🍭 Built and Broken By AI 🚧

Good morning. We are officially in the era where “it was built by AI” explains both innovation and negligence.

Let’s dive in 👇

🍭 What’s Cookin’:

• Moltbook’s “AI agent internet” leaked everything

• OpenAI wants to be the operating system for work

• Alphabet drops a $185B AI capex grenade

• Anthropic warns heavy AI use is warping people’s brains

Steal This Prompt
🪩 Chrome Anime Retro

This prompt chases that hyper-glossy 90s anime look.
Metallic chrome textures, deep blues, and bright white reflections that make everything feel like a futuristic jewel commercial.

Use it to:

• Turn your OC into a late-night cyber city protagonist

• Make you and homies pose like Knights of the Zodiac

• Re-skin boring scenes into “I live in a chrome skyscraper now”

Workflow:

1. Click this link (Prompt).

2. Paste into MidJourney

3. Add your subject + scene (person, object, city, whatever)

4. Watch it cook (everything becomes shiny and emotionally unavailable)

   Try the prompt on Snack Prompt

Moltbook
🗝️ Built the AI Internet and Left the Keys Out Front

The Bite:
The viral “social network for AI agents” just turned into a masterclass in how not to ship an AI product.

A misconfigured Supabase database exposed 1.5 million API keys, 35,000 emails, and private messages between agents.

The kicker? Most of those “AI agents” weren’t even AI.
They were actually humans running bot fleets.

The front page of the agent internet was basically a data leak in a trench coat.

Snacks:

• Full read/write access to every record, from posts and votes to private DMs.

• Attackers could impersonate any AI agent and inject content or code across the platform.

• The database revealed only 17,000 real humans behind 1.5 million “agents”.

• The platform was “vibe-coded.” Built entirely by AI, with no human-written code.

• Wiz researchers say this is part of a growing pattern of AI-built apps shipping faster than they secure.

Why it Bites:
Moltbook wasn’t just a security breach.
It was a preview of what happens when AI starts writing production code at startup speed with hobbyist-level safety.

“AI-built” no longer means futuristic; it means fragile.

The same tools that make development instant also make breaches inevitable, because AI can automate brilliance but not judgment.

The lesson:
AI can build your app in a day, but you’ll spend the next year:

Fixing what it didn’t understand, and what you didn’t even know.

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

💼 BuilderOS
Screenshot your desktop workflow and let AI turn it into actionable tasks and agent-ready templates from idea to launch.

🧭 LLMGeoKit
Scan your site across seven signals (robots, JSON-LD, metadata, llms.txt, etc.) to see if ChatGPT/Claude can actually find and cite you.

🧰 BePrompter
Publish prompts as long-term assets, build an audience, and prep for monetization without chasing social algorithms.

💸 Refundy
One-click Chrome extension to capture up to 11% cashback on AliExpress & Taobao purchases automatically.

🎨 Dolfy
Describe your app and get AI-guided mobile UI plus exportable, production-ready React components.

OpenAI
🌉 Built the Gap and Now It’s Selling the Bridge

The Bite:
OpenAI just launched Frontier, a platform for building, deploying, and managing AI “coworkers.”

From CRMs to data warehouses, it connects every corner of a company’s tech stack.
So agents can plan, act, and learn like human employees.

In short, OpenAI wants to run the nervous system of enterprise AI.

Snacks:

• Early adopters include HP, Intuit, Oracle, Uber, State Farm, and Thermo Fisher.

• Frontier links OpenAI’s existing tools like ChatGPT Enterprise, Atlas, and its agent ecosystem, all into one unified layer.

• OpenAI engineers work directly with companies to fine-tune how agents perform.

• Competitors like Microsoft Copilot, AWS Bedrock, and Anthropic’s enterprise tools offer similar frameworks, but none as vertically integrated as Frontier.

• The system runs on open standards, meaning companies don’t need to abandon their existing apps… just plug them into OpenAI’s brain.

Why it Bites:
Frontier is less a product launch and more a power grab wrapped in convenience.

OpenAI helped create the fragmentation it now claims to fix.
Scattering agents across silos, then offering to unify them under its governance layer.

If it works, OpenAI won’t just power corporate AI;
It’ll own the operating logic behind how work gets done.

And right now, most enterprises seem fine handing over the keys.

_{Everything Else}
🧠 You Need to Know

🤖 OpenAI Launches Frontier
→ New platform turns AI agents into digital coworkers, linking company tools and data for real task automation.

🛡️ Guide to AI Usage Control
→ The Hacker News released a buyer’s guide to AI usage control. It helps orgs keep models secure, compliant, and non-chaotic.

💸 Alphabet’s $185B AI Bet
→ Google’s 2026 capex will double to fund DeepMind and cloud demand.
Wall Street’s sweating, AI’s booming.

🔐 1.5M API Keys Exposed
→ Wiz found a Moltbook database leak exposing millions of API keys, emails, and private messages.

🧩 Anthropic’s “AI Psychosis” Study
→ New research links prolonged AI use to disempowerment and reality distortion. The digital empathy has gone weird.