an experimental feature, it could change or be removed at any time.
#browser #googlechrome #techtips
Have you seen Python libraries “powered by Rust” and wondered how you could do it, too?
My new article walks you through every step of the way. It shows you how you can build a Rust-based MCP tool served by a Python FastMCP server.
medium.com/google-cloud...
@thisweekinrust.bsky.social
My new article walks you through every step of the way. It shows you how you can build a Rust-based MCP tool served by a Python FastMCP server.
medium.com/google-cloud...
@thisweekinrust.bsky.social
Python and Rust interoperability: A walkthrough for building a high performance MCP server
You’ll learn step-by-step instructions for including Rust code with your Python code. We’ll build a tool for AI agents compliant with MCP.
medium.com
October 20, 2025 at 1:55 PM
Have you seen Python libraries “powered by Rust” and wondered how you could do it, too?
My new article walks you through every step of the way. It shows you how you can build a Rust-based MCP tool served by a Python FastMCP server.
medium.com/google-cloud...
@thisweekinrust.bsky.social
My new article walks you through every step of the way. It shows you how you can build a Rust-based MCP tool served by a Python FastMCP server.
medium.com/google-cloud...
@thisweekinrust.bsky.social
October 17, 2025 at 9:33 PM
vLLM TPU is now powered by tpu-inference!
* Broader model coverage and feature support
* 5x faster performance than Feb 2025 version
Learn more: blog.vllm.ai/2025/10/16/v...
* Broader model coverage and feature support
* 5x faster performance than Feb 2025 version
Learn more: blog.vllm.ai/2025/10/16/v...
October 16, 2025 at 8:41 PM
vLLM TPU is now powered by tpu-inference!
* Broader model coverage and feature support
* 5x faster performance than Feb 2025 version
Learn more: blog.vllm.ai/2025/10/16/v...
* Broader model coverage and feature support
* 5x faster performance than Feb 2025 version
Learn more: blog.vllm.ai/2025/10/16/v...
Rust ⬆️
Python ⬇️
That's my prediction 2 years out. Why?
1. Library support is no longer a moat. Existing libraries will proliferate in more languages or be bypassed with vibe-coding.
2. A language's perceived difficulty will not be the barrier to adoption it once was, if it offers unique value.
Python ⬇️
That's my prediction 2 years out. Why?
1. Library support is no longer a moat. Existing libraries will proliferate in more languages or be bypassed with vibe-coding.
2. A language's perceived difficulty will not be the barrier to adoption it once was, if it offers unique value.
October 14, 2025 at 1:02 PM
Rust ⬆️
Python ⬇️
That's my prediction 2 years out. Why?
1. Library support is no longer a moat. Existing libraries will proliferate in more languages or be bypassed with vibe-coding.
2. A language's perceived difficulty will not be the barrier to adoption it once was, if it offers unique value.
Python ⬇️
That's my prediction 2 years out. Why?
1. Library support is no longer a moat. Existing libraries will proliferate in more languages or be bypassed with vibe-coding.
2. A language's perceived difficulty will not be the barrier to adoption it once was, if it offers unique value.
💥 Stop manually spinning up GPU clusters. Deploy vLLM on GKE more reliably with Terraform.
Learn:
✅ AI inference with IaC
✅ Spot + on-demand GPU node pools
✅ Persistent model caching with PVCs
✅ GitOps CI/CD pipelines via GitHub Actions
Full working code + architecture:
medium.com/google-cloud...
Learn:
✅ AI inference with IaC
✅ Spot + on-demand GPU node pools
✅ Persistent model caching with PVCs
✅ GitOps CI/CD pipelines via GitHub Actions
Full working code + architecture:
medium.com/google-cloud...
Deploy Faster with Terraform: Your Guide to vLLM on GKE with Infrastructure-as-Code
This article is a practical guide on how to use Terraform for agile ML engineering with IaC.
medium.com
October 13, 2025 at 2:02 PM
💥 Stop manually spinning up GPU clusters. Deploy vLLM on GKE more reliably with Terraform.
Learn:
✅ AI inference with IaC
✅ Spot + on-demand GPU node pools
✅ Persistent model caching with PVCs
✅ GitOps CI/CD pipelines via GitHub Actions
Full working code + architecture:
medium.com/google-cloud...
Learn:
✅ AI inference with IaC
✅ Spot + on-demand GPU node pools
✅ Persistent model caching with PVCs
✅ GitOps CI/CD pipelines via GitHub Actions
Full working code + architecture:
medium.com/google-cloud...
Is Terraform worth learning? Is it useful even for small projects? Here’s my take.
youtube.com/shorts/qXsAJ...
#DevOps #MLOps #AI
youtube.com/shorts/qXsAJ...
#DevOps #MLOps #AI
Should you learn Terraform?
Is Terraform worth learning? Is it useful even for small projects? Here’s my take.
#DevOps #MLOps #AI
youtube.com
October 10, 2025 at 7:46 PM
Is Terraform worth learning? Is it useful even for small projects? Here’s my take.
youtube.com/shorts/qXsAJ...
#DevOps #MLOps #AI
youtube.com/shorts/qXsAJ...
#DevOps #MLOps #AI
Are you up to speed yet on agent security? Catch up with Ayo Adedeji and Aron Eidelman on:
- Real-world attack vectors
- Practical implementations
- Multi-agent considerations
youtu.be/nxezufaezHw
- Real-world attack vectors
- Practical implementations
- Multi-agent considerations
youtu.be/nxezufaezHw
The Agent Factory - Episode 10: Agent Security
YouTube video by Google Cloud Tech
youtu.be
October 9, 2025 at 1:22 PM
Are you up to speed yet on agent security? Catch up with Ayo Adedeji and Aron Eidelman on:
- Real-world attack vectors
- Practical implementations
- Multi-agent considerations
youtu.be/nxezufaezHw
- Real-world attack vectors
- Practical implementations
- Multi-agent considerations
youtu.be/nxezufaezHw
What API call are you making 10x a day? 🤔 Turn it into a simple /command.
With Gemini CLI extensions, you can build your own shortcuts to speed up your work:
/fetch_jira ABC-123
/deploy_staging
Learn how easy it is to get started:
geminicli.com/docs/extensi...
With Gemini CLI extensions, you can build your own shortcuts to speed up your work:
/fetch_jira ABC-123
/deploy_staging
Learn how easy it is to get started:
geminicli.com/docs/extensi...
October 8, 2025 at 3:56 PM
What API call are you making 10x a day? 🤔 Turn it into a simple /command.
With Gemini CLI extensions, you can build your own shortcuts to speed up your work:
/fetch_jira ABC-123
/deploy_staging
Learn how easy it is to get started:
geminicli.com/docs/extensi...
With Gemini CLI extensions, you can build your own shortcuts to speed up your work:
/fetch_jira ABC-123
/deploy_staging
Learn how easy it is to get started:
geminicli.com/docs/extensi...
Has Gemini ever felt like it's losing focus as your conversation goes on? Naturally, more context means more topics to cover.
Use the /compress command to keep the Gemini CLI on track. It prunes the history without a full reboot.
Get started today with the Gemini CLI: npx @google/gemini-cli
Use the /compress command to keep the Gemini CLI on track. It prunes the history without a full reboot.
Get started today with the Gemini CLI: npx @google/gemini-cli
October 7, 2025 at 5:03 PM
Has Gemini ever felt like it's losing focus as your conversation goes on? Naturally, more context means more topics to cover.
Use the /compress command to keep the Gemini CLI on track. It prunes the history without a full reboot.
Get started today with the Gemini CLI: npx @google/gemini-cli
Use the /compress command to keep the Gemini CLI on track. It prunes the history without a full reboot.
Get started today with the Gemini CLI: npx @google/gemini-cli
Not only is Nano Banana 🍌 production ready,
it now supports 10 aspect ratios!
Landscape: 21:9, 16:9, 4:3, 3:2
Square: 1:1
Portrait: 9:16, 3:4, 2:3
Flexible: 5:4, 4:5
it now supports 10 aspect ratios!
Landscape: 21:9, 16:9, 4:3, 3:2
Square: 1:1
Portrait: 9:16, 3:4, 2:3
Flexible: 5:4, 4:5
Gemini 2.5 Flash Image now ready for production with new aspect ratios- Google Developers Blog
Our state-of-the-art image generation and editing model which has captured the imagination of the wo...
developers.googleblog.com
October 3, 2025 at 2:30 PM
Not only is Nano Banana 🍌 production ready,
it now supports 10 aspect ratios!
Landscape: 21:9, 16:9, 4:3, 3:2
Square: 1:1
Portrait: 9:16, 3:4, 2:3
Flexible: 5:4, 4:5
it now supports 10 aspect ratios!
Landscape: 21:9, 16:9, 4:3, 3:2
Square: 1:1
Portrait: 9:16, 3:4, 2:3
Flexible: 5:4, 4:5
Who’s your audience, and how are you offering value in your message?
AI tooling is immensely useful as a partner, but your engagement and insights remain essential.
hbr.org/2025/09/ai-g...
AI tooling is immensely useful as a partner, but your engagement and insights remain essential.
hbr.org/2025/09/ai-g...
AI-Generated “Workslop” Is Destroying Productivity
Despite a surge in generative AI use across workplaces, most companies are seeing little measurable ROI. One possible reason is because AI tools are being used to produce “workslop”—content that appea...
hbr.org
October 3, 2025 at 11:55 AM
Who’s your audience, and how are you offering value in your message?
AI tooling is immensely useful as a partner, but your engagement and insights remain essential.
hbr.org/2025/09/ai-g...
AI tooling is immensely useful as a partner, but your engagement and insights remain essential.
hbr.org/2025/09/ai-g...
Reposted by Karl Weinmeister
Dublin 🇮🇪 we're coming!
Learn to build AI agents and deploy your MCP servers to production scale.
Register: goo.gle/accelerate-ai-dublin
Seats are limited!
Shir Meir Lador @kweinmeister.bsky.social @caseywest.bsky.social
#AI #AIagents #MCPServers #CloudRun #Workshop @GoogleCloudTech #DublinEvents
Learn to build AI agents and deploy your MCP servers to production scale.
Register: goo.gle/accelerate-ai-dublin
Seats are limited!
Shir Meir Lador @kweinmeister.bsky.social @caseywest.bsky.social
#AI #AIagents #MCPServers #CloudRun #Workshop @GoogleCloudTech #DublinEvents
September 30, 2025 at 12:51 PM
Dublin 🇮🇪 we're coming!
Learn to build AI agents and deploy your MCP servers to production scale.
Register: goo.gle/accelerate-ai-dublin
Seats are limited!
Shir Meir Lador @kweinmeister.bsky.social @caseywest.bsky.social
#AI #AIagents #MCPServers #CloudRun #Workshop @GoogleCloudTech #DublinEvents
Learn to build AI agents and deploy your MCP servers to production scale.
Register: goo.gle/accelerate-ai-dublin
Seats are limited!
Shir Meir Lador @kweinmeister.bsky.social @caseywest.bsky.social
#AI #AIagents #MCPServers #CloudRun #Workshop @GoogleCloudTech #DublinEvents
Learn about Sparse Attention in DeepSeek-V3.2-Exp:
* O(L²) → O(L·k) with similar performance to V3.1 Terminus
* Lightning indexer scores previous tokens
* Top-k selector picks top 2k sparse tokens from 128k window
📄 Paper: github.com/deepseek-ai/...
🎬Video:
youtube.com/shorts/CLsju...
* O(L²) → O(L·k) with similar performance to V3.1 Terminus
* Lightning indexer scores previous tokens
* Top-k selector picks top 2k sparse tokens from 128k window
📄 Paper: github.com/deepseek-ai/...
🎬Video:
youtube.com/shorts/CLsju...
DeepSeek Sparse Attention Explained
YouTube video by Cloud with Karl
youtube.com
September 30, 2025 at 1:23 AM
Learn about Sparse Attention in DeepSeek-V3.2-Exp:
* O(L²) → O(L·k) with similar performance to V3.1 Terminus
* Lightning indexer scores previous tokens
* Top-k selector picks top 2k sparse tokens from 128k window
📄 Paper: github.com/deepseek-ai/...
🎬Video:
youtube.com/shorts/CLsju...
* O(L²) → O(L·k) with similar performance to V3.1 Terminus
* Lightning indexer scores previous tokens
* Top-k selector picks top 2k sparse tokens from 128k window
📄 Paper: github.com/deepseek-ai/...
🎬Video:
youtube.com/shorts/CLsju...
You have four autonomy dials you can tune in your agentic AI system. Are you using them?
medium.com/google-cloud...
medium.com/google-cloud...
The Agency Spectrum: An AI Risk Management Framework
We can now build autonomous systems that pursue meaningful, high-level goals. Yet, for every inspiring success story, there is a…
medium.com
September 26, 2025 at 4:25 PM
You have four autonomy dials you can tune in your agentic AI system. Are you using them?
medium.com/google-cloud...
medium.com/google-cloud...
If you're a Google AI Pro or Ultra subscriber, your daily limits for Gemini CLI and Code Assist just got a nice bump. Spend less time worrying about quotas, and more on building! 💻
blog.google/technology/d...
blog.google/technology/d...
Google AI Pro and Ultra subscribers now get Gemini CLI and Gemini Code Assist with higher limits.
Google AI Pro and Ultra subscribers now get higher limits to Gemini CLI and Gemini Code Assist IDE extensions.
blog.google
September 24, 2025 at 4:05 PM
If you're a Google AI Pro or Ultra subscriber, your daily limits for Gemini CLI and Code Assist just got a nice bump. Spend less time worrying about quotas, and more on building! 💻
blog.google/technology/d...
blog.google/technology/d...