In partnership with
Good morning. ChatGPT just stopped being a text box. OpenAI is putting GPT-6 in front of every user, and the answers now come with buttons, charts and small working tools. Elsewhere, Google opened its AI watermark checker to the public and SpaceX is teaching Grok Bot to borrow its rivals' models.
1,000+ Proven ChatGPT Prompts That Help You Work 10X Faster
ChatGPT is insanely powerful.
But most people waste 90% of its potential by using it like Google.
These 1,000+ proven ChatGPT prompts fix that and help you work 10X faster.
Sign up for Superhuman AI and get:
1,000+ ready-to-use prompts to solve problems in minutes instead of hours—tested & used by 1M+ professionals
Superhuman AI newsletter (3 min daily) so you keep learning new AI tools & tutorials to stay ahead in your career—the prompts are just the beginning
TODAY IN AI
1. OpenAI brought GPT-6 to every ChatGPT user, with a new "Intelligent UI" that answers in charts, forms and tappable buttons instead of plain text. Paid plans (Plus, Pro, Business, Enterprise) got GPT-6 Sol on Oct 7. Free and Go users get the cheaper GPT-6 Luna starting Oct 8, and the rollout reaches accounts gradually. OpenAI's examples include a recipe with a tickable cooking schedule, a bill splitter and a savings calculator, all built inside the chat. It also says GPT-6 can start answering while it is still reasoning, cutting the wait on web-search questions by about 44% against GPT-5.6 Instant. Those speed figures come from OpenAI's own tests, and the company admits the model's design sense "still has room to improve." Older Windows and macOS desktop apps don't support the new layouts, so use the web or the latest mobile app.
2. Google opened its SynthID Detector to everyone, in English, after limiting it to journalists and researchers. You upload an image, video clip or audio file and it checks for invisible watermarks from Google, OpenAI, NVIDIA and Kakao, with Apple support promised soon. The limits matter. It allows roughly 10 checks per user per day, and a clean result does not prove a file is human-made, because some models never add a watermark and edits can weaken one. A hit also can't tell you whether AI generated the whole file or only touched it up.
3. SpaceX says Grok Bot will route work to rival models. In an Oct 7 post on X, Elon Musk named Anthropic's Claude Opus 5.5, Midjourney and Suno as back ends it could pick for a given task. RuntimeWire calls it a direction, not a launch. Musk didn't say how models get chosen, which tasks go where, or what data reaches outside providers. Grok Bot itself is an early-beta cloud agent that launched Aug 11 with automatic routing, and users can't pin a model.
Leave Granola and get up to 12 months free of Wispr Flow Notetaker + Dictation
If you have paid time left on an individual Granola plan, we'll match it with a Wispr Flow subscription that includes Notetaker and dictation, and add bonus time, up to 12 months total. Sign in or create a Wispr account and submit proof of your plan to check eligibility.
FROM THE FRONTIER
The laptop that runs the agent. Microsoft's new Surface Laptop Ultra is built around NVIDIA's RTX Spark Superchip, up to 128 GB of shared memory, and a pitch of "Run local. Run free." Microsoft says it can run local agents on models up to 120B parameters. It costs $2,599.99 and is open for pre-order with no shipping date on the page.
The part nobody sells. An agent that can act on your files needs a fence around it. Microsoft's open-source MXC sandbox is one attempt: policies set which paths an agent can read or write, and which network access it gets, across Windows, Linux and macOS. The repo is MIT-licensed, and its own README calls it an early preview. It says some generated policies are overly permissive and that "no MXC profiles should be treated as security boundaries currently."
The reframe. This week's theme is software that decides its own shape. ChatGPT picks a chart over a paragraph, Grok Bot picks a model, and a laptop picks local over cloud. Each choice is invisible to you. The pitch is convenience, but the real question is who checks the choice, and the sandbox's own README says that part isn't finished.
A deal has two sides. Your stack only sees one.
Aligned shows you the other half. See what buyers actually do between meeting. See what they open, revisit, and share; so you know the exact next step to move things forward. And your champion gets everything they need to win internally when you're not in the room.
IN THE KNOW
Games on demand. Google Labs' Playground lets you describe a game in chat and play or share it, with no coding. A subscription may be required, and it's limited to select countries and users 18 and over.
Grok Bot playbooks. SpaceXAI published guides for using Grok Bot in marketing, sales and recruiting, including setups for SDR outbound and founder-led sales.
Grok Bot pricing. Access comes through three paid plans: Cursor Premium Teams at $120 per seat per month, Cursor Ultra at $200, and SuperGrok Heavy at $300. Enterprise buyers join a waitlist.
Fewer visuals, please. If the new layouts annoy you, turn off "Layout and visuals" in web settings. It won't force text-only replies, and some visuals may still appear.
SpaceX's chip bill. SpaceX is reportedly seeking about $40B to buy NVIDIA chips, roughly $10B in bank loans and $30B in debt led by Apollo. Some investors balked at a two-page memo heavy on space photos.
PROMPT STATION
Turn a decision into a calculator you can reuse. GPT-6 can build small working tools inside the chat. Give it a real decision and your own numbers, and you get something you can tweak instead of a paragraph you'll forget.
COPY AND PASTE THIS PROMPT
I'm deciding whether to [DECISION, e.g. hire a freelancer, buy a tool, or raise my prices]. Build me a small interactive calculator in this chat. First ask me for the 5 numbers that matter most. Then show a simple chart of the best case, worst case and likely case, with a slider for the one assumption I'm least sure about. Finish with one line saying what would have to be true for this to be a bad idea.
Tip: run it with your real numbers, then move the slider you trust least. If the answer flips, that assumption is what you should research before you decide.