Good Morning! Here's what I have for you in today's newsletter:
OpenAI introduces GPT-6 Cyber to find and fix security vulnerabilities
Meta brings voice conversations to Muse, its personal AI agent
Google releases Gemini Omni 1.2 Flash, 15-second video at 30% lower cost
Stop AI companies from training on your website with Claude
4 new AI tools worth trying today
AI MODELS
OpenAI introduced GPT-6 Cyber to find vulnerabilities, test whether they’re exploitable, and help write fixes. Access starts with selected Daybreak Red customers through the API and Codex, expanding to all Red customers over the coming weeks.
OpenAI reports 48.5% on ExploitGym, compared with GPT-5.6 Cyber’s 32.1%. In its Linux kernel research, the model helped produce working exploits for 948 vulnerabilities, including 588 previously unknown.
In testing, it exceeded authorized scope less than 1% of the time, down from about 5%. OpenAI also raised its Preparedness Framework rating from High to Critical.
Mozilla, Palo Alto Networks, Calif, and Cisco are already using it. OpenAI says it found 81 previously unknown high-severity Firefox vulnerabilities, with Mozilla starting to ship fixes.
Finding a suspicious bug is one job. Proving it matters and getting it patched are two more. GPT-6 Cyber targets all three, with early results from Firefox and Linux research. Its improved scope compliance matters, but “less than 1%” still isn’t zero, and the Critical rating makes the controlled rollout worth noting.
AI TOOLS
Meta introduced Voice in Muse, letting users speak with its personal agent and hear responses in real time. The rollout starts in the US across mobile, web, and Mac.
Start a call from the phone icon at the top right of a chat. You can interrupt an answer, and Muse continues background tasks while you talk.
An animated avatar responds with facial expressions and lip movements matched to the speech. Keep it beside the chat or open it full screen.
Hundreds of voice styles are available, with the option to switch at any time. Meta says support for its AI glasses and upcoming Muse Charm device will follow.
Talking through a trip can be easier than typing every half-formed idea. Voice lets users change direction mid-answer while Muse keeps working in the background. The avatar adds a visual presence, but the practical question is how well it follows a messy, interrupted conversation and carries those instructions into the task.
AI AGENTS
Google released Gemini Omni 1.2 Flash through the Gemini API and Gemini for business. The updated video model generates 720p footage at $0.07 per second, 30% below Omni 1.1 Flash.
The maximum clip length rises from 10 to 15 seconds, with exact durations available between three and 15 seconds. Generate at 360p or 720p, then upscale to 1080p or 4K.
Follow-up prompts can change a background, subject, or lighting without starting from scratch. Google also reports better consistency for characters, objects, and scenes.
The update aims for more natural multi-speaker dialogue, closer lip-sync, and fewer unwanted jump cuts during continuous camera movements.
A shot can be almost right until the lighting changes or the camera suddenly cuts away. Longer clips and follow-up edits give creators more room to fix those problems without regenerating everything. The lower price also reduces the cost of testing variations, though consistency across edits still needs judging on the actual footage.
HOW TO AI
A proposed robots.txt group lets website owners ask AI companies not to train on their pages while preserving search crawler access. Support is rolling out, so keep existing crawler rules too. You’ll need website editing access or help from its maintainer, plus Claude with web search available.
Step 1: Check your file
Open yoursite.com/robots.txt. Ask Claude to retrieve that address, reproduce the file, and check for User-Agent: __usage_ai-training. If retrieval fails, paste the text yourself. Under the proposal, a wildcard User-Agent: * group alone does not express the shared training preference.
Step 2: Pick what stays open
Use Disallow: / for a site-wide preference or add Allow: /press/ to leave that folder open. Consider Gemini access: Google’s planned mapping treats the group like a Google-Extended block, which also affects Gemini app answers. An explicit Google-Extended allowance leaves those uses open.
Step 3: Add the group
Back up your file. Ask Claude to append these lines after a blank line, preserving everything else:
User-Agent: __usage_ai-training Disallow: /
Review the changes before saving. Don’t place Disallow: / under User-Agent: *: that requests a site-wide crawling block, including search crawlers. On WordPress with Yoast, look under Yoast SEO → Tools → File editor.
Step 4: Keep the old blocks
The guide lists Microsoft’s planned start as 1 February 2027, Google’s target as 2027, and OpenAI’s timing as “coming months.” Keep separate Disallow: / groups for GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, and CCBot, adjusting Google-Extended for your Gemini decision. GPTBot is separate from ChatGPT’s search crawler.
Step 5: Check it worked
Reload the live file and repeat the review in a fresh Claude chat. Check the two leading underscores, separate groups, and preserved existing rules. If an older version appears, wait briefly and reload. A correctly published preference remains a request; it does not prove compliance or undo past training.
P.S. You can access all my AI courses, trainings, prompts library, and $500+ worth of AI tool discounts here.
Anthropic introduced Personal Finance in Claude, in beta for US Pro, Max, and Ultra users. Users choose which accounts to connect, and Claude answers questions using balances and transactions. Access is read-only, so it can’t transfer or invest money.
Perplexity introduced Teammates for Max subscribers on the web. These agents use Computer tasks, skills, memory, and tools, each in an isolated environment. They can message coworkers in Slack and respond to changes in ongoing work.
Cursor added live apps on all paid plans. From a description, it builds the app, sets up a database, runs security checks, and deploys it. Finished apps can be shared through a link.
🎥 Krea: Rebuilt its browser-based video tools with camera controls, 100+ presets, UGC templates, editing, style transfer, and sound generation.
✂️ Motion: Added prompt-based editing for Premiere Pro, producing projects with cuts, captions, and keyframes that remain editable.
🔍 Lovable: Added Parallel for web search and deep research inside apps, with cited results and no separate API key to manage.
💻 Devin: Released Review 2.0 with typical PR reviews priced at $0.25 to $0.75, incremental re-reviews, and Autofix in Cloud and Desktop.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
✓ $500+ worth of AI tools