|
Happy Wednesday,
Today I'm giving you the prompt chain that got ChatGPT to fix my entire life in one afternoon, stuff I'd been avoiding for two years, then the Anthropic researcher who quit on Monday and the Anthropic safety lead who replied that there's a greater than 10% chance AI kills all humans (yes, this actually happened), and finishing with the three craziest things people have built with GPT-6 Astra this week.
Let's get into it.
|
IN THIS EMAIL
» ChatGPT can now fix your entire life in one day
» An Anthropic researcher quit because AI might kill us all
» The 3 craziest GPT-6 Astra builds this week
|
🧠 DO THIS TODAY
ChatGPT Can Now Fix Your Entire Life in One Day
Everyone has a list they don't write down. The subscription you've meant to cancel since spring, the dentist you keep not booking, the email you've drafted in your head forty times. None of it is hard. It's just been on your shoulders long enough that you've stopped noticing the weight.
So I asked ChatGPT to fix all of it in one day, and it knocked out stuff I'd been avoiding for two years in a single afternoon. Three prompts, and the second is where it stops advising and starts doing the work.
Step 1: Open ChatGPT, switch from Chat to Work at the top, turn on voice dictation, and paste this in. Then just talk.
|
Interview me about everything in my life that's broken, unfinished, or quietly stressing me out. Go area by area: money, health, home, relationships, work, admin, digital life, anything else. Three short questions per area, and nothing is too small, include the stuff I'd be embarrassed to say out loud. When I say DONE, turn all of it into one list called The Weight: one line per item, grouped by area, with how long I've been avoiding it next to each one.
Then turn The Weight into a one-day plan. Sort every item by how fast it can be fixed, fastest at the top, except anything with a hard deadline or a health issue, which goes first regardless. For each item give the time it takes, exactly what to do, and a label: ME (only I can do it), YOU (you can do it for me right here), or SPLIT (you prepare it, I finish it). Put it in Canvas as a checklist I can tick off, with a start time for each block starting at [9am].
|
Talk your answers out instead of typing them, you'll say three times as much and this whole thing lives on how honest you are here. What comes back is every single thing sitting on you, in one place, with how long it's been there next to each one. Mine had a few with a two next to them. Years. Then it turns the whole thing into a checklist with times on it, quick wins at the top so you've ticked five things off before your coffee's done. And about half of it is going to say YOU.
Step 2: Now the trick. Paste this in.
|
Now do the YOU part of everything marked YOU or SPLIT, in order. Draft every email and text ready to send. For each subscription, find the exact cancel page and write the steps. For each appointment, find the booking page and prep what I need. Use the browser for anything that needs it. Details: I live in [CITY], the subscriptions are [APPS]. Ask me before you log into, pay for, send or cancel anything. Check each item off in the Canvas as you go and tell me exactly what's left for me.
|
And then it just starts. It drafts the texts you've been avoiding, finds the cancel page for every subscription you forgot about, pulls up the booking pages for the appointments you keep putting off, and ticks them off in front of you one by one. It checks with you before anything gets sent, paid for or canceled, but the hard part, the part that's been stopping you, is done.
Step 3: One more, because this is the one that gets people.
|
Now tell me one thing on this list that would change my life the most if I actually did it today. Explain why in three sentences, then give me the first ten minutes of it, step by step, with the clock running from 0:00.
|
It's never the thing you expect and it usually hits harder than you'd like. Do the ten minutes. Four more prompts in the guide, including one that adds up what every subscription you're not using has cost you since you stopped using it.
Get my full fix-your-life prompt guide →
🚨 THIS IS SERIOUS
An Anthropic Researcher Just Quit Because He Thinks AI Could Kill Us All
|
This is the one that absolutely broke the internet today. On Monday a 27 year old researcher named Jacob Coxon quit Anthropic and posted why, and his reason is going to blow your mind. He'd spent three years actually building these models himself, first at OpenAI then Anthropic, and he said in his post that neither company is acting responsibly, that they're racing straight to self-improving superintelligence and gambling with our lives, and that the people building AI earnestly believe it could kill us all by the end of the decade. Then the next day Evan Hubinger, who runs alignment science at Anthropic, whose actual job is making sure these models are safe, replied and agreed with him. His words: "Jacob is correct here, we really do earnestly believe AI could kill all humans." He put his own odds at more than 10% within the next decade, and said Anthropic does not yet have a plan to solve this for superintelligence and is not clearly on track to.
That isn't a guy who's angry about his job. That's the person in charge of safety at the company that markets itself as the safety company, saying on the record that they don't have a plan. He did add that the risk from today's models is low and what worries him is what comes after, models improving themselves faster than anyone expected.
The part that matters isn't the scary quote though. It's why nobody stops. There's an old thought experiment called the prisoner's dilemma, and it is exactly what's happening here. Two people get arrested and put in separate rooms. If they both stay quiet, they both get a light sentence. If one talks and the other stays quiet, the talker walks free and the other gets the maximum. If they both talk, they both get a heavy sentence. From inside either room, whatever the other person does, you're better off talking. So both talk, and both end up somewhere worse than if they'd trusted each other.
Now swap the prisoners for AI labs. Everyone in this race agrees slowing down would be safer. But if OpenAI slows down and Anthropic doesn't, OpenAI loses. Zoom out a level and if America slows down and China doesn't, America loses. So the sensible move for every player, on their own, is to keep going, and everyone agrees it should stop and not one of them can be the one to stop. You don't need a villain for this to go wrong. You just need a lot of smart people making the reasonable choice for themselves, over and over, until together they've run somewhere none of them would have chosen. And people inside a dilemma like that can't fix it from the inside, which is why Coxon's only answer was a coordinated pause, which lands on governments, and you've seen how fast those move. Anthropic was literally founded by people who left OpenAI over safety, they built the company to escape this exact trap, and this week the guy who runs safety there said they don't have a plan either.
|
And here's how fast the race is actually moving. On Monday OpenAI posted that their agents had solved a famous math problem with a million dollar prize on it, and buried in the post was this line: the proof was produced using a next generation model "significantly more capable than GPT-6 Astra." Astra came out a week ago. So the most powerful model on the planet already isn't, and I'd bet the reason is simple: Astra is good enough to help build whatever comes after it, and that one will help build the next. That's the cycle, and it doesn't end.
I'm pro AI, you all know that, I spend every day teaching people to use this stuff and I'm not stopping. But I agree with him, and I'm not going to sit here and tell you this is fine. What do you actually do with it? The same thing you were doing. The models on your phone today are not the danger, and the people who understand these tools will be in a far better position than the people who don't, whatever comes next. Keep learning it, keep using it, and keep paying attention to what the people building it say when they think they're only talking to each other. Although after all that, it might be time to go buy a bunker.
Read the full story here →
🤯 THIS IS INSANE
The 3 Craziest GPT-6 Astra Builds This Week
On Sunday I told you GPT-6 Astra is the most powerful AI model on the planet. Six days later people are completely losing their minds over it, because it's way better than even the launch made it sound. Here are the three builds I couldn't stop looking at this week.
1. The entire human body, in 2,234 pieces
via @ftcarpe on X
A developer called Ashe gave Astra one prompt and it built a full 3D website of the human body you can pull apart into 2,234 individually modeled pieces. Every bone, every organ, fifteen body systems, and you can rotate it, click any piece and read what it does. It's live right now and free, and medical schools pay six figures for software like this.
2. Playable video games from one sentence
via @luckeyfaraday on X
People are typing a single sentence and getting full playable games back. One guy has a test he runs on every new model, ask for Minecraft in one prompt, one shot, one file, and he's run it on Fable, Grok, Gemini, all of them. Astra is the first one that actually came back with Minecraft. Someone else built a Mario Kart style racer you can play in your browser, someone built Fall Guys in a couple of minutes.
3. Nine bad photos of an office, turned into a 3D model
via @rpnickson on X
Someone uploaded nine bad photos of his studio, just snapped on a phone from different corners, and asked Astra to build the space. It worked out where everything was in relation to everything else, stitched the photos together, and handed back a fully interactive 3D model of the room you can walk around in. Nobody told it the dimensions or where the desk was. It figured it out from the pictures. This one is just genuinely insane and blew me away.
All three of these blew my mind, and trust me, there are so many more I didn't even cover. Go try GPT-6 Astra right now inside ChatGPT. On Plus it's under Work rather than your normal chat, so switch that at the top.
Try Astra today →
Run the first prompt tonight and actually talk to it for twenty minutes. Watching the second one clear half your list is the best feeling I've had using AI in a while.
See you Sunday.
Maverick
P.S. Want your business to actually show up inside AI? The brands that rank are seeing up to 3x the conversions. Get your free AI visibility score and see exactly where you stand inside AI.
P.P.S. If you're new here or you ever skip one, I've got a newsletter archive of all my past emails you can catch up on anytime.
|
