Good Morning! Here's what I have for you in today's newsletter:
OpenAI adds Locked Chats to keep sensitive ChatGPT conversations private
Perplexity Computer now runs fully local on your Mac
ElevenLabs launches Eleven Realtime for voice agents
Turn your brand guidelines into a reusable Claude design system
4 new AI tools worth trying today
AI TOOLS
OpenAI rolled out Locked Chats in ChatGPT, letting you save sensitive conversations behind a lock and keep them out of regular history and search. It’s rolling out across Free, Go, Plus, Pro, Business, Enterprise, and Edu plans.
Unlock conversations with a PIN on web and desktop. Supported phones and tablets also offer Face ID, Touch ID, or your device unlock.
New messages won’t update the memory used in regular chats, keeping those details from carrying over into everyday conversations.
Unlike Temporary Chat, which doesn’t save the conversation in your history, Locked Chats lets you come back later. Unlock it and pick up where you left off.
A resignation draft probably isn’t what you want someone finding while borrowing your laptop. Locked Chats gives conversations like that a place to stay saved but out of sight, without feeding new details into your regular chats. On web and desktop, set up your PIN before putting anything behind the lock.
AI AGENTS
Perplexity updated Computer so both the model and the agent can run entirely on your Mac. The local option is available now in the Perplexity Mac app.
Computer can work with native Mac apps and local files while keeping sensitive data off Perplexity’s cloud.
The earlier Hybrid Compute setup divided work between cloud and local models, reserving private-file tasks for your Mac. Now the whole job can run there.
Perplexity recommends a Mac mini as an always-on home for Personal Computer, letting agents keep working while you start tasks from your iPhone.
An agent is more useful when it can work with your files, but uploading client contracts or bank statements can be a sticking point. Local Computer lets it use those documents on your Mac. You also avoid figuring out which parts of a task can stay local, because the entire job can run there.
AI MODELS
ElevenLabs launched Eleven Realtime, a family of speech-to-speech models for voice agents with two options: v1 Mini and v1 Max. Both are available through ElevenAgents and the Realtime API, supporting more than 90 languages and 17,000+ library voices.
Max is built for demanding calls, including customer support and lead qualification. It scored 98.8% on Big Bench Audio and starts speaking in 0.57 seconds.
Mini offers lower-cost conversation, scoring 75.9% on the same test. For tool-use tasks, Max completed 87.1%, compared with Mini’s 60.3%.
The API follows OpenAI’s Realtime API specs, allowing compatible apps to switch endpoints and model names. ElevenAgents supports deployment to phone lines, websites, WhatsApp, and contact centers.
Nobody enjoys a long silence after asking a simple question. Mini offers a cheaper option for straightforward conversations, while Max is aimed at calls that require reasoning or actions through tools. Teams already using OpenAI’s Realtime API can try ElevenLabs with their existing code before committing to a bigger rebuild.
HOW TO AI
Save your colors, fonts, and layouts once instead of explaining them in every chat.
A design system gives Claude a reusable reference for your brand: build one from your existing files, check the details, then select it for new pages and decks. You'll need an eligible Claude plan and a few approved brand references.
Step 1: Check your access
Look under Settings, then Design systems. Anthropic's help center lists Pro, Max, Team, and Enterprise support, with organization controls on managed accounts.
Use the chat workflow. Standalone Claude Design is scheduled to close on December 14, 2026.
Step 2: Gather your references
Collect your logo, brand guidelines, font details, and one finished page or deck. The finished example shows how your spacing, headings, buttons, and cards work together.
Use approved files without confidential content.
Step 3: Build the system
Attach the references in a new chat and ask Claude to create “[Brand name] Design System.”
Request exact colors, fonts, and component styles, with inferred details listed separately. Review the artifact it creates.
Step 4: Check the details
Compare hex codes, typography, components, and spacing with your references. Give specific corrections rather than asking it to “make it more on-brand.”
If the overall style is off, add another finished example and explain what Claude should take from it.
Step 5: Test a design
Start a new chat, select your system, and request a landing page with a hero, three feature cards, and a sign-up form.
Check the result against the system. Correct a wrong system rule there; fix a page-specific mistake on the page. On Team plans, an owner can set an available organization-wide system as the default.
P.S. You can access all my AI courses, trainings, prompts library, and $500+ worth of AI tool discounts here.
OpenAI added team tasks to ChatGPT for recurring engineering updates, such as weekly reports on what shipped and what’s stuck. Teammates can review the results and adjust the next update’s scope, sharing the upkeep instead of leaving it to one person.
Meta brought Muse Wonder to Cline, Cloudflare, Cognition, Cursor, OpenRouter, Oracle, and Vercel. Pricing is $4 per million input tokens and $20 per million output tokens. The company says more partners will follow.
Firecrawl introduced free People Data in Alexandria, opening more than 270 million company and people records to ChatGPT and Claude agents. The Firecrawl plugin includes 1,000 free monthly credits for building prospect lists, finding work emails, and researching employment history.
🎙️ Mistral: Released Voxtral 3 Transcribe and Speech, the models behind its new API for building voice agents.
🔗 Reason Network: Connects Claude Code, Codex, Grok Bot, Muse, and other agents through a free, end-to-end encrypted peer-to-peer network with an open-source engine.
🎨 Krea: Released Krea 3, which completes most 2K image edits in under eight seconds for $0.06 per image.
🪞 HeyGen: Professional Voice Clone learns your voice from 30 minutes to two hours of recordings. Available
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