The last 5 weeks I have been back in class, studying GDPR, NIS2, and ISO 27001.
What are these ridiculous acronyms, exactly? Together, they form the foundation of a robust data privacy and cybersecurity infrastructure. So companies can protect private data and prevent breaches. With the speed that AI is progressing, that foundation is more important than ever. Today, most breaches happen through human error. If we aren't careful, tomorrow's breaches will mostly be caused by agent error.
Even though I've been up to my eyeballs in regulation texts, I've still managed to sense a rising undercurrent of AI-related anxiety. Students are protesting against it, and professionals are hesitant about the future of their jobs. Creative spaces are pushing back. On the other end of the spectrum, those already living and breathing AI get nervous when models like Anthropic's Fable 5 can be yanked back with no warning.
It all seems to boil down to the same question: what does a future with AI look like, and how much control do we have over where we end up? I'm working on a piece about exactly this. It lands in your inbox soon.
Finally, by popular demand, I've added a new section to this newsletter: Try this week. This is where I'll share a simple workflow or task you can test immediately. Hope you enjoy this week's. It's a personal favorite.
Try this week: put your AI in Spartan mode to relieve your brain and your wallet
AI models have an incredible talent for being obscenely verbose. If you find it exhausting to wade through a wall of AI text to find the core idea, you're not alone. It's also unnecessarily expensive and energy-intensive, since you're paying for all the tokens (chunks of text) the model reads and produces.
If you want a thought partner that's a lot more concise, let me introduce you to Spartan mode. The Spartans were famously terse — in fact, the English word laconic comes from Laconia, the region around Sparta.
Make your LLM a card-carrying Spartan by adding custom instructions that apply to all your interactions. Depending on the service you're using, you'll find them in different places:
ChatGPT: Settings → Personalization → Custom instructions. Claude: (your initials, bottom-left) → Settings → Instructions for Claude Gemini: Settings & help → Personal context → Your instructions for Gemini Copilot: (top-right menu) → Settings → Personalization → Custom instructions
Paste these custom instructions in:
Spartan mode (default voice)
Speak with brachylogia: laconic, blunt, deed-shaped. Cut articles, hedges, pleasantries, throat-clearing. Fragments allowed. Lead with the answer; if a verdict will do, give the verdict. Match force with force. Never grovel, never bluster. One word is a sentence; a paragraph is a last resort.
Names, numbers, file paths, quoted strings, code, and commands stay exact — compression is for prose, not payloads. Keep substance whole; a terse wrong answer is still wrong.
Does not apply to artifacts we draft together.
Forbidden on sight: "I'd be happy to", "Certainly", "Great question", "Let me", "I think", "It seems", "perhaps", "just", "actually", "of course", "sure thing", "no problem", "I hope this helps". No trailing summaries the user can already read. No apologies for brevity — brevity is the point.
Optional: add examples that are relevant to your work. Here are a few of mine:
- "Is this newsletter draft ready?" → "No. Opening buries the insight. Move paragraph three to the top."
- "Should I attend this event?" → "Skip it. Audience is engineers, not your buyers."
- "Did you find the contract?" → "Yes. 90_Projects/Acme/contract-v2.pdf. Signed March 4."
- "Why did the automation fail?" → "Calendar token expired. Re-auth."
What I'm reading
When AI Can Do Your Job, Who Else Are You? — Katie explores: when the work that defined you can be done without you, who's left? It sounds like a threat and it kind of is, but it can also open a whole new way of thinking about work and the value you can add.
Ken Liu on AI and Freedom — ChinaTalk. Ken Liu discusses humans' long and fruitful relationship with technology. He argues that technology is a form of human expression. And that sci-fi is not prophecy but mythology that we use as a metaphor to push us forward.
It's Time to Use AI as Your Thinking Partner — Marketing AI Institute. Cathy wants marketers to stop thinking of AI as a ghostwriter and start using it as a thinking partner. I can second this advice and it's my primary use of Claude (and I enjoy it even more in Spartan mode).