• 80/20 AI
  • Posts
  • Google's Gemini 4 Carbon Goes Public

Google's Gemini 4 Carbon Goes Public

PLUS: Claude joins your meetings + turn any call into a to-do list

November 11, 2026

Sponsored by

Forwarded this email? Get 80/20 AI free and learn the 20% of AI that matters, in 5 minutes a day.

Good morning, there. Google just made Gemini 4 Carbon, its most powerful model, available to paying developers and AI Ultra subscribers, at an intro price that ends on Nov 30. Meanwhile, Anthropic wants Claude to sit in on your meetings and build your spreadsheets. Here's what's worth your time.

IN TODAY'S 80/20

  • 🧠 Gemini 4 Carbon goes public — what it costs and when to use it

  • 📅 Claude joins your meetings — plus Sheets and Whiteboards

  • 🎓 Skill of the day — turn any meeting transcript into owners and deadlines

  • ⚡ Quick hits + trending AI tools

Read time: 5 minutes

Find where your agent breaks, before your users do

Your agent works in the demo. Production is a different story. Plurai is the simulation infrastructure for self-improving agents. It stress-tests your agent across realistic scenarios and long-tail edge cases, pinpoints exactly where it breaks, and suggests the fix you apply in one click. Then it post-trains small open-source models for fast, low-cost evals and guardrails in production. Works with any model and any agent platform, right inside Claude Code, Codex and GitHub Copilot. Teams using Plurai have cut failure rates by 82% and eval costs by 97%.

GOOGLE

🧠 Gemini 4 Carbon is now open to paying users

Gemini 4 Carbon key art

Image source: Google

The 80/20: Google opened up Gemini 4 Carbon on Nov 10, two weeks after a release limited to cybersecurity teams. It's built for the hardest jobs: long agent workflows, tough debugging, security work and scientific research.

The details:

  • Where to get it: Google AI Studio and Gemini Enterprise for paying developers and businesses, and rolling out to AI Ultra subscribers in the Gemini app, AI Mode in Search and Antigravity.

  • Price: $2 per million input tokens and $10 per million output tokens until Nov 30, then $4 and $20.

  • Benchmarks (Google's numbers): #1 on CursorBench 4.0, Zapier's AutomationBench and Perplexity's WANDR, at more than 2x lower cost per task than the next-best outside model. On an honesty test it passed 92% of tasks, vs. 70% for Gemini 4 Argon.

  • New /swarm mode in Antigravity (preview) turns one prompt into a team of agents: Carbon plans and checks the work while faster models do tasks in parallel.

  • Not for everything: Google says Argon is still the better pick for most coding and knowledge work, because Carbon uses more tokens to explore and verify.

Why it matters: The top models are getting cheaper and more specialised. Use Argon for everyday work, and save Carbon for the one hard, multi-step task you've given up on. If you want to test it on real workloads, do it before the price doubles on Dec 1.

🎓 AI SKILL OF THE DAY

Turn any meeting into owners, deadlines and a follow-up email

The 80/20: You don't need a meeting bot to get great notes. A transcript and one prompt gets you a clean action list in under 2 minutes, in Claude, ChatGPT or Gemini.

Step-by-step:

  1. Download the transcript from Google Meet, Zoom or Teams (or paste your rough notes).

  2. Paste it into your AI chat with the prompt below.

  3. Copy the table into your team chat or tracker, and send the drafted follow-up email.

Copy and paste this prompt:

Here is a meeting transcript. Do three things:

1. List every decision we made, in one line each.

2. Make a table of action items with columns: Task | Owner | Deadline | Source quote. If an owner or deadline wasn't stated, write "MISSING" instead of guessing.

3. Draft a short follow-up email to attendees that recaps the decisions and asks each owner to confirm their tasks and dates.

Transcript:
[paste here]

Pro tip: The "MISSING" rule is the secret. It shows you every task nobody actually owns, which is where most projects quietly stall.

ANTHROPIC

📅 Claude now joins your meetings and builds your spreadsheets

The 80/20: Anthropic launched Claude Meetings, Sheets and Whiteboards in beta on Nov 10, pushing Claude further into Google and Microsoft's office territory.

The details:

  • Claude Meetings (Team and Enterprise) joins Google Meet calls as a visible participant, then leaves notes and action items in claude.ai. It won't join calls with outside guests unless an admin allows it.

  • Claude Sheets (all paid plans) builds spreadsheets with tabs, formulas and dropdowns, and exports to Excel. Anthropic calls it "fairly basic" for now, so it's not for heavy financial models.

  • Claude Whiteboards (all paid plans) is a planning board that Claude can turn into a doc or slide deck. Claude Data (formerly Dashboards) is now generally available.

Why it matters: Claude is turning from a chat window into a full workspace. Before you invite any AI bot into calls, check your company's recording and data rules, and tell attendees it's there.

⚡ QUICK HITS

  • 🧠 Gemini 4 Carbon — Google's strongest model for long, hard tasks, at half price until Nov 30

  • 🐝 Antigravity /swarm — turn one prompt into a coordinated team of coding agents (preview)

  • 📊 Claude Sheets — build working spreadsheets by describing them, then export to Excel

  • 🗂️ Claude Whiteboards — plan on a shared board, then turn it into a doc or slides

A CRM so smart, it updates itself.

HubSpot's CRM is so smart, it now updates itself. Calls get logged and summarized the moment they end. New leads get researched and a first outreach drafted, automatically. Deals move forward on their own, with next steps flagged before you have to ask. 

Even the one repetitive task you've been meaning to fix can now run itself, no code required. Customers using it are already seeing the difference. 

This isn't the CRM you have to work hard to figure out. It's the one already working before you log in.

💬 ALAMIN'S TAKE

Two launches, one pattern: the AI labs no longer just want to answer your questions, they want to be where the work happens: your meetings, your spreadsheets, your codebase. My advice is to pick one workflow you repeat every week and test it with these tools, instead of trying everything at once. That's the 80/20 of keeping up.

How was today's 80/20 AI?

Login or Subscribe to participate in polls.

Know someone who should read this? Send them this link.

See you tomorrow,
Alamin

ADVERTISE WITH 80/20 AI

Reach 65,000+ engaged AI professionals. Book a placement.