Good Morning! Here's what I have for you in today's newsletter:
Anthropic opens Salesforce in Claude beta with 37 prebuilt sales skills
Google releases Gemini 3.8 Live, voice models that think while talking
Meta launches Meta One subscription bundling AI usage across its apps
Stop Claude Code from ignoring files you've already built
4 new AI tools worth trying today
AI AGENTS
Anthropic opened Salesforce in Claude in beta on all paid Claude plans, the first product from its Claudeforce partnership with Salesforce, following a pilot with GitLab, Siemens, and Legora covering 7,000 sellers.
The plugin carries 37 prebuilt sales skills, including a morning brief, meeting prep, deal health review, pipeline dashboards, forecast drafts, win/loss analysis, and Salesforce hygiene checks.
Claude reads accounts, opportunities, and pipeline under each seller's existing Salesforce permissions, combining them with Slack threads and email to assemble context before a call.
Write-backs such as logging calls, updating opportunities, or creating follow-up tasks require seller approval and route through Salesforce so business rules still apply, with an admin connecting Salesforce once.
If your company runs on Salesforce, the CRM work that eats a seller's morning, digging up account history, reconciling Slack and email, updating records, can happen in the same Claude window where the call notes already live. Skills for service, marketing, and commerce teams are scheduled to follow.
AI MODELS
Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two native speech-to-speech models available through the Gemini API, Google AI Studio, the Gemini app, Search Live, and Google Workspace.
Gemini 3.8 Live is the speed and cost option, handling mid-sentence interruptions, switching across 97 languages on the fly, and reading live camera input to give step-by-step audio guidance.
Extended Thinking reasons and speaks in parallel, running tool calls and multi-step tasks like event planning in the background while narrating progress, and ranks first on Artificial Analysis' Speech-to-Speech leaderboard.
Developer pricing is $0.005 per minute of audio input and $0.018 per minute of output, and Extended Thinking reaches Google AI Pro and Ultra subscribers in Docs, Gmail, and Keep Live.
Voice becomes a way to get work done, not a novelty mode. You can point your camera at a leaking pipe in Search Live and get talked through the fix, or ask Gemini to plan something while it keeps chatting with you. For builders, the per-minute pricing makes voice agents cheap enough to ship.
AI TOOLS
Meta introduced Meta One, a subscription service spanning Facebook, Instagram, WhatsApp, and Meta AI that bundles expanded AI usage with more than 50 premium features for individuals, creators, and businesses.
Meta One Core costs $7.99 a month and includes Instagram Plus, Facebook Plus, and WhatsApp Plus benefits plus extra Meta AI usage credits, while Premium at $19.99 expands AI usage further.
The AI allowance covers image creation and editing, video generation, deeper reasoning in Meta AI, and in-app tools such as Instagram's Restyle editor.
Business plans top out at Expert for $149 and Max for $499 a month, adding team account access without shared passwords, higher WhatsApp broadcast credits, and more Meta Business Agent capacity.
Meta AI stays free, so nothing is taken away. What changes is that heavy image and video generation, longer reasoning, and creator tools across Instagram, Facebook, and WhatsApp sit behind one monthly bill. If you already pay for two Plus tiers, Core is the cheaper way to get all three plus the AI headroom.
HOW TO AI
Most of what you've built is invisible to your own AI. Here's how to find it.
Step 1: Map a folder
Pick the folder your work actually runs on, not a raw export you downloaded once.
Read every single file in [THIS FOLDER] and build me a map of it. Do not skip any. For each one, work out what it is about and which other files cover the same ground. Then write MAP.md in this folder with four parts: every topic and the files covering it, how many files nothing points at, the connections I would not have spotted myself, and a header with the file count and date. Mark every connection FOUND or GUESSED. Never present a guess as a find.
Step 2: Read the map
Show me Part One and Part Two of the map you just wrote, and tell me honestly if the top-referenced topic actually matches what you'd expectPart Three is where it gets specific, every connection marked FOUND or GUESSED, the difference between a fact and a hunch.
Show me Part Three, and separate the FOUND connections from the GUESSED ones into two lists
Step 3: Fix it
Read MAP.md. Take everything in PART THREE. For each one: what breaks if I leave it, and the smallest change that fixes it. Then do the ones I approve. Ask before you touch anything you are not sure about.Step 4: Make it stick
Add this line to the top of the CLAUDE.md in each of my folders: READ MAP.md IN THIS FOLDER BEFORE ANY JOB HERE, AND APPEND WHAT YOU LEARNED WHEN YOU FINISH. If a folder hasn't got a CLAUDE.md, make one with that line in it.Step 5: Share it
Push these folders to a new private GitHub repoInvite the people you actually work with. From there, they clone it once, and git pull is the only command they ever need to type again.
P.S. You can access all the AI trainings (including the full version of this one), prompts and workflows if you upgrade.
OpenArt launched OpenArt Arena, a global leaderboard that has creative professionals blindly evaluate leading AI models on the work creators care about, from film and ads to animation, motion design, and video editing.
Vercel made v0 model-agnostic, so builders can pick Claude, GPT, Kimi, GLM, Grok, DeepSeek, and more through Vercel's AI Gateway, mixing frontier, cheap, open, and fast models inside one app project.
Obsidian shipped 1.14.2 in early access, letting users open and edit any Markdown file on disk without moving it into a vault, with support for every file format Obsidian handles.
📌 Krea: added Pinterest to Krea Agent, so pins and boards can be brought in as context and generations follow a saved mood board.
🎤 LM Studio: added local realtime voice transcription on Mac, Windows, and Linux, so agents take spoken input without sending audio to the cloud.
✏️ Imagine: can edit the text on any image in beta, swapping copy on event invites, posters, and ads without regenerating the visual.
📋 Notion: added workspace-wide default agent instructions, so house rules apply to every agent without repeating them per page.
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