Good morning guys, happy Friday, and before we jump in—I saw a thing called Kickbacks this week and stole the idea. And how I did, you might find valuable.

I don’t feel bad either. This newsletter is all about stealing ideas from products and companies that have shown proof of growth, and then applying our own flavor given our context. Good artists copy, great artists steal—or something like that.

So, what did I steal? You can find out here on bits.the-diff.com , but TLDR…

It’s for product builders and engineers using Claude/Codex in the terminal who want to learn while you wait. Instead of just …Discombobulating...you get one-liner headlines and ideas from the topics you care about—right in your Claude Code status line while the agent works.

You pick your topics (i.e Venture, Startups, Tech) , run one install command, and you’re done. Stay informed with what’s going on and get bite size ideas while your agent finishes a task.

One line to install. Uninstall anytime.

Open source. 100% free. No account or sign up.

It’s in beta, but I made it for you because you’re here reading 5-Bits, meaning bit size ideas and TLDRs matter to you. I’d love for you to try it, and let me know what you think.

In the next 5 minutes over your coffee: the tokenmaxxing AI bill that just came due (and how companies are slamming the brakes after splurging), the day ChatGPT stopped being the default, the App Store fee switch most builders overpay straight through, the 37-day clock almost everyone is ignoring but could cost millions in violations, and where the data shows app screenshots fail.

Remember the roaring 20s of every CEO telling their team to tokenmaxx, to burn as many AI tokens as humanly possible? The invoices are in—and they’re not going to do it anymore. Uber torched its entire 2026 AI budget by April, then capped spend at $1,500 per person per tool on Claude Code and Cursor. Microsoft pulled most of its Claude Code licenses. And one company ran up a ~$500M Claude bill in a single month after forgetting to set usage limits. …. you’re fired

The incentives went feral. Uber ranked teams on a leaderboard by AI usage. Amazon reportedly scrapped its own usage tracking after catching staff firing the most advanced models on Earth to check the weather. Uber's own COO admitted he can't draw a line between all that spend and shipping anything useful. Per-token prices keep falling, but agents consume so much more that the bills balloon anyway, and Goldman thinks token usage 24x's by 2030.

Watch your spend. Set limits. If you’re overseeing a team…right now start seriously treating your AI line item like it's payroll, because effectively it now is. Set per-seat caps before you need them, route cheap tasks to cheap models, and treat the broad middle moving from low to moderate usage as the real ROI, not your power users gorging. The FinOps crowd already flipped from “go fast” to “where did it all go.”

I made a free research tool (well, 4 of them), but this one is my favorite—the App Competition Calculator shows you the volume of new apps in a category, the MRR growth of that category, and the competition heat level.

You can also find apps like yours with keywords and see your app name’s ranking in the App Store — and drill down by who’s new, who’s releasing more frequently, and who’s getting the most downloads.

Build in partnership with my friends at RevenueCat, who have by far the largest dataset of app and subscription data out there.

Dang, Yang Yang is beating me at my own game. 🙉 Well played sir.

Just drop in your category and it shows the newest, top, and trending apps you're actually up against, and how hard the space is to crack.

I also built 3 other free ones. Like one to dig into the unit economics of how much your app can make, plus an app health score and an ask-anything tool. Poke around the whole set.

For the first time ever, ChatGPT is under half the AI-assistant market: 46.4%, down from a clear majority in January. Gemini's at 27.7% (riding Google's distribution), Claude at 10.3% with the highest paid-conversion rate in the field (13% of users pay). ChatGPT still dwarfs everyone at 1.1B monthly users, but the single-app era is over.

First, optimizing for a single assistant now misses most of the market. ChatGPT alone reaches 46%, adding Gemini gets you to 74%, adding Claude 84%. Second, and the wild part: switching is now driven by values, not just features. When OpenAI signed a Department of Defense deal in February, US ChatGPT uninstalls spiked ~200% and Claude downloads jumped. People voted with their installs.

If AI assistants are a distribution or discovery channel for you (they are), you're in the businesses of optimizing for at least three now. Test how your product (and how it shows up in AI answers) lands across ChatGPT, Gemini, and Claude. And take the trust signal seriously—your stance on data, ads, and who you partner with is becoming a feature people will switch over.

Everyone has been watching the model wars, but the app stores have changed the money math. Google split its cut into a 10% service fee + 5% billing fee, and with Apple, a huge share of devs are now overpaying by 2x. Avoidable though..

For most apps under $1M, the 30% “Apple tax” is a choice now. The Small Business Program halves your commission to 15% from day one. On real revenue, that's a real reason to spend five-minutes doing it.

Go check your enrollment status today on the App Store Connect agreements, and for the real details and how to’s 👇

I wish the answer was just as simple as my content got better. But the reality is I had the tools to help me figure out who was reading (and who wasn’t), what was working (and what wasn’t), and use two tools I never had on Substack: Segmentation and Automations.

And I didn’t even have to click and do it myself—all in Claude via their MCP.

beehiiv truly is the backbone of my operation. It’s helped me be more creative and expressive with design, more savvy with how I do emails, and earn well over 2-3X more a month here than I was on Substack (despite having more readers there).

Not all recommendations are equal—but if you’ve ever thought of sharing something you have to say with others—including influencing customers through data-driven email newsletters—beehiiv is the place to try it

On August 2, a big chunk of the EU AI Act goes live, including the transparency rules. The big one that’s easy to default on…you have to disclose when someone is talking to an AI, and label AI-generated content. Miss it and the penalty ceiling is €35M or 7% of global revenue.

A recent survey suggests 78% of organizations haven't taken a single meaningful step toward compliance.

It applies to you even if you're nowhere near the EU. If your product touches EU users, you're in scope. And it's not just the model labs: ship a chatbot, a generated-content feature, or anything AI-adjacent and the disclosure duties are yours.

Fine. I am not compliant. Little Moments has AI and I do not disclose it in the app. But 7% of global revenue for me is doable. Fun fact…that’s a $7 fine.

So let’s together spend one hour this week on the boring stuff. Taking inventory of every AI feature we’ve shipped and labelling anything generated.

When people are scrolling through apps the gallery is the workhorse for your pitch. Never the description. Ever. People are about 10x more likely to swipe your screenshots than read a word of your copy, with just ~2% ever tapping “read more.”

Everyone sees screenshots 1–3 but almost nobody scrolls past the third frame. Yet most apps burn prime frame 2 real estate on a logo slide or a second hero shot.

Most including me…

What I learnt this week → Frame 1 is the hook (singular promise). → Frame 2 is where you either show the product actually working or lose them.

The holy trinity sequence that converts for an app is Value → Usage → Trust: the promise, then the app in motion, then proof. Broadly though for any product, people want to to see cleanly a “shot” of the core action you sell paying off. Don’t bury what people get (visually) with clever tags.

This weekend, I’m working on storyboarding my 5-frame narrative…see what that is 👇

› For parents who want to capture the little moments — I’ve noticed an interesting usage pattern in my app, the most active users are parents capturing their kids growing up. If you have a little one, do you want to forget the little moments with them? (No, show me how your app helps with that).

› Nobody likes the guesswork on how you’re doing compared to others — and access to data that answers that for your product can be a chore to sift through, which is why I did it and packaged it for you in our free set of tools— the app revenue/economics calculator, an app health score meter, and the competition heat calculator from up top (See how your app compares).

› My go-to stack with the vetted tools I trust everyday — the products I genuinely lean on to build things at work, my apps, and run my newsletter business (Explore my personal app store of recs).

› Fresh product finds that seem cool — these are not vetted by me personally like the above, but I like to watch new launches from the week and support, and these three interesting

››note.md (a local-first research and note taking space for you and your LLMs)

››Buddy Figma Agent (the plugin to go from Claude Design to Figma)

››Locofy.ai (Build frontend 10x faster with design-to-code AI all in localhost).