AI got rebranded to SI

what is up guys 👋,

Apologies to everyone who’s been asking if I’ve gone MIA. I had some personal stuff to deal with and needed a bit of a break, but I’m back with the weekly newsletter and we’ve got a lot to catch up on.

Claude actually killed editors, and this time it’s not hype: I actually fired my editor (don’t want to include a screenshot; it's kind of sad) and used Claude Opus 5.5 to edit an entire YouTube video.

I’ve also been autistically obsessed with Instagram carousels, which led me to build Top Carousels, the only tool in the world that has all the viral carousels. It’s literally a cheat code for making carousels (more on that below).

The hype around AI agents is back, so I’ve been testing a lot of them, and I do have a winner.

Also, AI finally got a rebrand, so I guess it’s called SI now?

On September 29, Trump signed an executive order telling US executive branch agencies to replace “Artificial Intelligence” with “Super Intelligence”, or SI, in their communications. NASA is already updating its pages.

Elon Musk is also a fan of the rebranding.

The whole outcome only benefits Slovenia, which owns the .si domain. These domains are now selling for millions of dollars, so that’s really good for its economy.

The actual difference in meaning is very fluffy: AI is designed to perform specific intelligent tasks, while SI (superintelligence) is a hypothetical AI that surpasses the smartest humans in virtually every intellectual ability.

Apart from the hype, nothing has changed, so it’s business as usual.

Opus 5.5 has probably been the best and most reliable model I’ve ever used. I’ve been building a lot with it lately, and editing is probably my favourite use of it so far.

A fun fact about Opus 5.5 is that you get 5x more value from Anthropic subscriptions when using it compared with GPT-6.1 Sol. I’ve literally been building my SaaS with it all week and doing client work, and I haven’t hit any limits.

The recent product updates are worth catching up on too:

For me, this is probably Claude’s most important update.

Claude can split work into separate chats that share the same files and keep working in the cloud when your laptop is closed.

For my content process, I set up a project with one chat researching ideas from Instagram, X and Reddit, another drafting posts, and another checking links, while a board shows what’s finished and what’s waiting for me.

This one is for the nerds who use Claude Code.

Basically, mods are little dashboards in your Claude Code terminal that can do lots of small, handy things, like the examples below.

More Claude updates, quickly:

I love carousels, and I got a little too obsessed with figuring out what makes one go viral.

So I collected and analysed over 10,000 Instagram carousels while building Top Carousels. The library now has 4,000+ outliers, with new posts being added every day.

An outlier is a post that did unusually well compared with that creator’s normal results. A celebrity getting 50K likes might be completely ordinary for them, while a smaller creator getting five times their usual engagement is something I’d study.

Here are a few patterns that stood out in the analysis:

Find an outlier outside your niche → Recreate the structure with AI

For example, I found @themiagabriella’s “4 methods to get your sh*t together” carousel, which reached 18.3x her usual engagement. It used a big, centred headline with objects arranged around it and one underlined word.

I kept that layout and made “5 of these AI agents are a waste of money”, with the agent mascots arranged around the headline.

Mine has since passed the original’s like count, and the numbers are still going up.

That’s what I want the library to help you do: find a structure you can adapt for your own ideas. You can see the full before-and-after here.

If you want to try it, you can start for free here :)

What is Jev AI (System One)?

Jev is TypeSafe’s first System One model. You give it text and a specific question, and it returns a category, score or probability that your software can use to decide what happens next.

What is OpenAI Decisions?

OpenAI Decisions is a developer feature for making similar decisions with OpenAI’s models. It accepts both text and images, so you can ask questions about what’s actually visible in an ad.

What is the difference between them?

The practical difference here is that Jev works with text, while Decisions can inspect images too. Here are three ways I’ve been building around that.

Real & practical use cases 🚨

→ Jev: qualify leads before reaching out

I built a starter that takes saved Apollo contacts or a CSV, compares their company and role with your offer, and labels them as a good fit, a poor fit or needing more research. You define what a good customer looks like, so you can prioritise who to contact.

Get my lead classifier on GitHub.

→ Jev: study a creator’s Reels

I used this on Alex Hormozi. My tool turns each Reel’s speech into text, then Jev labels the hook, structure and CTA. Those labels sit alongside the views and likes, so you can find his most-viewed stories, compare them with his step-by-step videos, and study the openings when planning your own content.

The published comparison covers 431 Reels. Get Creator Lab on GitHub or read the full Hormozi breakdown.

→ OpenAI Decisions: organise Meta ad images

I built an ad-research project around this because it can look at the images. It can label product photos, before-and-afters, people using the product and discounts shown in the creative.

Add your ad account’s sales and spending figures, and you can compare which image categories brought in more sales for the money spent. The image tells you what’s in the ad; your account numbers tell you how it performed.

Get my ad-research project on GitHub.

I’ve been testing ChatGPT, Grok, Muse, Instinct, Hermes and Manus, and ChatGPT Dots was my personal pick until recently. With the latest updates to Grok Bot, though, I’d now say it’s the best overall AI agent.

That's it for this week.

See you next Sunday,