Mon, April 20 at 6:00 AM
OpenAI just fired back at Mythos with a cyber-only GPT 🛡️
.....wait, is this just a reaction to Anthropic?

Basically, yeah. Days after Anthropic's Mythos spooked the banking world, OpenAI quietly rolled out GPT-5.4-Cyber, a variant of its flagship model tuned specifically for defensive security work.

What's your favorite family of AI models?

Login or Subscribe to participate

OpenAI limits access to new model as firms warn of AI-driven cyberattacks
youtu.be
🤔 so who actually gets their hands on it?

Not you. OpenAI is expanding its Trusted Access for Cyber program to thousands of vetted defenders and hundreds of teams protecting critical software. The company also said its Codex Security agent has already contributed to fixing over 10,500 critical vulnerabilities.

and what's stopping bad actors from using it?

That's the nightmare scenario, and OpenAI knows it. A model fine-tuned to find bugs for defenders can be reverse-engineered to find zero-days for attackers. That's why the rollout is so slow: they want the safety rails to grow alongside who gets access.

Claude Code got a glow up and can now run jobs while you sleep 💤
what actually changed?

Anthropic fully redesigned the Claude Code desktop app on April 14. You can now run multiple Claude sessions side by side, with a new sidebar, integrated terminal, in-app file editor, and faster diff viewer.

Claude Code for Desktop is the BEST vibe coding tool ever (full tutorial)
youtu.be
and the whole "works while you sleep" thing?

That's Routines, which dropped the same day in research preview. A routine is a saved prompt, repo, and connector bundle that runs on Anthropic's cloud infrastructure, not your laptop.

In theory, this should work even better with the latest release of Opus 4.7. However, there are a lot of mixed reviews about this model so far, so we’ll see if it actually gives Claude Code Desktop more capabilities or weighs it down.

weekly scoop 🍦

🇪🇺 EU regulators open talks with Anthropic over Mythos risks

💊 AI is helping researchers spot patient-reported outcomes in cancer trials

👟 Allbirds sells its shoes, pivots to AI, stock jumps 800%

Shoemaker Allbirds Suddenly Says It’s An AI Company
youtu.be

🩺 More Americans turning to AI for medical advice

🗳️ AI assistant texted 1.1M Wisconsin voters before the Supreme Court race

🛒 AI-driven traffic to US retailers jumped 393% in Q1

🗣️ weekly challenge: let two AIs argue about your life decision
what's the challenge?

You know that decision you've been avoiding? The job offer, the move, the big purchase, the "should I quit." This week, you're going to make two AIs debate it out loud while you listen.

Here's what to do:

📝 Step 1: Brain dump Open a doc and dump everything about the decision. Pros, cons, gut feelings, deal breakers, salary numbers. Five minutes of mess is enough.

📤 Step 2: Fire up NotebookLM Head to notebooklm.google.com (free) and create a new notebook. Upload your doc plus any articles, offer letters, or Reddit threads you've been doomscrolling.

⚔️ Step 3: Generate a Debate In the Studio panel, click Audio Overview, hit the pencil icon, and select Debate format. Two AI hosts will argue opposing sides using your sources.

🎧 Step 4: Listen on a walk Put it in your ears while doing dishes or walking. The point isn't a verdict, it's noticing which side you silently root for.

✍️ Step 5: Decide Write one paragraph on what shifted and what you'll do next. Most people find they already knew the answer.

Will OpenAI's gated rollout hold up once jail breakers start pressing? And, is agentic coding at 2 AM the future, or just a faster way to break production?

Zoe from Overclocked 

Keep Reading