This website uses cookies

Read our Privacy policy and Terms of use for more information.

IN PARTNERSHIP WITH

Cloudways by DigitalOcean

Cloudways Managed AI Agents runs OpenClaw or Hermes on its own private server, live in minutes once you add your API key. And this Black Friday, get 80% off for 2 months across Cloudways, add-ons included, taking the entry plan to about $2 a month. Use code BFCM802 until December 4.

Welcome back. The FBI warned that a free AI hacking tool got into company networks in under three hours. Now Google has found an AI that turns a security patch into a working attack with no human help. Hot take: the slowest part of your security is now you, not the hacker.

Today: AI hackers speed up, who just bought into Anthropic before tonight's IPO pricing, a 7-cent model from Tencent, and a 20-minute lockdown for your AI tools.

TODAY IN AI

1. 📈 Anthropic's IPO update: demand is about six times the shares: quick follow-up to yesterday. The final price is set tonight, and CNBC's sources expect it above the $70 to $78 range. Trading starts tomorrow under ANT. The new detail: AMD, Google, Microsoft, Nvidia, TSMC and others are buying $26 billion of stock privately at the IPO price, so the chip and cloud companies Claude runs on now own a piece of it.

Why it matters: the companies that sell Anthropic its compute are now also its shareholders. Don't build your whole business on one AI provider: keep your key prompts and workflows written so they run on a second model too.

2. 🚨 AI now turns security patches into attacks by itself: Google found an AI setup, linked to a suspected Russian-speaking operator, that reads a fresh security patch and builds a working exploit for it with no human in the loop. Separately, the FBI warned that attackers used Artex AI, a free open-source tool, to break into networks in under three hours, including a US IT company and South Korean financial firms.

Why it matters: the gap between "update available" and "someone is using it against you" is now hours, not weeks. Turn on auto-updates for your website plugins, browser, password manager and anything your AI agents connect to.

3. 🔥 Tencent's new model costs 7 cents per million tokens: Hy4 flash reads text, images and more, holds up to 1M tokens, and Tencent prices input at about $0.07 per million tokens. It's built for fast everyday jobs. Same day, Hugging Face teamed up with OpenRouter so one API key can send each request to the cheapest or fastest provider that's up.

Why it matters: cheap models keep getting cheaper every week. Don't lock any automation to one model. Use a router so you can switch with one line when a cheaper one passes your test.

What this means for you: AI is making attacks faster and models cheaper at the same time. Spend some of what you save on locking things down.

OUR BLACK FRIDAY DEAL

One context profile shared across ChatGPT, Claude, Gemini, Grok and Perplexity

NjectAI saves your business, audience and voice once, then injects that context into your prompts in ChatGPT, Claude, Gemini, Grok and Perplexity, and into every prompt in the God of Prompt library. Black Friday: your first month is $0.90 instead of $9, already applied at checkout, until November 27.

FROM THE FRONTIER

OpenAI just let outside watchdogs sit inside its labs.

The good news. Six Apollo Research evaluators will work inside OpenAI for four months, checking whether models quietly work around their safeguards or hide what they're doing. OpenAI also signed deals with METR, Transluce and Goodfire. Outside eyes on how agents behave is exactly what this year needed.

The bad news. Apollo itself says its access and what it can publish are limited. And this lands the same week California sued OpenAI over 80 incidents where agents went beyond what they were allowed to do. Watchdogs help, but they won't be watching your agent.

My call: no AI agent touches money or sends email without your OK

Let agents draft, sort and prepare. Keep the final click on payments, refunds, outgoing emails and deletions with a human. It costs you a few seconds per action. I'd change my mind once outside testers can publish freely what they find.

IN THE KNOW

𝕏 God of Prompt @godofprompt

You burn your weekly Claude limit by Wednesday. Most of it is the top model doing easy work. On the API, 1,000 tasks cost $60 on one top model and $9.96 when each goes to the cheapest model that can do it…

♥ 11 · ↻ 1 · View on X

𝕏 FBI Dallas @FBIDallas

The #FBI issued a FLASH advisory warning that cybercriminals and state-aligned hackers are exploiting ARTEX, a super intelligence-powered penetration testing framework…

♥ 44.6K · ↻ 15.6K · View on X

𝕏 Watcher.Guru @WatcherGuru

JUST IN: Google and Amazon to own a combined 28% of Anthropic after its IPO.

♥ 4.5K · ↻ 394 · View on X

The AI news without the noise

TLDR AI covers the model releases, research, and tools that actually matter, curated by engineers from OpenAI, Adobe, and ex-Google DeepMind. It’s a 5-minute read, delivered every weekday morning.

TOOLBOX

🎬 Creatify: paste a product link and get a ready video ad with an AI presenter. Make ten versions for the cost of one shoot.

📹 Fliki: turn a blog post or script into a voiced video with stock footage in minutes. Good for repurposing one newsletter into a week of shorts.

📈 Surfer SEO: shows exactly what your page is missing compared to the pages that rank above it, then helps you rewrite it with AI.

TUTORIAL

Lock down your AI tools (20 minutes)

  1. List every AI tool you've connected: check Google, Microsoft and Slack for "third-party apps" and write down each one with what it can access.

  2. Remove what you don't use: anything you haven't opened in 30 days goes.

  3. Run the prompt below with the rest. It flags which access is too wide.

  4. Cut access to the minimum: read-only where the tool only needs to read, one folder instead of your whole drive.

  5. Turn on auto-updates for your browser, website plugins and password manager.

  6. Set a monthly reminder to repeat steps 1 and 2.

Here are the AI tools and apps connected to my accounts:
[paste your list: tool, what it connects to, what access it has]

For each tool:
1. Rate the risk HIGH / MEDIUM / LOW if this tool got hacked tomorrow.
2. Say what the smallest access is that still lets it do its job.
3. Tell me exactly where to click to reduce it (e.g. Google Account > Security > Third-party access).
Then give me a table: tool, current access, risk, change to make.
Flag anything that can send email, move money or delete files first.

PROMPT STATION

Suspicious Email Check

When an email asks you to pay, log in or click, and something feels slightly off.

## Role
You are a security analyst who checks emails for small businesses.
## Task
Tell me if this email is safe to act on.
## Context
The email: [paste full text, sender address and any link addresses]
What it asks me to do: [pay / log in / download / reply with info]
Who I normally deal with at this company: [name, usual email address, or "none"]
## Output
1. Verdict: SAFE / SUSPICIOUS / SCAM, in one line.
2. Every red flag you found, with the exact words or address that shows it.
3. How to check it without using anything in the email (e.g. call the number on their website).
Never tell me to click a link in the email to verify it.

👉 Browse more prompts like this. Built for ChatGPT, Claude, Gemini, Grok.

Want one written for your business instead? Describe the task in a sentence and the prompt generator writes it, now UNLIMITED with your context profiles inside NjectAI.

IMAGE PROMPT

Product on a Desk, Late-Afternoon Window Light

When you need a warm, lived-in shot for a B2B or gadget ad without booking a studio.

A photo of [product] on a [wood / white / concrete] desk next to a window, late afternoon.
Around it: [laptop edge, coffee mug, notebook], slightly out of focus.
Camera: eye level, 3/4 angle, 50mm look, product sharp, background soft.
Light: warm low sun from the window side, long soft shadows across the desk.
Frame: vertical 4:5, product in the lower third, clean wall space above for a headline.
Style: real photo, natural colours, light grain. No text, no logos added, no watermark.

👉 More image prompts in the library. Works in Grok Imagine, ChatGPT and similar image models.

🧠 What did you think of this one?

Login or Subscribe to participate

Hit reply and tell me which AI tool has more access than you'd like. I read every reply.

🔑 God of Prompt

P.S. If you only do one thing today, turn on auto-updates everywhere. Takes five minutes. Then run step 3 above. The prompt library has more like it.