
Doing AI tutorial videos early this year required explaining a dozen different steps, configuring variables, patching together connections, actually opening and editing files, etc. Now pretty much everything boils down to: Ask your agent to do it 😂

God first. ✝️ Leveraging AI in businesses 🧠 Partnerships/Brand ✉️: cristian@ccarchitecture.ai

Doing AI tutorial videos early this year required explaining a dozen different steps, configuring variables, patching together connections, actually opening and editing files, etc. Now pretty much everything boils down to: Ask your agent to do it 😂

Looks familiar @theo @t3dotcodes #nightly and #stable
You can now opt into more frequent builds of Claude Desktop on Windows and Mac. Switch to the Latest release channel in Settings > Desktop app. You'll get fixes and improvements faster, but these builds can be less reliable than Stable, which stays the default.

GPT-6.2 is out as two models. Astra is for the hardest projects, Sol for everyday work. OpenAI's chart has Astra at 78.3% for $2.89 a task. GPT-6 Astra: 73.1% at $8.99. Early testers are split, and it is not in regular Chat. On a paid plan, open Work.

You can now put your own AI sales rep inside ChatGPT. OpenAI's Business Agents are for US online stores and online services. OpenAI's policy says the agent has to say it's AI and can't invent prices or reviews. Test it with your top 5 buying questions. #chatgpt

ChatGPT now shows businesses how often it mentions them. OpenAI launched Businesses in ChatGPT in beta on Nov 10. Claim and verify your business, and once you pass a minimum you see mentions and website clicks from the past 28 days. #chatgpt

Claude builds real spreadsheets now. Claude Sheets launched in beta on November 10 on every paid plan: tabs, formulas, formatting, drop-downs, export to Excel or CSV. Claude Whiteboards came out the same day and turns a plan you talk through into a timeline. #claude

The best Claude Code upgrade I found this year was an app, T3 Code. Claude Code and Codex in one window, one thread per project, and from my iPhone I can send instructions that run on my main computer back at the office. Hats off to @theo and @t3dotcodes. #claudecode

Claude can edit your videos, but it can't hear them. It only sees a sound file's name and length, so every sound effect is a guess. Earshot by @lexnlin is a free, open source tool that lets Claude Code listen. It checked all 16 sounds on one of my shorts. #claudecode

Every new Claude Code session starts from zero. @mattpocockuk built /grill-with-docs for that. It interviews you, writes the terms you agree on into GLOSSARY.md and hard calls into decision records in docs/adr/. Next session the AI reads them. #claudecode

Someone asked if Obsidian beats context-mode for Claude. They fix two different problems. Context-mode keeps huge reads out of Claude's context in one session. Obsidian holds what Claude should remember across sessions. I use both. #claudecode

Agreed… just added like 10 new ones , currently have on “Nyx” from @Deepusleepy
If you are not T3 code pilled yet, I promise you you will be now!

ARETE 📈
Sat down with @0xLinnyx0 to chat about how @arete_a4 lets your agents use @solana safely. Shout out @superteam @SuperteamBLKN for the opportunity


Alex Finn's (@AlexFinn) setup: plan with Fable 5.2, build with Opus 5.5. The step people skip is writing the plan down. Anthropic's model guide says switching from Fable to another model drops its reasoning, so Opus only sees what's written. #claudecode

Stop typing "go" when Claude asks for permission. Anthropic's Fable 5.2 guide says to spell out what you want. Their example: "Yes, run the migration on the staging database" where you'd have sent "go" or a thumbs-up. #claudecode

Old "never" rules in your CLAUDE.md can make Claude worse. Anthropic's Fable 5.2 guide says it treats any instruction with "never" in it as absolute. Rewrite "never delete files" as "ask for approval before deleting files" so it checks with you and keeps working. #claudecode