Good morning guys, it's our 2nd edition of 5-Bit Friday’s rebooted, it’s June 12, and has anyone else noticed LinkedIn is becoming like YouTube shorts? I’m seeing video everywhere…brace yourself for “90 things I learnt about b2b sales from my grans 90th birthday” videos.

In the next 5 minutes over your coffee: ChatGPT's soft launch ad system and what it means, Apple's billion-dollar-a-year brain rentals, the funnel the top 1% of apps have moved to, the acquisition that lets agents work for weeks, and the IPO sprint that changes your vendor math.

Jaryd Let's get into today's 5-Bit!  — Jaryd

Five things to know and think about before the weekend .

1
2
3
4
5
What To Know Why It Matters Applying It Tools to Try Done ✅
<5 min read

PS—Did someone forward you this email? Join 27K curious readers here. Feedback is a gift, and I read every email — let me know how I'm doing by replying to any email.

Together with our awesome partners

I added my weekly + 3-Day Trial this week with them

Last time I wrote about this big gap in the app market, where less than 15% of apps are using weekly plans, yet weekly plans are driving most app revenue. And I also said I was going to test it on Little Moments. I am, and that’s why I love RevenueCat

  • they power my entire sub ecosystem

  • they make is simple to experiment with offers and test placements

  • they make it easy to try new versions of my paywall

  • plus their new AI, Rico, is an underrated growth tool

If you’re building out an app even for fun, add RevenueCats free SDK

See how they can help build and grow your app business  ↗

Check out RevenueCat, free under $2.5k MRR

+1 TLDR / ChatGPT just became a performance ad channel

Not brand awareness ads. Real, conversion-optimized, pay-when-they-buy campaigns went live for early advertisers on June 5, and the pilot reportedly crossed $100M in annualized ad revenue within six weeks of launching.

→ Why it matters

This is the early-Facebook-ads moment people will claim they saw coming. ~700M weekly users, a pixel, a Conversions API, and CPA-style pricing where you pay for the action, not the impression. Criteo's early client data showed AI-referred conversion rates approaching 2x traditional search in some retail categories. The inventory is cheap and the auction is empty right now. That never lasts

→ Applying it

If you spend anything on Google or Meta, set up the OpenAI Pixel or Conversions API this weekend and run one small test on a low-friction action (trial start, signup, content download). See how CAC compares against your existing channels.

+1 TLDR / Apple just told us the model isn’t the moat

Has Apple come to terms with what we’ve all know has been a total joke for years? The new Siri unveiled at WWDC runs on a custom Gemini model that reportedly costs Apple ~$1B a year. Xcode 27 ships with coding agents from Anthropic, Google, and OpenAI built in. Even Apple's AI image watermark is Google's SynthID.

→ Why it matters

The most vertically integrated company on the planet looked at the model layer and decided to rent it. What Apple kept for itself is the part that compounds: the device, the distribution, and the personal context (your messages, photos, and screen) that makes the assistant actually useful. If Apple won't bet the farm on owning the weights, "we built our own model" is probably not your startup's moat either.

→ Applying it

Audit your own pitch. If your differentiation is the model you use, you have none. Differentiate on proprietary context, workflow, and distribution, and keep your model layer swappable underneath.

+1 TLDR / The top apps have stopped growing on the app store

From a session with 150 app growth leaders this week: 82% of top-grossing apps now run web2app funnels, and users acquired on the web generate nearly 2x the LTV of users who convert purely in-app.

→ Why it matters

Web2app means you skip the 15–30% Apple tax, get full-funnel attribution back, and can run paywall experiments without waiting on App Review. The store is becoming the endpoint of the funnel, not the funnel. The same session's other consensus point: most teams over-invest in acquisition and under-invest in everything after the install. The two compound—own the funnel, then own the activation.

→ Applying it

Pick one paid campaign and route it through a simple web onboarding + Stripe checkout instead of straight to the store listing. Compare blended CAC and 30-day LTV. (I'm running this exact math on my own app right now, so misery loves company.)

btw, those web2app funnels…you can do them in RevenueCat

Build custom onboarding, surveys, and checkout flows that unlock seamless in-app access — all for free under $2.5k a month in sales

+1 TLDR / Agents are graduating from minutes to days

OpenAI announced yesterday it's acquiring Ona, whose tech gives agents secure cloud environments to work in over long stretches. Codex is now at 5M+ weekly users, up 400% since earlier this year, and its most valuable work increasingly runs for hours or days, not one chat turn. (Same new value prop with Anthropic’s Fave btw)

→ Why it matters

The unit of AI work is shifting from "answer my question" to "take this job and come back when it's done." That changes a lot downstream: the product surface (an inbox of finished work beats a chat window), the infra (agents need their own workspaces and permissions), and the pricing (remember last week's bit on metered agent billing—this is why).

→ Applying it

Find the one job in your product a user would happily hand off overnight, and sketch the delegate → review → approve loop for it. That loop is the next onboarding flow.

+1 TLDR / The AI labs are sprinting to Wall Street

Anthropic filed a confidential S-1 on June 1. OpenAI filed its own on June 8, last valued at $852B. Two foundational labs racing to the public markets within a week of each other…? The cost of rented intelligence is not rent stabalized

→ Why it matters

Public companies answer to quarterly revenue, and you're the revenue. Connect the dots from this edition alone: ads in ChatGPT, metered agent pricing, infra acquisitions. The era of subsidized intelligence is winding down, and the pricing you're building your AI features on today is a snapshot, not a guarantee.

→ Applying it

Re-run the unit economics on your AI features with a margin buffer. If a feature only works at today's token prices, it doesn't work. Re-check quarterly like you would any supplier.

save yourself a ton of tokens on your next app build

I burnt through tokens like the Joker burnt that pile of cash in early apps builds, trying to get my stuff to work, things to be compile, not crash and actually have a stable foundation.

After a few runs, I built this master app prompt (one copy and paste) that sets all the configs, has context, and churns out a real working app on Sim that works on device — letting you start the creative part much faster with less token usage.

+ICYMI / The latest edition

Less than 15% of paid/subscription apps offer a weekly plan, yet weekly plans are now printing 55.5% of total subscription app revenue. It’s the fastest growing share of revenue across all plan types.

I love this because a) Nobody is really doing it, b)The data proves it works, and c) It's trivial to steal. If you missed this idea, tap below!

ICYMI: Why weekly paid plans are winning  ↗

💭 annual and monthly is leaving money on the table

+My Stack / Check out some of the products I love

Today, quick s/o features for these ones…

  • Loom— quick videos for reporting bugs, sharing feedback, sending proposals to sponsors

  • Steel.dev— best browser infrastructure for agents (think AI being able to take over browsers and to stuff for you)

  • Expo—my build and submit pipeline for Little Moments and the cross-platform stack I use to vibe code native apps

Jaryd

See you next time, and thanks for reading!

— Jaryd

Liked this? Forward it to one builder  ·  Say hi on LinkedIn

Reply and tell me how you might try this one. I read them all.

Subscribe for more editions  ↗

🤝 Join 27,000+ PMs, founders, and builders

Keep Reading