Make your AI do your morning check

Signal Preview

⚡ Set up your AI to do your morning check
🗣️ Hear Dessa talk with real people
🏗️ See what Skyfall says its AI can do for engineers
📈 What Anthropic's IPO filing means for your Claude plan
🍓 A browser that runs on your ChatGPT plan
🧂 Plus: check any big AI claim with 3 simple questions

You've sent me a lot of frustration about scheduled tasks. The same four came up again and again:

The last one happened to me. I asked Grok to run a task at 8:15 every morning. It said "Automation created." Then I looked at the card: Daily, 1:45 PM. It had saved 8:15 in UTC, not my time. If I hadn't checked, my morning brief would have come after lunch.

So I made one page that fixes all four:

Not sure a scheduled task is for you? The prompt's first question helps you find one job worth giving it. If it doesn't save you time in 7 days, delete it. The page shows how.

Open it on your computer, with your AI app in the next tab.

3 AI tools launched this week that I'm really excited about. Here's who each one is for, and one line from me on why it caught my eye.

You ask ChatGPT to find 20 companies, and it gives you a list. Then you spend an hour opening LinkedIn and pasting each one into a sheet.

Strawberry is a new browser for Mac and Windows with AI helpers built in. They work inside the pages you're already logged into. For example, it can look up companies on LinkedIn, add them to a Google Sheet and rate how well each one fits.

Already pay for ChatGPT? Then you can bring your own account. Connect your paid ChatGPT plan, and Strawberry runs your tasks on it instead of burning its own credits on every job.

On Claude or Grok? You still get 500 free credits a month, and normal browsing uses none. Paid plans start at $20 a month.

Why it caught my eye: I always say use the AI you already pay for before you buy another one. This browser gets more out of the ChatGPT plan you already have.

One tip before you start: it asks what it may do without you, app by app. Keep "ask me first" on for your email until you trust it.

You make a poster in ChatGPT, and the headline has a typo. You ask it to fix just the word, and the whole picture changes.

SurgePix splits your image into layers: the person, the background and the text. You fix one layer by chatting, and the rest stays exactly as it was.

You can also save your brand colors and fonts once, and every new image stays on-brand. It's made for posters, social posts, banners and simple infographics, even if you're not a designer.

Why it caught my eye: I make a lot of images with AI. Redoing a whole picture for one wrong word costs me time and empties my limit.

ChatGPT and Grok can make you a short clip. But a 2-minute video means 10 clips and a whole evening stitching them together.

MOVO is a video agent for iPhone and the web. You tell it your idea in chat. It plans the scenes, makes them, keeps the same characters in every shot and puts the whole video together.

It makes stories of 2 minutes and longer. And you can lock your phone while it works. You come back, check the video and ask for changes in chat. New users get 100 free credits.

Why it caught my eye: I do a lot of my work from my phone. A tool that keeps going while my phone is locked is my kind of tool.

You use voice mode, you stop to think, and the AI jumps in. Or you try to stop it, and it keeps talking.

Block (the company behind Square and Cash App) showed Dessa on October 23. It listens and talks at the same time, so it can say "yeah" or "got it" while you're still talking, like a friend on the phone.

Two things caught my eye. Block says it learned only from real people talking, with no text at all. And it runs on a laptop (a MacBook Pro M4 Max or newer), so your voice can stay on your own computer.

You can't use it yet. It's a research preview, and only a small group of safety researchers get early access. There's a public waitlist, no price, and Block hasn't shown how it compares with other voice AIs.

So nothing to change this week. But you can hear it today. Block posted live recordings of real people talking to Dessa. Dessa picks what to say and when, in its own voice. Listen to one and tell me if it sounds like a person to you.

Every week some new AI says it beats ChatGPT. This week it's an Indian-founded startup, and the numbers are huge.

Sam Pasupalak, co-founder and CEO of Skyfall AI, announced SKY-001 on October 19. It's made for CAD and BIM, the drawings engineers and architects use to design parts and buildings.

Skyfall says it's 51 to 125 times faster and 150 to 210 times cheaper than frontier models like GPT-6 Astra on its engineering tests. For example, one CAD drawing costs $0.005 with SKY-001 and $0.75 with GPT-6 Astra.

It works one step at a time. It breaks a product into parts first, then designs each part. A team of 25 across India, San Francisco and Toronto built it in 18 months. They're the same team behind Maluuba, the AI lab Microsoft bought.

For now it's only a waitlist. No API, no price, no way to try it yourself. Full launch is planned for early 2027.

Not an engineer? Watch Sam's 2-minute video anyway. Close to 5 million people have. Then scroll down to the hype check, where I check those "125x faster" numbers.

Maybe you use Claude every day. Yesterday, the company behind it filed to sell shares on the stock market.

Anthropic filed its papers with the SEC (the US stock market regulator) on October 26. It plans to list on Nasdaq under the name ANT.

The Wall Street Journal says it could raise $100 billion or more, at a value of $1.5 to $2 trillion. The first trading day could be the week of November 16. These numbers come from the Journal's sources, not the filing, so they can still change.

The filing does show one real number: $16.3 billion in revenue in the first half of 2026. A year earlier, it was $1.16 billion. That's 14 times bigger in one year.

So what changes for you today? Nothing. Your Claude app, your plan and your price all stay the same.

I'm not a financial advisor, so I won't tell you whether to buy shares. I pay for Claude because it's the brain of my AI work. Owning a piece of the company is a different decision.

Every week, one big AI claim, and how to check it in 30 seconds. This week it's SKY-001: "51 to 125 times faster, 150 to 210 times cheaper."

I ask 3 questions:

So basically, it might be true, for one kind of work, and we can't check it yet. It's use case specific. A big number like that doesn't make your ChatGPT plan old overnight.

You can run the same check on any AI headline. Paste the article into your AI with this: