Addy Osmani is an Irish Software Engineer and leader currently working as a director at Google Cloud AI, focused on helping developers and businesses succeed with Gemini, Vertex AI, and the Agent Development Kit (ADK). After nearly 14 years leading developer experiences in Chrome, he now bridges Google DeepMind, engineering, product, and developer relations teams. A developer for 25+ years at Google, he is passionate about AI-assisted engineering and developer tools. Follow for motivation, productivity and engineering.

Addy Osmani

Introducing our biggest Google Maps upgrade yet - powered by Gemini!

Today Google Maps is getting its biggest upgrade in over a decade!

By combining our Gemini models with a deep understanding of the world, Maps now unlocks entirely new possibilities for how you navigate and explore. 

Here's what you need to know:

🗣️ Ask Maps A new way to get answers to your complex, real-world questions with a simple conversation. Try asking things like:

"My phone is dying - where can I charge it without having to wait in a long line for coffee?" 🔋

"Is there a public tennis court with lights on that I can play at tonight?" 🎾

"I'm headed to the Grand Canyon, Horseshoe Bend, and Coral Dunes - any recommended stops along the way?" 🏜️

No more sifting through research and reviews - just tap "Ask Maps" and get your questions answered with a customized map. (Starts rolling out now in the U.S. and India on Android and iOS, with desktop coming soon).

🚗 Immersive Navigation (that first demo!)

The Maps driving experience is also evolving with clearer visuals and intuitive guidance. You’ll be able to see the buildings, overpasses, and terrain around you in a vivid 3D view, made possible with help from Gemini models.

You’ll also be able to:

1. See more of your route to prepare for what’s next.
2. Understand trade-offs for alternate routes to pick what works best for you.
3. Arrive easily with helpful details like parking and entrance information.

Really excited about these updates! Go Gemini!

#ai #programming #softwareengineering

Director, Google Cloud AI. Best-selling Author. Speaker. AI, DX, UX. I want to see you win.

You can instantly 10x your AI-generated frontends just by learning what different UI components are called.

The Component Gallery (https://component.gallery) is a comprehensive visual reference for standard UI patterns and their proper names ⚡ 

When delegating tasks to coding agents, the quality of your output is bottlenecked by the specificity of your input. If the only vocabulary you use in your prompts are generic words like "menu" or "button," the model will naturally return generic results.

To get high-fidelity interfaces, you need the right shared language for spec-driven development. Knowing when to ask for a "combobox" instead of a "dropdown" or a "segmented control" instead of a "toggle" gives the AI the exact context it needs to execute your vision. The Component Gallery walks through this really well.

UI Design Brain is also an open-source taxonomy for UI skills and component design. (https://lnkd.in/gYYbhv5q) based on the Component Gallery.

Mastering this terminology is a small investment that massively upgrades your prompt engineering.

#ai #programming #softwareengineering

"WebMCP - MCP Is Coming to the Browser"

The Model Context Protocol just got a proposal to bring it natively to the web.
WebMCP is a new JavaScript API that lets web apps expose their functionality as "tools" that AI agents can invoke directly in the browser. Think of web pages as MCP servers - but running entirely client-side, using your existing frontend code. Available as a Google Chrome Early Preview.

Why this matters:

→ Web developers can declare what their app does in a machine-readable way using APIs they already know 
→ Browser agents can discover and call these tools - no backend MCP server required 
→ Two approaches: 1) a declarative API for standard actions via HTML forms  and 2) an imperative API for complex dynamic interactions via JavaScript 
→ Users and agents share the same UI context, enabling cooperative workflows where you delegate the tedious parts and stay in control of the decisions

The use cases click immediately: an agent that can actually search flights, configure options, and handle bookings through structured data. Customer support tickets that get auto-filled with the right technical details. E-commerce flows where an agent navigates checkout with precision instead of fumbling through DOM elements.

Pietro Schirano built a great demo showing this in action - a DoorDash-style app where the agent adds items to cart and checks out with the right address and promo code, without ever needing to "see" the UI. That's the shift really - agents shouldn't have to screenshot or dump HTML to use your app. Just expose your tools.

This is also potentially huge for accessibility - assistive tech that can understand app capabilities semantically, offering users higher-level actions instead of navigating individual UI elements.

We went from "MCP for desktop tools" to "MCP for the entire web" in months. The pace of this protocol's adoption is remarkable.

If you're building web apps, this is worth watching closely.

🔗 WebMCP Proposal: https://lnkd.in/gxbhzPQ9 
🔗 Chrome Early Preview: https://lnkd.in/gySqDEKS
🔗 Syntax.fm deep dive: https://lnkd.in/gbDfWuFn
🔗 Pietro's WebMCP starter: https://lnkd.in/gMWCksvZ

#ai #programming #softwareengineering

An engineer at Anthropic wrote a spec, pointed Claude at an Asana board, and went home. Claude did the rest.

Claude broke the spec into tickets, spawned agents for each one, and they started building independently. When an agent got confused, it ran git-blame and messaged the right engineer in Slack.

By Monday, the agents had finished the feature.

This is the part that fascinates me most: the agent's learning loop mirrors how a new team members onboards. It reads the codebase, identifies who knows what, and asks the right humans. Slack and email - that's all it needed.

We're approaching a world where developers orchestrate dozens of AI agents in parallel. And when that happens, the differentiator between a good engineer and a great one comes down to one thing:

The quality of your specs.

You can't write a spec that holds up at that scale without genuinely understanding what you're building at a deeper level. Architecture, edge cases, failure modes, integration points. The spec IS the product thinking.

Having agents run over the weekend to ship work will start to become entirely normal. There's zero reason you wouldn't add two extra days of output with a couple of prompts and a machine running in the background.

That said - I'm always a little skeptical of how viable these patterns broadly are but do think this is the direction we're heading in. Verification remains the biggest unsolved challenge. Tests that pass on the pre-patch codebase, UI verification, pathological coding practices leaking into the context window. We need slightly better models and much better evaluation infrastructure for this to become the default. I believe we will get there.

But the trajectory is clear: the next-gen developer who understands fundamentals, can architect well, and knows how to orchestrate agents is going to be a force-multiplier.

#ai #programming #softwareengineering

Anthropic has launched free courses to master AI with certificates for $0! 

You can find the full catalog here: https://lnkd.in/gxXNxesb

Yes, they're worth it - especially if you're a web developer. These free Anthropic Academy courses are short (mostly 1-2 hours each), entirely self-paced with videos and quizzes, and provide official certificates of completion when you're done. There's even Claude on Vertex guidance on there.

For developers, the technical tracks are highly practical. They teach you how to move past basic prompting and start building real AI integrations into your daily workflow:

🔹 Claude Code in Action: Learn how to integrate Claude Code directly into your development workflow, master CLI tools, and handle complex context management.
🔹 Building with the Claude API: A comprehensive, hands-on look at working programmatically with Anthropic models.
🔹 Introduction to Model Context Protocol (MCP): Master MCP’s core primitives (tools, resources, and prompts) to connect Claude with external services using Python.
🔹 Claude 101 & AI Fluency: Strong foundations for collaborating with AI systems efficiently and safely.

Understanding context management, agent skills, and API workflows is quickly becoming a critical engineering baseline. These courses give you a phenomenal foundation for AI-assisted development. Plus, the certificates come directly from a top-tier lab, offering a solid signal of real, applied skills to add to your LinkedIn profile or resume.

Enroll and knock out a couple today. 

For those building in the Google ecosystem, make sure to check out the "Claude with Google Cloud's Vertex AI" It's an excellent, comprehensive guide to working with Anthropic models at enterprise scale!

#ai #programming #softwareengineering

Chrome just became massively more agent-friendly 🔥 Your real, signed-in browser can now be natively accessible to any coding agent (with a toggle)

One toggle in chrome://inspect/#remote-debugging and your real, signed-in browser is now natively accessible to any coding agent. Just use this with Chrome DevTools MCP auto connect. OpenClaw just added support too.

No extensions.
No headless browser.
No screenshots.
No separate logins.

Check this out: https://lnkd.in/gNcDHFnV

Your agent finally sees your web the exact tabs, cookies, dashboards, Linear boards, Notion pages, and internal tools you already have open.

This is the moment the biggest friction in browser agents disappeared. 

Google shipped native DevTools MCP live-session attach (Chrome 146+), and tools like OpenClaw and Browser Use instantly plugged in.

https://lnkd.in/gwnKE6fd

Tested it instantly: told my OpenClaw to check my live Google Analytics traffic dashboard. It opened the tab I already had logged in, read the real numbers, and dropped them straight into a weekly report. Zero extra steps. Note: I give my agents their own accounts (e.g. Google account, specific permissions on their accounts for other sites) as this gives me better security isolation.

The same toggle works for Claude Code, gemini-cli, and any MCP-native agent. 

This closes the "my agent can browse the web… but not my web" gap.

#ai #programming #softwareengineering

Introducing the Google Workspace CLI: https://lnkd.in/gpRfeFvR - built for humans and agents.

Google Drive, Gmail, Calendar, Sheets, Chat and every Workspace API. 

The repo ships many Agent Skills - one for every supported API, plus higher-level helpers for common workflows and 50 curated recipes for Gmail, Drive, Docs, Calendar, and Sheets. 

Great work here by Justin Poehnelt 🙏  

#ai #programming #softwareengineering

Boris created Claude Code. His point here is important -  when AI handles the code generation, the engineer's value shifts to the decisions above the code: 

1. what do we build?
2. why? for whom? 
3. and how it all fits together. 

The bottleneck was always judgment, taste, and systems thinking. AI just made that more obvious. I continue to believe that as engineering and product roles evolve (and even converge) through this time, taste is so important.

For me, the line is roughly:

Agents handle
• code co-authoring
• low-risk code reviews
• tactical planning inside a clearly defined feature

I stay very involved in
• technical / UX design
• feature briefs, high-risk reviews
• app-level intent and constraints

The leverage feels highest when humans stay crisp on "what problem, what constraints, what good looks like" and let agents explore the solution space aggressively underneath that.

#ai #programming #softwareengineering

"I was a 10x engineer. Now I'm useless." here's my take:

AI will encourage software engineers to be more deliberate about what thinking we protect, what code we read and what bottlenecks only we can still unblock.

This may be work with high-dimensional search spaces, where the bottleneck isn't code at all but judgment, taste, domain knowledge, or where hard questions that have to be answered before the code even matters. 

On code, I don't think the choice is between reading all the code and reading none of it. I still read lots of code - just not all of it. The question is developing judgment about which parts actually need your eyes on them. Even if I have multiple agents working together to write code, I always account for some review time and some quality inspection.

A lot of developers have spent years building a professional identity around their ability to figure things out. Mo's video is about what happens when you quietly let something else do the figuring out, and then look up one day and realize the skill isn't there anymore.

Video credit is Mo Bitar (https://lnkd.in/gPdByU9x) who I've enjoyed listening to lately and is worth a follow.

#ai #programming #softwareengineering

Tip: Stop using /init for AGENTS.md to get better performance

My latest deep-dive: https://lnkd.in/gkmZ3HJs ✍ 

As we transition deeper into AI-assisted engineering and background agent orchestration, there's a common ritual: setting up a new project, running an auto-generation tool, and committing a comprehensive AGENTS.md context file. It feels like the responsible thing to do.

But recent 2026 research from ETH Zurich and others reveals a different reality: auto-generated context files can actually reduce task success by 2-3% while inflating costs by over 20%.

Why? Because coding agents can already discover your directory structure, tech stack, and module explanations on their own. Handing them an auto-generated summary just adds noise, burns tokens, and dilutes their attention from the actual task.

So, what actually earns a line in your AGENTS.md?

1. The undiscoverable: Tooling gotchas that change operations
2. Operational landmines: "The legacy/ directory is deprecated but imported by production - do not delete."
3. Non-obvious conventions: Custom middleware patterns that shouldn't be refactored to standard conventions.

If an agent can discover it by reading your code, delete it from the file.

A better mental model: Treat your AGENTS.md as a living list of codebase smells you haven't fixed yet. If an agent keeps reaching for the wrong dependency or putting utilities in the wrong folder, don't just add a prose instruction to the context file. Restructure the code, add a linter rule, or improve the test coverage. Treat it as a diagnostic tool, fix the underlying friction, and then delete the line.

I dive into the data, the "pink elephant" problem of context anchoring, and why monolithic context files need to evolve into dynamic routing layers in my latest post. 

How is your team currently managing context for your background coding agents? Are you relying on static files, or exploring more dynamic context loading? Let me know in the comments.

#ai #programming #softwareengineering

Announcing Gemini 3.1 Pro! Google's smarter model for your most complex tasks just shipped!

Last week, we released a major update to Gemini 3 Deep Think to solve modern challenges across science, research, and engineering. Today, we are releasing the upgraded core intelligence that makes those breakthroughs possible: Gemini 3.1 Pro.

Gemini 3.1 Pro is now available for developers, users and enterprise. I'm sharing some fresh demos showing off just a little of what it can do :)

Building on the Gemini 3 series, 3.1 Pro represents a massive step forward in core reasoning. It is designed for tasks where a simple answer is not enough. The model delivers a 77.1% score on the ARC-AGI-2 benchmark, which effectively doubles the reasoning performance of Gemini 3 Pro.

This is a noticeably smarter, more capable baseline for complex problem-solving. Gemini 3.1 Pro excels at workflows where the model must decide for itself how to build the right solution. 

Whether it is modeling complex systems, orchestrating multi-step logic streams, or acting as a more powerful foundation for creative coding, this model makes tough tasks feel simpler.

We are shipping 3.1 Pro at scale starting today:

For developers: Available in preview via the Gemini API in Google AI Studio, Gemini CLI, Android Studio, and our agentic development platform, Google Antigravity. Note that these are in progress of rolling out.

For consumers: Available via the Gemini app and on NotebookLM for Pro and Ultra subscribers.

For enterprises: Available in Vertex AI and Gemini Enterprise.

While we have made a lot of progress, we are already looking ahead to further improving how our models work in broader, more ambitious agentic workflows. We will continue to refine performance around accurate tool selection, complex instruction following, and extended coding tasks.

Whether you are a developer, a student, or just brainstorming a big idea, we can't wait to see what you build and discover with this new level of intelligence.

#ai #programming #lifeatgoogle

We used Gemini 3.1 Pro to build a realistic city planner app, and the multimodal reasoning on display is incredible!

Some great work from Google DeepMind, Michael Chang and team using our new 3.1 Pro release. I wanted to share a little more about how this works.

You can watch how the model tackles complex terrain, maps out infrastructure, and simulates traffic to generate a high-quality visualization. It's a fantastic look at how we can use AI to reason through large-scale, real-world spatial constraints.

Here is a breakdown of what the model is handling under the hood:

🗺️ Terrain analysis: Gemini writes individual programs to generate topography and discover shorelines. It analyzes the flat areas and water placement to build a foundational map in the app's dashboard. 

🌉 Infrastructure mapping: The model seeds the map with urban hubs based on resource locations, layering in basic wireframes before algorithmic generation takes over. 

🚗 Traffic simulation: To build the roads, the app simulates thousands of trips between these hubs. By analyzing how people would theoretically travel, it generates the optimal road networks around the natural terrain. 

✨ High-quality visualization: The finale generates a polished, photorealistic satellite view of this fictional city. From there, clicking anywhere on the map prompts Gemini to imagine and generate detailed "aerial photography" perspectives of specific neighborhoods.

As noted in the demo, Gemini 3.1 Pro has taken a massive leap forward in caring about user intent - seamlessly combining text prompts, spatial math, and image synthesis. We'll also see if we can open-source some of the prompts involved :)

Gemini 3.1 Pro is now available and rolling out to your favorite developer tools.

#ai #programming #softwareengineering

Announcing the new Google Gemini powered Docs, Sheets, Slides, and Drive! More context, fulled editable AI slides, and new grounding sources!

Available today to G1 Pro and Ultra users!

🚀 Google is fundamentally reimagining how we work across Google Docs, Sheets, Slides, and Drive. 

I'm incredibly excited to see Google Workspace become more contextual, helpful, and collaborative. We've all stared down a blinking cursor or an empty spreadsheet - getting started is often the hardest part. Now, you can simply choose your sources, and Gemini will securely connect the dots and uncover useful insights for you.

Here is a look at the new way of working:

📝 Docs is your new writing partner
Get a customized first draft instantly by asking Gemini to draw from your existing files. You can effortlessly refine sections, strengthen your message, and seamlessly match your unique voice or a specific document format.

📊 Sheets build complex sheets 9x faster

Describe what you need, and Gemini will set up the entire spreadsheet, pulling details directly from your inbox and files. You can even populate missing data instantly by accessing real-time web results - saving hours of manual research. (Plus, Gemini in Sheets is now #1 on the SheetsBench leaderboard!)

🎨 Slides let you go from concept to Presentation-Ready faster

Go from a rough concept to a polished deck with a simple prompt. Gemini acts as your design collaborator to generate beautiful, fully editable slides that align with your theme and pull context from your existing work. (Generating entire decks from a single prompt is coming soon!)

📂 Drive is your active collaborator

When you search in everyday language, Drive now generates a summarized AI Overview right at the top of your results. You can also use "Ask Gemini" to have a multi-turn conversation with your files and folders - all without opening a single document.

These features are rolling out today in beta for Google Workspace Gemini Alpha customers, as well as Google One AI Ultra and AI Pro subscribers.

#googleworkspace #programming #ai

SDE & AI @Amazon | Simplifying software engineering, system design, and generative AI | Building Hungry Minds: the newsletter to stay ahead in software 👇

Tell me about the last time you felt growing as a software engineer. Really.

𝙽𝚞𝚕𝚕𝙿𝚘𝚒𝚗𝚝𝚎𝚛𝙴𝚡𝚌𝚎𝚙𝚝𝚒𝚘𝚗? Here's why:

We all face information overload as software engineers. 
There's too much content and too little time.
You have to stay relevant, but it's hard.

Right now, you're either working or taking a break.

You are hoping to get a dopamine shot from scrolling your feed.

You can find:

💡 A practical tip to implement
📚 An inspirational quote
😆 Something funny

And then what?

→ Scroll past it? 
→ Forward it blindly to a friend?
→ Forget about it in the bookmarks?

The harsh truth: nothing.

What you consume passively doesn't make you learn.

How do I know?
It was the same for me for years.

How do you stay updated with AI, stay relevant in the job market, and continuously grow?

That's the problem I set out to solve with Hungry Minds - a free weekly newsletter to help software engineers learn and stay updated, without the noise.

I'm doing that by:

1. Writing to you daily on LinkedIn to 40K+ followers and over 5M impressions last year
2. Compressing 1 week of software deep dives and trends in 1 free newsletter issue
3. Sharing my thoughts and learnings being promoted to SDE II in 1.5y at Amazon

If that sounds interesting, check out Hungry Minds or DM me! I'm always down for a chat.

Save 50+ hours of your week here: 𝗵𝘁𝘁𝗽𝘀://𝗵𝘂𝗻𝗴𝗿𝘆𝗺𝗶𝗻𝗱𝘀.𝗱𝗲𝘃 (6K+ engineers from startups to big tech are reading already)

DISCLAIMER: Opinions expressed are solely my own and do not express the views or opinions of my employer.

My career background:

I'm currently an L5 software engineer at Amazon in Madrid, working on core Search CX and Personalization. 
In 2022/2023, I delivered 2 successful billion-scale projects with multiple teams across different time zones.

I have worked with multiple code languages (although this doesn't make a difference), and with tech stack in the lines of AWS CDK, React applications, scheduling systems, backends with GraphQL, and simple AI applications.

My education background:

- I finished a Master's in Machine Learning and AI with l'Ecole Polytechnique and IPP
- I completed a Master's in Computer Science and Data Science with ESILV Paris
- I got a Master's in business development and marketing with EMLV Paris

Alexandre Zajac

My typical day as an AI engineer:


9 AM: Ask Claude to write the code
9:05 AM: Ask Claude to fix the code Claude wrote
9:30 AM: Explain to Claude why its fix broke the original fix
10 AM: Give up and write it myself
10:15 AM: Realize Claude was right the first time
11 AM: Meeting to discuss "AI strategy"
12 PM: Manager asks if AI can replace the meeting
1 PM: Deploy to prod using vibes
2 PM: Prod is down
2:01 PM: Ask Claude to fix prod
3 PM: Write a LinkedIn post about how AI 10x'd my productivity
5 PM: Read Hungry Minds when Claude can’t fix prod: hungryminds.dev


What did I miss?


📚 Repost to help others use Claude
👨‍🔬 Follow Alexandre Zajac + turn on notifications

No alternative text description for this image

Leetcode has more than 3525 coding questions.

But only a few make you learn the important patterns.

Here are 25 of my favorite questions that will help you pass tech interviews:


___
btw my free interview cheatsheet: https://lnkd.in/d_eCGEJR
___


Graphs:

0. Clone Graph: https://lnkd.in/dV2P2TEs
1. Course Schedule: https://lnkd.in/de8Q3NBS
2. 01 Matrix: https://lnkd.in/dV2P2TEs
3. Number of Islands: https://lnkd.in/drT2MpTz
4. Rotting Oranges: https://lnkd.in/dUQVwJ-d

Arrays:

5. Insert Interval: https://lnkd.in/dfcEDFwB
6. 3Sum: https://lnkd.in/duGvuCjf
7. Product of Array Except Self: https://lnkd.in/dkGkjQVk
8. Combination Sum: https://lnkd.in/d3iStbGc
9. Merge Intervals: https://lnkd.in/dmFZxrVQ

Stacks:

10. Evaluate Reverse Polish Notation: https://lnkd.in/d-y7Zw4C
11. Min Stack: https://lnkd.in/dqbh7PeV
12. Trapping Rain Water: https://lnkd.in/dS_svBAm

Binary Trees:

13. Binary Tree Level Order Traversal: https://lnkd.in/dM-VYbVB
14. Lowest Common Ancestor of a Binary Tree: https://lnkd.in/dUvJykgA
15. Serialize and Deserialize Binary Tree: https://lnkd.in/dW2cP5Wn

Dynamic Programming:

16. Maximum Subarray: https://lnkd.in/dvjYye6E
17. Coin Change: https://lnkd.in/d7zZRg7H

Binary Search:

18. Search in Rotated Sorted Array: https://lnkd.in/dEuh3gie
19. Time-Based Key-Value Store: https://lnkd.in/dbERGKUB

Strings:

20. Longest Substring Without Repeating Characters: https://lnkd.in/d_vZrZda
21. Minimum Window Substring: https://lnkd.in/de8aeeQD

Heap:

22. K Closest Points to Origin: https://lnkd.in/dUtCqYf4
23. Find Median from Data Stream: https://lnkd.in/ddDgWqUv

Recursion:

24. Permutations: https://lnkd.in/dTUqmAfy


What are you studying right now?


_______
📚 Save this for later & repost to help others!
👨‍🔬 Follow Alexandre Zajac + turn on notifications

SDE & AI @Amazon | Building Hungry Minds to 1M+ | Daily Posts on Software Engineering, System Design, and AI ⚡

taplio

The Battle

Addy Osmani

Alexandre Zajac

Followers

Posts

Engagements

Likes

Performance ScoreAddy Osmani's LinkedIn Performance

Performance ScoreAlexandre Zajac's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Addy Osmani

Alexandre Zajac

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);}Addy Osmani's LinkedIn Performance

Performance ScoreAlexandre Zajac's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreAddy Osmani's LinkedIn Performance