karozi

30 Creators Tested What Claude Opus 5.5 Can Do with One Vague Prompt

I asked everyone to start with the same short prompt. The reported satisfaction scores ranged from 0 to 9.5 out of 10. A read for builders, PMs, and anyone who refuses to ship without thinking.

TL;DR: I asked Substack writers to give Claude Opus 5.5 one vague prompt to make a video, change only their publication name and URL, and rate the result from 1 to 10. Twenty-one participants gave a numeric score, covering the entire scale from 0 to 10, with a median of 6/10 and a mean of 5.36/10. Nine landed in the lukewarm middle (4 to 7), six were satisfied (8 to 10), and six were disappointed (0 to 3). Anna Levitt revised her video, and her score went from 5 to 7.5. My conclusion: one vague prompt often gets you a decent first draft, but rarely a finished video, and the people who already had saved skills and brand context weren't running the same experiment as everyone else.

0 to 10.

That’

karozi

Write for Product With Attitude

Your next (or even first) Substack article could run on Product with Attitude. A read for builders, PMs, and anyone who refuses to ship without thinking.

Hey

Your next (or even first) Substack article could be a contribution to Product with Attitude.

I’m opening this opportunity to paid subscribers as a membership perk. Your publication’s size doesn’t matter. I care about the topic, your perspective, and the attitude you bring to it.

karozi

AI SEO: I Ran 6 Reddit Experiments Over 18 Months. Here's What Worked.

One experiment brought 438 subscribers. Still, my long-term strategy took a different direction. A read for builders, PMs, and anyone who refuses to ship without thinking.

Your reader has 14 things to do today. Finding your article competes with all of them. So does reading it.

You’re busy too. (If you’re not, please reach out. I need you as my guru.)

The traditional advice was to “meet your readers where they are.” TikTok, Reddit, Facebook. The entirety of the internet, it would seem.

karozi

OpenAI Dots Might Be the Biggest Deal for AI Beginners Since ChatGPT Itself.

AI PM turning everyone into critical AI thinkers & builders. Join tens of thousands readers from 146 countries, and a community developing critical AI literacy through practice. Use AI to strengthen the branch you’re sitting on, not cut it off. A read for builders, PMs, and anyone who refuses to ship without thinking.

OpenAI recently introduced Dots, always-on AI agents that get to know and track your goals, and can take action on your behalf.

Dots are absurdly well-equipped for something with such a tiny name. GPT-6 Astra. Their own cloud computers and browsers. Thousands of app integrations. Background agents. And no need for our laptops to stay awake.

Your Dot learns what you need. Then tracks all activities. Keeps working on your projects even when you’re busy bein

karozi

How to Build a Halloween Game for Your Readers With ChatGPT Sites. Before Saturday.

41 words, 30 minutes, and the reverse interview method that turns a vague idea into a working app. A read for builders, PMs, and anyone who refuses to ship without thinking.

I want to build a Halloween web app where a reader enters their email, opens two treasure chests, and finds a reward. Three chests. Two offers. One empty chest. Ask me one question at a time until you understand my requirements.

41 words were enough to kick off the build with ChatGPT Work.

30 minutes later, I had my first playable draft.

karozi

OpenAI Just Turned AI Answers Into Tools. I Want This as a User. As a Publisher, I’m Worried.

GPT-6's Intelligent UI makes the source visit optional. This may become one of 2026’s consequential shifts for publishing. A read for builders, PMs, and anyone who refuses to ship without thinking.

TL;DR: On October 7, 2026, OpenAI began rolling out GPT-6 with Intelligent UI in ChatGPT: answers that arrive as charts, controls, calculators, and games. As a user, I want this. As a publisher, I’m worried the interface makes the source visit optional, citation or not. Pew’s March 2025 data already showed the pattern: users clicked a Google result in 8% of visits with an AI summary, versus 15% without. This post separates link, answer, and interface displacement, and lists what to test if you have access.

Oh, fellow writers and publishers, you’re not going to like this.

I really hope I’m wrong and someone finds a hole in my logic.

karozi

You Don't Need an AI Newsletter.

You only need 3 steps to start building with AI. You don't need an AI How-to newsletter. How to tell which AI newsletters are worth paying for. Plus 6 AI newsletters that pass the test. A read for builders, PMs, and anyone who refuses to ship without thinking.

TL;DR: You can start building with AI by describing the result you want, asking it to interview you one question at a time, and reviewing the proposed plan before it builds. Then test the result yourself. AI newsletters are worth paying for when the writer brings original experiments, useful tools, and thinking that challenges your own. Below: three steps to get started and six writers who give you reasons to keep reading.

What I’m about to share today might surprise you and raise an eyebrow or two among fellow writers.

I write an AI newsletter, and I don’t think you should need one open beside you to start automating your work with AI.

karozi

What Is Jev? Why Is Everyone Crazy About It? What I Built for Less Than $1.

LLMs talks through a problem. Jev makes a small, structured decisions. You can keep using Claude Code or Perplexity Computer and add Jev to the workflows you build with them. A read for builders, PMs, and anyone who refuses to ship without thinking.

TL;DR: On September 15, 2026, TypeSafe launched Jev, the first System One judgment model, and it became the fastest-adopted model in Vercel AI Gateway history: nearly 13% of paid teams used it within 24 hours. Pricing is $0.042 per million input tokens, with output tokens free. Measured on my own account: $0.000126 per four-question scoring request, so a $5 sign-up credit covers roughly 39,000 requests. The non-technical version: Jev is not another chatbot. It is a specialist at a sorting station. You hand it information

karozi

What Jev Means for AI Agent Security. The Security Gamble Behind “Jev Can’t Hallucinate”.

A CISSP's field guide to using Jev, TypeSafe's System One model, in AI agent security: where cheap AI judgment fits, stops, and what attackers will target. A read for builders, PMs, and anyone who refuses to ship without thinking.

TL;DR: Jev is TypeSafe’s System One model: it returns typed, probabilistic decisions instead of free-form text, fast enough and cheap enough to call on every agent action. In AI agent security, that makes it a judgment layer: a cheap sensor for suspicious tool calls, risky requests, and review routing. It is not an authorization mechanism, it can be confidently wrong, and attackers have every reason to manipulate its inputs. This guide covers where Jev helps, where existing security controls still matter, and how to deploy it without giving a model permission it should never have.

Last week, I shared two articles:[one on governing AI agent fleets](https://karozieminski.substack.com/

karozi

Weekly Long-Tail Question Prompt for AI Search Visibility

How one simple recurring task can improve your AI search visibility. A read for builders, PMs, and anyone who refuses to ship without thinking.

I’ve written countless Product with Attitude articles about questions I was trying to answer for myself.

For a while, I trusted myself to remember those questions. This was optimistic. So I gave the remembering job to AI.

I now have a recurring task across ChatGPT Work, Claude Cowork, and Perplexity that sends me every long-tail question I asked during the previous week.