SEPTEMBER 10, 2025

• PromptCraft: AI Companion Directories; Psychological Time Bomb Defuser Prompt; and Help Beating Procrastination​
• The Best Way to 'Teach AI' in the Classroom​
• THE HEADLINES
• Tool Spotlight: AgentSea: Privacy-First Chat Lets You Test Multiple Models

AI Companion Directories; Psychological Time Bomb Defuser Prompt; and Help Beating Procrastination

• ​User-built directory of AI companions: Community post launching a review directory for AI companion platforms, with categories and user feedback to compare features quickly. Useful for surveying options and spotting new entrants; invites submissions if a platform is missing.
• ​Trauma “time bomb” defuser prompt: A PromptFuel lesson frames AI as a “psychological bomb squad” to map triggers, assess risk, and guide de-escalation steps. It’s a structured, narrative prompt for reflective work—not therapy—emphasizing detection methods, defusal protocols, and clear next-step reporting.
• ​PromptCard: quick-insert ChatGPT prompt shortcuts: Workflow to save prompts and inject them with a #shortcut typed directly in ChatGPT’s box. Cuts copy-paste friction, supports very long templates, and adds a small community library for sharing reusable prompts.
• ​Calculus + AI tutor SPN notebook: A detailed “system prompt notebook” for a calculus tutor that always sets prerequisites, uses analogies, provides worked examples, and bridges to AI/ML applications. Includes structure, tone, academic-honesty guardrails, and a repeatable lesson template.
• ​Twelve prompts to crush procrastination: A pragmatic set for business writing and execution: instant briefs, two-minute starter tasks, conversion rewrites, angle generation, 2-hour action blocks, brand voice guides, distraction mitigation, repurposing pipelines, and momentum resets.

The Best Way to 'Teach AI' in the Classroom

Hint: It Has Nothing to Do with ChatGPT

I started teaching AI in June 2023, teachers were frequent students in my classes.

Common questions at the time centered on how to detect when a student was using ChatGPT and how to prevent students from using it on assignments.

An honorable mention: are there any tools that can tell when work was done with AI?

(That last one has an easy answer. Nope.)

Anyhow, I’ve long believed they were asking the wrong questions. The best solution: teach students how to use AI effectively.

But, believe me, I get the frustration.

I taught high school for a bit in a time when ChatGPT didn’t exist. There was nothing more frustrating than when the dumbest kid in class turned in a perfectly written essay.

You knew immediately it was bullshit, but was it really worth the effort to open that can of worms? (Especially when it was almost guaranteed his otherwise incommunicado parents would take his side if you called him out on cheating.)

Believe me, teachers, I’m not judging you for asking the wrong questions. But I do want to help you find a more productive way to navigate this AI thing. First:

Stop trying to stop them from using it.

You’re not going to.

Children know more about technology than most adults.

For us Gen Xers and Millennials, technology was something we “learned.” With younger generations, it’s built into their DNA.

They’ll never know a world where the Internet doesn’t exist. Any obstacle you come up with, they’ll eventually find their way around.

Technology is a language they learn from the time they’re old enough to walk, and no matter how far ahead you are, they will catch up and surpass you.

So, instead of preventing them from using it, teach them the when and the how.

To do that effectively, you have to do one thing that has nothing to do with ChatGPT or any other genAI tool.

​CHECK OUT THE REST ON SUBSTACK​

HIGHLIGHTS: NotebookLM Tips; Cognition AI's $400 Million Raise; Major Orgs Sign Up for AI Education Support; and Orgs Are Using ... Less ... AI!?

• ​The United Arab Emirates Releases a Tiny But Powerful AI Model​
• ​ASML becomes biggest Mistral investor in boost to Europe’s AI ambitions​
• ​6 ways to use NotebookLM to master any subject​
• ​Cognition AI defies turbulence with a $400M raise at $10.2B valuation​The United Arab Emirates Releases a Tiny But Powerful AI Model​
• ​ASML becomes biggest Mistral investor in boost to Europe's AI ambitions​
• ​6 ways to use NotebookLM to master any subject​
• ​Cognition AI defies turbulence with a $400M raise at $10.2B valuation​
• ​Lilly launches TuneLab platform to give biotechnology companies access to AI-enabled drug discovery models built through over $1 billion in research investment​
• ​Anthropic endorses California’s AI safety bill, SB 53​
• ​Nebius signs $17.4 billion AI infrastructure deal with Microsoft, shares jump​
• ​Major Organizations Commit to Supporting AI Education​
• ​AI adoption rate is declining among large companies — US Census Bureau claims fewer businesses are using AI tools​
• ​Hackers threaten to feed data to AI if their demands aren’t met​
• ​Ralph Lauren introduces Ask Ralph, a new conversational AI shopping experience​
• ​PwC’s U.K. chief admits he’s cutting entry-level jobs and taking a ‘watch and wait’ approach to see how AI changes work​
• ​The Projected Impact of Generative AI on Future Productivity Growth​
• ​Anthropic’s Landmark Copyright Settlement: Implications for AI Developers and Enterprise Users​
• ​AI Mode in Google Search expands to more languages​
• ​When AI Speaks for the Dead​
• ​What Happened When Researchers Co-Founded a Startup with AI​
• ​Can researchers stop AI from making up citations?​
• ​How Simplify in the Google app makes complex text easier to understand​
• ​A People-First AI Fund: $50M to support nonprofits​
• ​AI will consume all of IT by 2030—but not all IT jobs, Gartner says​
• ​Publishers fear AI summaries are hitting online traffic​

AgentSea: Privacy-First Chat Lets You Test Multiple Models

AgentSea is a unified, privacy-first AI chat that lets you use the latest models without losing memory or context. You can switch models mid-conversation, keep one continuous thread, and tap built-in tools like image generation and social search across X, Reddit, Google, and YouTube.

A growing marketplace offers hundreds of community agents to jump-start tasks. The standout is Secure Mode: chats run on open-source or AgentSea-hosted models, helping keep prompts and outputs out of third-party training pipelines—great for sensitive or professional work.

Launched with free options and a first-month-for-$1 promo, AgentSea aims to replace app-hopping with one safer, smarter workspace.

​https://agentsea.com/​

​Read All mAIn Street Back Issues Here​

Your input fuels the content. Let me know how I did today!