Hugo Lopes

Personal Intelligence @GoogleDeepMind. Giving Gemini a Memory and learning about you through time.

[Hugo Lopes @HJDLopes]

Remembering is not understanding. 🧠

We can give an AI "long-term memory" to store 10,000 facts about you, but that doesn't mean the system actually gets you. A thread on why AI memory needs a rewrite. 🧵

[Marc Stogaitis @marcsto]

Overwhelmed by the kind notes about the Android earthquake alerts in Venezuela. Thank you! I’m passing your kind words along to the incredible team behind the scenes who build this system.

We're grateful it could provide some help. Thinking of everyone in Venezuela right now.

[Hugo Lopes @HJDLopes]

Thank you @marcsto . I'm honored to work on the same company where such a team exist. This is exactly what to strive for - helping people when they need the most.

[Marc Stogaitis @marcsto]

Excellent breakdown by the @nytimes explaining how our Android Earthquake Alert System worked during the recent earthquakes in Venezuela.

https://t.co/mpc46IyU6r

[Logan Kilpatrick @OfficialLoganK]

Gemini Omni Flash is SOTA at image to video, text to video, and video editing : )

Excited to get this to developers in the API soon!

[Hugo Lopes @HJDLopes]

@garrytan 's approach is also extremely interesting and it is worth checking out. But with so many options on the market the most important thing is to just "start". Don't procrastinate.

[Hugo Lopes @HJDLopes]

Remembering is not understanding. 🧠

We can give an AI "long-term memory" to store 10,000 facts about you, but that doesn't mean the system actually gets you. A thread on why AI memory needs a rewrite. 🧵

[Hugo Lopes @HJDLopes]

You can check out open-source memory frameworks like @honchodotdev (by @plasticlabs ) to see this reasoning-first approach in action. Or create your own and integrate into @NousResearch 's Hermes assistant. Great weekend project.

[Hugo Lopes @HJDLopes]

It’s the difference between an AI simply stating, "I remember you live in Zurich," and an AI automatically structuring its code explanations to match your specific engineering style without being prompted.

[Hugo Lopes @HJDLopes]

Most "memory" today is just a standard dump-and-retrieve vector database. To actually learn about users, we need to shift to a "reasoning-first" architecture.

Here are 3 ways to tackle "learning how to learn" in AI systems: 🛠️

[Hugo Lopes @HJDLopes]

And this is our small model :) amazing capabilities. We're cooking even more around personalization

[Logan Kilpatrick @OfficialLoganK]

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own.

We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!

[Josh Woodward @joshwoodward]

Most Al chatbots give you basic "projects." Gemini just built you a second brain. 🧠

Introducing Notebooks: some of the magic from @NotebookLM, integrated directly into @GeminiApp.

Here's what changes for you today:

📚 Upload 100 sources for free

📂 Organize your chats - the wait is officially over :)

🔄 Sources, chats, and emojis sync

People are using Gemini and NotebookLM in tandem, and we'll keep building both.

To manage capacity, we're rolling this out NOW on the web and going from Ultra ➡️ Pro ➡️ Plus ➡️ Free. (Mobile, EU, and Workspace are up next!)

With Google I/O right around the corner, we are just getting started. Enjoy!

[Demis Hassabis @demishassabis]

Now it’s even easier to switch to the @GeminiApp ! 😎

[Hugo Lopes @HJDLopes]

That's why I love working solving this hard problem. We're getting closer @kar

[Andrej Karpathy @karpathy]

One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.

[Hugo Lopes @HJDLopes]

We're cooking even more!

[Sundar Pichai @sundarpichai]

We’ve heard how much people appreciate getting more personalized help with things like shopping, trip planning & more with Personal Intelligence.

So now it’s available for US users in AI Mode in Search and expanding to more users in the US in the @Geminiapp, and Gemini in Chrome.

https://t.co/NXKVcRsJEr

[Hugo Lopes @HJDLopes]

This has been the most challenging yet most rewarding project I've ever taken! Finally building what I wanted to have for years!

[Josh Woodward @joshwoodward]

The end of the "one-size-fits-all" response.

Last week, Gemini added memory globally.

Today, we go further. Everyone can now turn on Personal Intelligence in the 🇺🇸USA across Web, Android, iOS, and Chrome and connect Gmail, Photos, and more.

We're moving beyond generic answers and providing personal responses uniquely relevant to your life.

More countries coming soon!

[Demis Hassabis @demishassabis]

Excited to launch Gemini 3.1 Pro! Major improvements across the board including in core reasoning and problem solving. For example scoring 77.1% on the ARC-AGI-2 benchmark - more than 2x the performance of 3 Pro. Rolling out today in @GeminiApp, @antigravity and more - enjoy!