Google’s best new AI feature is just a really good to-do list

Google’s big I/O event came and went last week, stuffed to the gills with new AI announcements and functionality. Most of it left me cold.

But one — and only one — of those Gemini announcements is actually making a difference for me in the week following Google I/O, and it’s relatively humble: Daily Brief, a Gemini-generated daily to-do list based on your Google Workspace data.

Gemini’s Daily Brief is an AI feature done right

Sitting in the left-hand sidebar of the Gemini app, Daily Brief works hand-in-hand with Google’s Personal Intelligence feature, which connects core Google services like Gmail, Google Calendar, and Google Drive to Gemini. You’ll need to enable Personal Intelligence (tap your profile icon in the Gemini app, then tap Personal Intelligence) to make the Daily Brief appear.

Now, I’ve been a vocal critic of Personal Intelligence in the past because it felt so gimmicky, popping up in Gemini chats with useless, pushy asides like “Since you’re already renovating your New York apartment…” that felt more like showing off than actually being helpful.

But Daily Brief does a terrific job of making Personal Intelligence truly useful, scanning your Gmail inbox, your Google Calendar events, and recent Gemini chats to deliver a once-a-day summary of to-dos.

Aside from a smarmy “Hey Ben here’s what today looks like” heading punctuated by a jumbo-sized emoji, Daily Brief is refreshingly direct. You get a series of bulleted lists with action items culled mainly from your Gmail and calendar, telling you where you need to go, who you need to reply to, what appointments you need to prepare for, and which chores need to be tackled.

I’m a sucker for to-do lists, but what makes Daily Brief even better are the action items under each entry. Event invitations will have an “add to calendar” button, while chores and tasks get icons for pulling up the relevant emails.

The killer feature for me, though, is the “Mark complete” button hidden in the three-dot “overflow” menu. (There’s also a Chat button that spawns a Gemini conversation about the item.) Tap “Mark complete,” and the associated Daily Brief to-do will be grayed out. It’s a deeply satisfying action that triggers a flood of endorphins in my brain.

I’ve been using Daily Brief for roughly a week now, and it’s been mostly spot-on about catching my key appointments and to-dos, as well as highlighting some important tasks that might otherwise have fallen through the cracks.

Daily Brief is accessible to users on Google’s affordable $20-a-month AI Pro plan (AI Ultra users can use it too, of course), and yes, it’s easy to turn off if you don’t want Gemini poking through your Gmail. Personally, I’m all in.

More in AI this week

• For his first encyclical, Pope Leo XIV held forth on the promise and perils of AI, and in doing so made a crucial point: AI can be a “valuable tool,” but only if used with “vigilance” and thoughtfulness. For everyday users like us, that means keeping our neurons firing while working with AI rather than switching them off. (PCWorld)
• Claude has been nagging its users to take a break and get some sleep, interrupting night owls and early birds alike. Anthropic says it’s looking into it. (PCWorld)
• It looks like Apple’s long-promised AI update for Siri is finally about to happen, complete with the ability to access your personal data, analyze on-screen content, add calendar events, launch apps, and search the web. (Bloomberg, subscription required)
• So, is corporate America having second thoughts about tokenmaxxing after burning through their AI budgets? Not everyone thinks so. (Simon Willison’s blog)
• You’ve heard of AGI (artificial general intelligence), meaning an AI system that could theoretically match or surpass humans in all meaningful thinking tasks. Well, the cool new thing is RSI: recursive self-improvement, or an AI that automatically and continuously improves itself without any humans in the loop. Yeah, it’s worrisome. (TechCrunch)

Prompt of the week: The “wrong question” meta-prompt

Not entirely satisfied with the answer you got from ChatGPT, Claude, or Gemini? Maybe you were asking the wrong question. Too specific, and the AI may give you a narrow answer that misses the big picture. Too vague, and you’ll end up with a wishy-washy reply. Or maybe you’re asking the model to solve a problem that doesn’t actually need fixing. 

Next time, try this: a “wrong question” meta-prompt that makes the AI interrogate your original prompt, challenging its preconceptions, structure, and framing while bridging the gap between what you’re asking for and what you really want the AI to do.

That’s all for now! 

Thanks for reading the latest issue of Prompt Mode. Want more next week? Don’t forget to sign up to start receiving this newsletter in your inbox.

See you next time!