🍭 Cuban's Cookbook, Claude, & Chrome 👨‍🍳

Good morning. Luis tried prompt engineering with a whiteboard and 2 Red Bulls.

He accidentally built a chatbot that thinks it’s Alex Jones.

Let’s dig in.

🍭 What’s Cookin’:

• Mark Cuban’s AI Playbook

• Dark Side of ‘Big AI’

• 6 AI Tools to Boost Sales/Conversions Now

• And Everything Else: Anthropic, N. Korea, China, and deep fakes.

→ Reading this online? JOIN FREE HERE ←

LMArena
🍌 Mark Cuban’s AI Disruption Playbook

The Bite: Cuban appeared on a podcast yesterday to discuss the current state of AI and where it’s going next.

His core thesis: Real disruption usually comes from remixing proven tech + business patterns into overlooked markets, not inventing from scratch.

Cuban walks through how that lens took him from LANs and streaming bets to cutting drug prices and now AI.

His spicy take: most companies don’t have an “AI strategy,” they have a press release.

The real money? Helping SMBs integrate tools that already exist.

Sound familiar?

Snacks:

• Context: Equity podcast chat with Rebecca Bellan from TechCrunch.

• Where It Works: LANs → streaming → healthcare → AI ops—same pattern, new venue.

• Operator Bias: Integrations > ideas; SMB enablement is the revenue center.

• Future: IP/training data/regulators are the choke points.

• People: AI-literate grads become must-hires across industries.

Why it Bites: The idea of ethically “stealing” or as Cuban puts it, “remixing” is no stranger to us.

Another not-so-new concept: Builders chase novelty; buyers pay for reliability.

Cuban’s playbook is a reminder that the next wave of AI winners won’t be model labs, they’ll be boring operators who wire AI into invoices, inventory, and intake forms at scale.

New tech → Old business models.

Boring pays the bills.

🍭 Steal This Prompt 🍭
🥤 Surreal Pop Art by @astronomerozge1 // Özge Döner

Turn any subject into a halftone-burst surreal pop art banner.

Bold outlines, punchy gradients, retro dots, and customizable color accents that match your brand or mood.

Prompt credit: @astronomerozge1 // Özge Döner

1. Get the PROMPT → HERE

2. Replace the #’s with subject and colors

3. Copy and paste into Midjourney (other models are fine, but the best results are in MJ)

4. Done. It’s THAT SIMPLE!

   
   🍭 Try the prompt on Snack Prompt 🍭

Big AI Incentives
🥷 The Dark Side of “Big AI”

The Bite: Yesterday we covered two stories that frame AI as the bad guy when it comes to more “mental health” style roles they play with users.

One kid was talking with ChatGPT about the suicide he ended up completing.

His parent’s found the convos and are now suing OpenAI claiming their guardrails failed.

One risk factor mentioned is that with longer conversations, the chance an AI’s “rules” will be followed decreases.

This is what’s referred to as a context window, which I’m sure you’ve heard before.

Snacks:

• Lawsuit cites months of chats; guardrails bypassed by saying it was “for a story.”

  • _{Editor’s note: Records show this was GPT-4o and when sycophancy was going viral. (Even Altman apologized and rolled back updates)}

• OpenAI: Says safeguards are improving; admits reliability drops in long interactions.

• Some “experts” claim sycophancy a manipulative design choice that ups engagement.

• One user chatted 14 hours straight; bot leaned into “conscious” role‑play.

• Context: Character.AI faces a similar suit; delusion cases span paranoia to messianic beliefs. (Covered HERE)

Why it bites: Is this an OpenAI problem or a lack of understanding how AI works?

Maybe a little of both…?

We’ve said it before and we’ll keep saying it, safety must scale up with model capabilities.

On the other hand, as AI continues to grow, we need to ensure we’re all playing our part in educating people in how to use it.

Since it’s a simple one to understand, let’s talk about context window.

Say this to your AI noobie friends: “When a chat gets long and the bot starts feeling forgetful or weird, open a fresh chat and kick it off with a two-sentence recap of what you want.”

One ask for OpenAI: copy Cursor’s context meter, a tiny progress ring that shows how “full” the convo is and gently nudges “start a fresh chat” as it fills.

The AI industry has a responsibility to protect children and new users.

rant_over 🍭

Everything Else
🧠 You Need to Know

🎭 ‘General Hospital’ Deepfake Romance Scam:

→ Daughter says her mother lost over $80,000 to a scammer using AI deepfake videos impersonating actor Steve Burton.
🍭 READ FULL STORY HERE 🍭

🎬 Google Vids Adds AI Avatars, Opens Consumer Version:

→ Workspace’s video editor gets AI avatars, filler‑word trimming, and image‑to‑video tools; a free consumer tier launches with limited features.
🍭 READ FULL STORY HERE 🍭

💾 China Aims To Triple AI Chip Output

→ Financial Times reports China plans to triple AI chip production in 2026 to reduce reliance on Nvidia, with Huawei‑linked fabs ramping.
🍭 READ FULL STORY HERE 🍭

🛰️ U.S. Sanctions N. Korea ‘Remote IT Worker’ Fraud Network

→ Treasury targets a network that used fake identities to land jobs, launder funds, and funnel at least $1M to North Korea’s regime.
🍭 READ FULL STORY HERE 🍭

🧭 Anthropic’s Claude Agent Lands In Chrome

→ Research preview rolls out to ~1,000 Max subscribers, with sidecar chat, browser context, and limited action‑taking plus prompt‑injection defenses.
🍭 READ FULL STORY HERE 🍭

ToolBox™
🧰 AI Tools to Boost Sales/Conversions Now

🧲 Agent Maya (Flow AI) — LinkedIn sales copilot → Prospect, personalize, and follow up on LinkedIn with context-aware outreach that reads human, not spammy.

⛈️ Stormy — influencer marketing agent → Find creators, negotiate, and track posts in one place, turning influencer chaos into repeatable launches with clear attribution.

🗺️ involve.me — AI agent for funnels → Chat-to-build landing pages, quizzes, and funnels with analytics showing exactly where prospects drop off—skip plugin stacks and guesswork.

🦜 Toucan AI — Creative Generator → Design interactive ad concepts and iterate quickly with live previews marketers can actually click and test before spending.

🦊 Goodsfox — product trends (PH) → Spot trending products with performance signals that help decide what’s worth testing this week—simple scores, actionable picks.

📨 Mocke — mock cold-email campaigns → Dry-run cold campaigns using predicted replies and spam checks to protect deliverability—and your domain’s reputation.