Rig Labs

Operating at the bleeding edge of AI

By @dtsbourg and @hugorenaudin_

[Rig Labs @riglabs_]

The Edge, issue 5, is out.

→ Claude joins your Google Meet and starts the plan doc before you hang up
→ Gemini 4 Hydrogen, free in Google Search
→ One GitHub task: 1,421 tokens by command line, 59,417 by MCP

5 minutes, free: https://t.co/YgYrwcW8ob

[Rig Labs @riglabs_]

Blog: Gemini 4 Hydrogen, Google's new fast model, is free in Google Search and on the API free tier.

Google says it finishes tasks 2.5x faster than 3.8 Flash. The catch: on the paid API it costs twice as much.

Prices, where to use it, when to switch: https://t.co/kyATmdtICP

[Rig Labs @riglabs_]

The full guide: https://t.co/h0isyqpuCk

Give the setup file to your AI assistant and it does this for you.

[Rig Labs @riglabs_]

Guide: let Claude join your Google Meet calls, take the notes and build the plan doc while you talk.

It can stay up to 8 hours, and an owner decides if it may join calls with outside guests.

Setup, the 3 settings that matter, limits and a ready-made file for your assistant:

[Rig Labs @riglabs_]

Limits: Google Meet only for now. Not for orgs with zero data retention or HIPAA setups. It saves a transcript, not audio. Docs made in a meeting are free until Feb 10, 2027. Connected apps count toward your own usage.

[Rig Labs @riglabs_]

The Edge, issue 4, is out.

→ Make Notion the memory of Claude and ChatGPT: 1 guide page every assistant reads first, and what it can then see
→ United says no to agents that book flights: the rule to give yours
→ Gemini Decision 1

5 minutes, free: https://t.co/TndCGIRNCn

[Rig Labs @riglabs_]

Blog: check if Claude Code saved one of your passwords in a file that went to GitHub.

"Yes, and don't ask again" can store the whole command, password included. Apiiro found 1,122 public copies with secrets.

The git command that tells you, and the fix: https://t.co/9VI4ywBgry

[Rig Labs @riglabs_]

Blog: United Airlines says it will prohibit personal AI agents from booking its flights.

If your agent books flights for you, United is now off its list.

3 reasons United gave, 3 unknowns, and the rule to paste into your agent today: https://t.co/3xuh6c75k9

[Rig Labs @riglabs_]

Blog: make Notion the shared memory of Claude and ChatGPT, so you explain yourself once.

Notion's setup prompt adds 1 guide page every agent reads first.

The catch: the agent can read all your account can read.

What it creates, and the first step: https://t.co/dlYOob37m7

[Rig Labs @riglabs_]

Blog: Google released Gemini Decision 1, a model that only answers yes/no, pick-one and score questions.

$0.10 per million input tokens, output free.

For the small checks your app sends to a big model all day.

Price, limits, first step: https://t.co/VU0W13Z0qF

[Rig Labs @riglabs_]

The Edge, issue 3, is out.

→ Claude Sheets: a spreadsheet with tabs and formulas from one sentence, and why small trackers will start in a chat
→ Gemini 4 Carbon at $2/$10 until November 30, then double
→ 3 guides with setup files

5 minutes, free: https://t.co/JFdTfUP0AR

[Rig Labs @riglabs_]

Blog: if your agent reads files, git repos or web pages through the MCP reference servers, update them today.

v2.4.0 fixes 6 vulnerabilities. One let a repo you only looked at run a program.

The 3 fixed versions, and why one will not update by itself: https://t.co/4kMEyZo8xR

[Rig Labs @riglabs_]

Blog: Google's top model, Gemini 4 Carbon, is now open to every paying developer and to Ultra subscribers.

Until November 30 it costs $2/$10 per million tokens. Then it doubles to $4/$20.

Where to use it, when Argon is enough, and one test to run: https://t.co/jybjYvadVX

[Rig Labs @riglabs_]

Blog: you can now ask Claude for a real spreadsheet, with tabs, formulas and drop-downs, and edit it with your team.

Claude Sheets and Whiteboards: beta, on all paid plans. Claude Data: out of beta.

What each of the 3 does, the limits, a first prompt: https://t.co/OXLG7kIXWS

[Rig Labs @riglabs_]

Blog: every Claude Pro, Max, Ultra and Team user now has 1 free full reset of their usage limits, across all models.

Use it any time before December 10, 2:00 PM PT.

It is in Settings > Usage. Save it for the day you hit your weekly limit: https://t.co/LDFme5MLVz

[Rig Labs @riglabs_]

Each item with its limits and source: https://t.co/PFeBoA77kH

[Rig Labs @riglabs_]

New this week. Perplexity Teammates: always-on agents that own a job and work in Slack, on the Max plan. Personal Finance in Claude: read-only access to your US bank accounts through Plaid. Cursor: a cloud agent builds an app, hosts it and shares it by link.

[Rig Labs @riglabs_]

Claude Fable 5.2, Anthropic's top model, now costs $8/$40 per million tokens on the API, down from $10/$50.

It is the default Fable model in Claude Code, and is in Cursor and Perplexity Computer.

When to pick it over Opus 5.5, and how to prompt it: https://t.co/Lp909g42Hi

[Rig Labs @riglabs_]

Glasshand rents AI agents real iPhones and Mac minis in a data center: $60 and $90 a month. The agent sees the screen and taps and types through an API.

What it does, what it costs, and what to know before signing in to real accounts:
https://t.co/AVfQ3wWJ4Z