Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
https://t.co/qr2iP0GpfX
I give my AI full access to my computer, and I recommend you do the same.
That's bypass permissions in Claude Code and full access in Codex.
The reason is pretty simple.
I've used AI this way on a plethora of tasks, in my own work and in other businesses, and without fail it has never done something I couldn't undo or that cost me greatly.
Nothing horrible security-wise or privacy-wise, and no very important file deleted.
Why do I encourage it? Well, because you can TRULY walk away from your computer and it keeps doing great work without you there, which is the true benefit of AI.
It saves you time so you can go do the other stuff you want to do.
No one wants to sit there all day waiting to click "Allow" on an action the AI could have just decided itself.
All it comes down to is making sure you don't tell it to do something awful in a prompt, which most people don't do.
And even if I do ask for something really sensitive or a poor decision by accident, it still decides to ask me anyway, because these models are that smart.
3 key questions for every consumer AI product:
1. Does your product give people new superpowers?
2. Does your product get better as more people use it?
3. Does your product get better when models get better?
Today in AI: November 17
California Takes OpenAI To Court:
California sued OpenAI on Monday over how it tests powerful AI agents.
The state says the company failed to contain them and misled the public about the risks.
The complaint cites at least 80 cases in which OpenAI allegedly lost control of agents during tests.
That includes July's attack on Hugging Face, a site that hosts AI models and code.
These agents can act across computer systems, so a failed boundary can let a test spill onto someone else's network.
The state wants civil penalties and court orders to stop the alleged unsafe practices.
The court has not ruled on those claims.
OpenAI says it takes the problem seriously and argues that the lawsuit singles it out over a challenge shared across major AI labs.
Separately, OpenAI announced deals with four outside teams to study its models and safeguards from inside the lab.
Those checks are starting, not a clean bill of health.
Chip Controls Follow The Cloud:
The US Commerce Department released a rule aimed at closing another gap: renting restricted AI computing through overseas data centers.
A company can rent computing power abroad without importing the chips itself.
The rule would require licenses for certain advanced-chip shipments and transfers to data centers serving companies based in restricted countries, including China.
It also covers some cases where a data center starts renting out large amounts of computing power after getting the chips.
The server can stay put; the export paperwork still finds it.
Commercial recommendation systems, such as those used by online shops, are exempt from that remote-access license requirement.
The rule takes effect when published in the Federal Register, the government's official rulebook.
Existing covered arrangements then have 90 days to get a license.
AI video editing is BY FAR my favorite use case of AI.
That's coming from someone who has used AI extensively across all the various models for years, taught it personally and for other businesses, and started out as a freelance video editor who has edited and posted over 1,500 videos.
Why is it my favorite? Well, because of how much time video editing has taken from me, and how boring I find it.
I'd venture to say a lot of video editors don't like video editing either, because it's super tedious.
Crafting a story and making it all come together can be cool, but mostly you're moving your mouse around and hitting shortcuts.
Now AI can edit my videos so well, in way less time, and I can still tweak it exactly how I want.
Which means I get to focus my time and energy on the more important stuff (the idea, the script, the hook) and make that as phenomenal as possible.
And yes, every creator can now do the same if they put in the work and set up their AI right, which is good, because it means more creation, more ideas and a real boom for the creator economy.
If you thought it's at the peak, it's not even close yet.
If you're thinking about making content, have your AI do the editing. Give it a try.
P.S. I have a bunch of content about this on my YouTube channel and everywhere else under my name, Payton Bilodeau.
Oh shittttt! Robert Pattinson is the next James Bond! Solid pick. He’s a great actor. He’s been on a role with back to back amazing performances in huge movies lately. Excited to see how he does.
Today in AI: November 16
Qwen's New Hire Starts Work:
Alibaba has released Qwen4-Max, its new AI model for coding, office work and tasks that involve images or video.
The digital coworker has arrived with a very long résumé and a plan to cut the bill.
It can hold roughly a million tokens, the chunks of text AI processes, in one working context.
That gives it more room for large files, so you may spend less time splitting up the work.
Alibaba says it beat its previous model on the company's work-task tests while cutting average cost per task by about 60%.
Those are Alibaba's tests, so treat the savings as a result to check on your own work.
The test firm Artificial Analysis rates it above average for its price range, with plenty of rivals still ahead overall.
You can use it through Qwen's apps and developer service now.
Smaller Qwen4 models with downloadable core files are promised later; today's Max release stays hosted.
Apple Studies The Health Chat:
Apple and OpenEvidence have launched a US study of how people use AI for health questions.
Up to 200,000 adults can ask questions in a study app, get answers linked to medical research and rate the responses.
Researchers want to know what people understand, trust and do afterward.
An answer can sound reassuring and still need checking, which is a poor quality for a substitute doctor.
Apple says study data won't train AI models, and sharing personal HealthKit data is optional.
You must be 21 or older, with an eligible iPhone running iOS 26 or later and Apple Intelligence turned on.
The app is for research, not diagnosis or treatment, and doesn't replace a clinician.
I get a crazy amount of hate on my AI videos.
I'd say even MORE than the average creator, simply because I make content around AI and teach people how to use it.
So I'm thinking about starting a weekly short-form series where I bust myths, do comebacks and kind of clown the people who clown me.
I think that'd be pretty funny, and the roasting is really the fun part.
The broader goal is to turn people away from the deep cynicism and pessimism they have around AI, which I think is unjustified.
I think most of those people don't even know why they hate it. They haven't thought about it very rationally, they just have a lot of feelings around it.
Pessimism isn't wisdom.
I want to create a deep sense of rational optimism (not blind optimism) around AI: the benefits in healthcare and the economy, the lives it can and will save, the amazing things people will be able to build, and how it expands human skills, ideas and taste.
People forget those positives because the news and a lot of algorithms heavily bias towards the negativity.
Should I do it? Let me know in the comments.