Good Morning! Here's what I have for you in today's newsletter:
xAI releases Grok Voice 3.0, resolving 70% of Starlink support calls on its own
Anthropic upgrades the Claude Code desktop app for running sessions
Lovable adds ad launching and marketing assets to its app builder
How to cut 90% of your tokens with Jev and Claude Code
4 new AI tools worth trying today
AI MODELS
xAI released Grok Voice 3.0, a speech-to-speech model for voice agents that responds in under a second, with 40% lower latency on voice agent tasks.
At SpaceX, the model resolves 70% of Starlink support calls without handing the caller over to a human.
xAI says it matches leading voice models on agent tasks at a lower cost. Its Voice Agent API page lists a price of $0.05 per minute.
The API supports talking and listening at the same time, using tools, searching for information, and holding conversations in more than 25 languages.
Answering a call quickly is one thing. Solving the caller’s problem is what counts. If other businesses see results like Starlink’s, voice agents could take on more routine support and shorten the queue for human help. Resolution rates and the cost of each successful call will tell the real story.
TOGETHER WITH WISPR FLOW
Leave Granola and get up to 12 months free of Wispr Flow Notetaker + Dictation
If you have paid time left on an individual Granola plan, we'll match it with a Wispr Flow subscription that includes Notetaker and dictation, and add bonus time, up to 12 months total. Sign in or create a Wispr account and submit proof of your plan to check eligibility.
AI TOOLS
Anthropic updated the Claude Code desktop app to make multiple sessions easier to start, organize, and follow, so developers can split work across tasks without losing track.
Recent folders appear on the new session screen, putting projects you’ve been working on one click away.
Save a setup with several folders, an SSH host, or the cloud as an environment, then reuse it for new sessions. Each folder can have its own worktree.
Running sessions show their current activity, so you can see what each task is doing without checking every window.
One session can fix a bug while another writes tests, but keeping tabs on both can become a job of its own. Saved environments cut the repeated setup, and visible activity makes progress easier to follow. Anyone already juggling Claude Code across terminal tabs may appreciate having it together.
AI AGENTS
Lovable added marketing tools to its app builder, letting users turn their projects into images, videos, and ad copy, then launch campaigns on Google, YouTube, Facebook, and Instagram.
Launch ads directly from Lovable, keeping campaign setup in the same place where you built the product.
Generate visuals and copy using the app or site you’ve already created, giving the marketing tools product details to work from.
The update adds promotion to Lovable’s building tools, helping founders and small teams tackle the work that comes after an app goes live.
Finishing an app leaves another big task: getting people to try it. Lovable now lets founders create promotional material and put ads live without moving to a separate tool for each step. For a small team handling both product and marketing, that could mean less setup between launch and the first campaign.
HOW TO AI
Let Jev sort and shortlist your notes so Claude reads only what it needs.
Jev returns decisions and probabilities rather than written answers. Here, it organizes a folder of Markdown notes and selects three relevant files for Claude to read.
Step 1: Install Jev
Create a TypeSafe account and API key. Set the key in Terminal and keep it out of chat.
Install the official skill with claude plugin marketplace add typesafe-ai/skills and claude plugin install typesafe@typesafe-ai. Have Claude confirm the expected key variable without displaying its value.
Step 2: Build the brain
Create ~/second-brain with inbox, notes, people, and projects folders.
Ask Claude for a CLAUDE.md under 60 lines covering the layout, filenames, source citations, and what to do when an answer is missing.
Step 3: Sort incoming notes
Add six fictional notes to inbox. Have Claude build file-inbox, using Jev to choose notes, people, projects, or none.
Review the proposed threshold and moves before running it. Leave uncertain notes in the inbox.
Step 4: Find relevant notes
Build find-notes to rank files against a question and return three paths.
Add a CLAUDE.md rule to run it before answering questions about the collection. Claude should read the selected files and cite the evidence.
P.S. You can access all my AI courses, trainings, prompts library, and $500+ worth of AI tool discounts here.
Perplexity added Ultrafast versions of GPT-6 Astra and GPT-6.1 Sol to Computer for Max subscribers. Choose Custom and set the speed to Ultrafast for faster token generation that can shorten long tasks.
Meta released Immersive Web SDK 1.0, allowing coding agents to run and test web XR apps in a desktop emulator or on a connected Meta Quest, catching performance problems before users encounter them.
Hermes gained a third-party Phone Harness plugin for operating iPhones and Android devices. It can read screens, open apps, tap, type, and complete multi-step tasks through Mac mirroring, ADB, or cloud phones.
🏗️ Factory: Launched OpenCompute to manage sandboxes for remote agents inside your own VPC, giving teams a way to run the fleet on infrastructure they control.
🎨 Tesseract: Added color grading with Opus 5.5 and GPT-6 Astra, including ACES and LUT support. Layers remain editable in Premiere or After Effects.
🗂️ Hark Pro: Prepares approvals, receipts, and flight check-ins before you ask. The proactive assistant is free on web, iOS, and Android.
🎬 HeyGen: Released HeyGen Video, a business video model built on MiniMax H3 with additional training by HeyGen. Pricing starts at $0.01 per second through October.
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