Improvisation at its best: printed 300 business cards with an extra “r” in my email for a huge conference—so I bought the typo domain on the spot. Chaos contained.
Retreating back to the safety of LinkedIn QR codes.
Best,
Marrily
Just scrolling Marketplace and there’s a casually listed humanoid robot like it’s a used coffee table. Absolutely normal times.
It was fantastic meeting Priscilla Chan & Mark Zuckerberg last night. The Chan Zuckerberg Initiative will now concentrate on AI & scientific research, led by a network of research centers called Biohub to accelerate discoveries that design treatments to change lives.
In 2024, I visited the 1X offices and checked EVE out, their current robot at the time. I also saw NEO nearby, which looked remarkably human—almost like someone in a suit. The NEO PM told me, “Give it a year, and everyone will have one of these at home”.Well, here we are… pre-order is in 🤖
POV: Getting into an argument with Sora about who is the real content creator
sora.chatgpt.com/p/s_68e456a2...
Fast forward to today: 18 years (216 months) later, that adventure became the journey of a lifetime from moving continents, to a PhD in AI to joining big tech in 2013, to founding my own education & learning company, to becoming an author and getting to work with and meet incredibly talented people.
On this day back in 2007, I packed my life into a suitcase and left Greece — for what was supposed to be a 12-month adventure to leave everything behind and study computer science.
Me: I’ll give folks a few minutes to join my AI Agent webinar
Also me, 60 seconds in: 👀 473 participants already
✨ Experimentation is amazing but we need to also start thinking about moving from experiments → operationalization. It’s not about side projects. It’s about integrating AI into roadmaps, KPIs, and ROI.
✨ Biggest myths we had to unlearn:
-“AI is cheating”
-“AI will replace PM intuition”
Back on stage after mat leave at Women In Product's 10th Conference. 1000 people, AI everywhere. I shared thoughts on moving from AI hype → AI how:
✨ All PMs will be AI-first - this needs to be our default lens for solving problems.
🚨 It’s called a bootcamp for a reason 🏋♂️ For our September's AI PM Bootcamp cohort, I’m assigning an engineer to every single team — so you don’t just design a capstone project, you *actually* launch it.
lnkd.in/gdSTqQy2
Today I learned that AI can help you interpret a baby’s cry 🧐
Nano Banana is totally bananas 🍌 For folks that haven't tried this yet, it's an advanced AI image model from Google DeepMind that allows you to edit and combine photos with text.
Gemini comes to Google Home! This is the team I'm a part of.It was fun to explore what GenAI can do on communal devices.Some use cases I briefly worked on:“add the ingredients to make an authentic Italian lasagna to my shopping list”,”set a timer for perfectly blanched broccoli” 🥦
lnkd.in/eWxY9fQ5
Teaching a “Building AI Agents” workshop while also on hold for 90+ minutes waiting for 'the next available agent' to sort out an insurance bill. The irony...
Product Managers should view AI as a colleague or rather as a team of really smart AI interns.
Prompting ≠ Chatting with AI.
✨ Chatting is conversational. You ask a vague question ("Tell me about go to market strategies") and get a vague, textbook answer. It's fine for simple queries, but not that transformative.
✨ Prompting is directive and it's the skill of giving structured instructions.
Thought of the day: Use AI quietly. The less we talk about AI, the better. The more we talk about what it enables, the better.
Deep voice conversations with advanced AI are an underrated luxury most people haven’t discovered yet.
I got to meet Sundar Pichai last week! Still taking in everything from Google I/O — the learnings & energy — but getting the chance to meet our CEO after 12 years at Google was *easily* the #1 highlight. Sundar visited our AI Builders group at the AI Pavilion & the excitement was obvious. #googleio
AI Builders got their own actual, real TPU at Google I/O today. Coolest swag ever.
#googleio
🚨 Top AI announcements on Google I/O
✨ 𝗔𝗴𝗲𝗻𝘁 𝗠𝗼𝗱𝗲
✨ 𝗔𝗜 𝗠𝗼𝗱𝗲
✨ 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗠𝗮𝗿𝗶𝗻𝗲𝗿
✨ 𝗡𝗲𝘄 𝗽𝗲𝗿𝘀𝗼𝗻𝗮𝗹𝗶𝘇𝗲𝗱 𝘀𝗵𝗼𝗽𝗽𝗶𝗻𝗴 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲 𝗶𝗻 𝗔𝗜 𝗠𝗼𝗱𝗲!
✨ 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗔𝘀𝘁𝗿𝗮 ⭐
✨ 𝗧𝗿𝘆 𝗼𝗻 𝗬𝗼𝘂
✨ 𝗣𝗲𝗿𝘀𝗼𝗻𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗼𝗻 𝗚𝗲𝗺𝗶𝗻𝗶 𝗮𝗽𝗽
✨ 𝗩𝗘𝗢 3
✨ 𝗡𝗲𝗮𝗿 𝗿𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝘀𝗽𝗲𝗲𝗰𝗵 𝘁𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗶𝗼𝗻
✨ 𝗝𝘂𝗹𝗲𝘀
✨ 𝗙𝗹𝗼𝘄
#googleio
www.linkedin.com/posts/marily...
Lately, there’s been a surge in products branded as 'AI Agent for X.' But I believe we’re moving towards a future where simply saying 'X' will implicitly mean it involves AI, and inherently agentic capabilities. How long before users start expecting AI as a default, rather than a differentiator?
I use different AI tools for different use cases. My new discovery that has helped me a lot on every day dillemmas is Perplexity's 'social-only' search that summarizes relevant discussions (with links to actual threads) primarily from Reddit.
AI ≠ Product
AI Agents ≠ Chatbots
AI Product Management ≠ AI *for* Product Management
AGI ≠ Just a bigger, better AI
AI ethics ≠ a solved problem
your skills + AI ≠ your skills OR AI
AI in Healthcare ≠ Just diagnostic tools
Big AI models ≠ Inherently smarter
AI automation ≠ Job replacement
In AI Products,we are building 𝘂𝗻𝗳𝗮𝗺𝗶𝗹𝗶𝗮𝗿 experiences for users. Experiences they have never seen before.Truth is that users can tell you what’s broken about the path towards a use case, they won't tell you how to fix it.Listen, observe behaviors then design what they didn’t know is possible with AI
If you’re overwhelmed, then you're on the right track. It's when I feel I've got a good handle of things when I realized something is wrong. Don't wait, start with the fundamentals and get used to adapting as things shift.
Someone told me today they were waiting for AI to “settle in” before diving in. AI isn’t settling. It’s evolving. Rapidly. Every day brings new tools (Gemini 2.5, image editing, ghibly-style, new ways to get things done, new companies, new demos, new comparisons... and so on).
and then have them talk to each other!
Seriously considering an AI experiment: training two agents, InnieGPT and OutieGPT. My own digital "Severance".
InnieGPT trained on my pre-5pm work-related GenAI needs: content creation, PRD generation, efficiency hacks. OutieGPT, on the other hand, gets *the entire* rest of my search history..
