In partnership with
Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},
OpenAI helped track down two serious holes in PHP before attackers got to them, and the tricky part turns out to be getting everyone to patch in time.
Anthropic is also testing an agent that handles things like emails and shopping without being asked, and a free coding model that runs on your laptop was mostly built by AI itself.
We’ve got a few handy tools and a Chinese landscape painted entirely with code too.
Let’s dig in!
🤿 DEEP DIVE
OpenAI Found a Hole in the Internet
Finding vulnerabilities is getting faster. Fixing them is another story.
AI is getting pretty good at finding security holes. But there's a problem nobody can really ignore. What happens when it starts finding them faster than people can fix them?
OpenAI just gave us a pretty interesting example.
Over the past two months, OpenAI helped coordinate a security response around two critical vulnerabilities in PHP, the programming language used to run websites and web applications.
One flaw was identified by Calif using GPT-5.6 Sol through OpenAI's Patch the Planet initiative. Security company Wiz independently discovered and reported both vulnerabilities.
And according to OpenAI, frontier AI found these flaws before attackers did.
That's a pretty important detail.
Finding the bug was only half the job
The PHP maintainers led the actual fixes, but getting those fixes ready for the wider internet involved a lot more people.
OpenAI worked with researchers, security providers like Wiz and Cloudflare, infrastructure companies, and government agencies to help prepare protections before the vulnerabilities became public.
Because fixing a security flaw in widely used software isn't just about changing a few lines of code.
Some organizations need to update older versions. Hosting providers might need temporary protections. Security companies need time to prepare defenses. And everyone needs enough information to act without accidentally revealing the vulnerability too early.
OpenAI calls this opportunity the Defender's Window.
Basically, if AI finds a serious flaw before attackers do, defenders get a head start. But that advantage only matters if they can actually use the time.
So they're trying to make the response repeatable
OpenAI is proposing a five-stage framework for handling major vulnerabilities, from confirming and fixing the bug to notifying affected organizations, coordinating with governments, preparing updates, and helping people deploy them.
The idea is to avoid figuring out the whole coordination process from scratch every time AI discovers something serious.
And there's a tricky balance here.
Tell too many organizations too early, and sensitive details might leak. Tell too few, and some providers won't have enough time to protect their customers.
OpenAI's proposal leaves those decisions with maintainers, using criteria like who can actually provide protection and how much time they need.
There's a bigger AI problem underneath this
OpenAI says these PHP vulnerabilities are among several significant flaws it has already found, and it expects discoveries to increase sharply.
That's both promising and a little uncomfortable.
The same AI capabilities helping researchers find vulnerabilities can also help attackers develop exploits.
So OpenAI is putting $1 billion toward subsidized Daybreak access for frontline defenders with limited security resources, while also developing safeguards to reduce misuse.
It's also sharing a PHP version checker as a Codex skill to help organizations identify potentially vulnerable installations.
And I think that's the part worth remembering. Finding security flaws with AI is becoming one challenge. Building a system that can actually respond to all those discoveries is another.
The quick take
OpenAI says frontier AI discovered critical PHP vulnerabilities before attackers did.
PHP maintainers led the fixes, supported by researchers, security providers, and government coordination.
Five stages aim to help defenders prepare and deploy protections more consistently.
OpenAI says AI-driven vulnerability findings could increase sharply.
OpenAI has committed $1 billion in subsidized Daybreak access for organizations protecting essential services.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Discover laptops with built-in intelligence.
Shift your workforce into hyper-productivity mode with Dell Pro laptops powered by Intel® Core™ Ultra with Intel vPro®. Shop on Dell Premier.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
An AI Draws and Animates an Entire Chinese Landscape Using Code
Creator Chetaslua asked Claude Fable 5.5 to draw and animate a traditional Chinese landscape using code. Mountains, blossoms, clouds, and a tea-drinking woman gradually come together in a 44-second artwork that feels like watching a painting create itself.
AI used
Claude Fable 5.5
⚡ QUICK HITS
The latest in AI, tech, and productivity worth knowing
Anthropic has introduced Claude Duet, a proactive AI agent that handles tasks like reading emails and shopping without waiting for instructions. It's still in a research preview with a waitlist, and will eventually be available to paid subscribers.
Recursive has released Flint, a free coding model developed in under four months with help from its own AI research system, the Eureka Machine. The system helped with nearly every stage of development, including experiments and training improvements. Flint runs locally on a laptop, so developers can work offline and keep their code private.
👀 ON OUR RADAR
One more thing we thought you'd like
A fast AI answer isn't always a coordinated one. When AI resolves B2B issues without looping in the teams responsible for them, confidence outpaces alignment. A new Harvard Business Review Analytic Services briefing paper, sponsored by Front, looks at closing that gap.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
Turn your ideas into physical designs that are ready to make.
Why it’s useful: Handy for creating things like stickers, packaging, and pop-up cards without needing to figure out print dimensions, cut lines, or design software yourself.
🎯 Needle
Find potential customers and reach out using AI prompts.
Why it’s useful: Useful for founders and sales teams who want to find leads across LinkedIn, Instagram, and email without jumping between different prospecting tools.
Give your business data a shared source of truth.
Why it’s useful: Helps teams get more consistent answers to business questions by keeping company definitions, context, and previous analysis in one place.
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
