Your AI Doesn't Need Better Prompts. It Needs Persistent Memory.

Advertise to 40,854 E-Commerce Businesses

Promote your tool, service or upcoming event - Reach Founders

Good morning!

In today’s newsletter,

This issue takes 2 minutes to read.

Check out our DTC tool stack here

Let’s dive into it👇

Every time you message Claude or ChatGPT, it starts from zero.

And yes, even with a folder filled with all the relevant context.

This is called RAG. Retrieval-Augmented Generation.

You created the folder on your computer. Dropped in your brand guide, customer reviews, top performing ads, product descriptions, the full Klaviyo flow library.

But every time you send a query, Claude scans the folder. Pulls the chunks that look relevant to your question. Stitches them together. Writes an answer.

This works. But it never gets smarter about your business.

Nothing builds. Nothing compounds. The insight Claude found yesterday is not used to make today's answer better. It just searches again.

On April 2, 2026, Andrej Karpathy, a founding member of OpenAI, put out a tweet solving this exact problem.

19.6 million views in a week.

Instead of Claude reading your raw files every time you ask a question, you build a permanent memory of your business once. It updates itself every session. It compounds forever.

Layer 1 — The raw files. Your source of truth. Brand guide. Customer reviews. Top performing ads. Images. Product specs etc.

The LLM reads these but never touches them. They stay exactly as you uploaded them.

Layer 2 — The wiki. The memory itself. A folder of clean, organized notes the AI writes and maintains.

The LLM extracts the key insights from the raw data.

Layer 3 — The Schema. The rulebook. A single file that tells the AI how to read new sources, how to write new pages, and how to keep the wiki organized.

This is what turns a generic AI into a disciplined memory keeper for your business.

This has been the single best upgrade to my AI output in two years.

And the best part: you don't need to maintain this yourself. The AI does it automatically.

This is what our DTC Daily wiki looks like. Every dot is a piece of our brand. Every line is a connection the AI made between them.

The complete setup, file structure, and step-by-step guide to building this for your brand is below for AI Launch Codes subscribers.

When someone asks ChatGPT to recommend a non-toxic cookware brand or the best DTC supplement for sleep, the AI doesn't go hunting. It recalls what it already knows.

That recall is built from training data. The more your brand name appears across high-quality sources, in context, alongside the right topics, the stronger the picture AI builds of you. Brands mentioned rarely, or only on their own website, are ones AI can't describe accurately. Some it can't describe at all.

This is different from SEO. You're not trying to rank a page. You're trying to exist in AI's memory as the answer to a specific problem.

Industry data shows that 80 to 90% of AI responses rely on earned media rather than a company's owned content. Your product page alone isn't enough. 

Being mentioned in a Reddit thread, a YouTube review, a podcast, or a trade publication isn't just good PR anymore. It's how you teach AI what your brand is and what problem it solves. 

Action Summary:

Neven Eyewear burned through $30,000 to $50,000 testing promos before anything stuck. 10% off. 20% off. A standard BOGO. Nothing.

Then Jonathan switched to buy one, get two free. Day one: 20 orders. Day two: 40. It kept climbing.

But here's the part that doesn't get talked about — customers thought it was a scam.

People were DMing Neven on Instagram saying they didn't believe the brand was real. Jonathan was taking photos of his inventory to prove to strangers that sunglasses actually existed in his house.

Neven even added an FAQ entry just to answer the question before people had to ask: "Is the buy one get two free deal real?"

The offer was so good it stopped people from buying.

Two things fixed it:

Put the proof where the doubt shows up first. Reviews, a real-looking feed, signs that other people have actually ordered. If those aren't there, the deal doesn't matter.

Answer the objection before they ask. If your offer sounds too good, say so in your FAQ. "Yes, this is real. Here's how it works."

Action Summary:

A year ago, most marketers weren't thinking about AI search. Now it's one of the fastest moving channels in the industry and nobody has a playbook yet.

So we built one. We surveyed hundreds of marketers to find out how they're approaching answer engine optimization, where they're investing, what's actually working, and what isn't.

The result is the 2026 State of AEO Report. Real data. Real strategies. A clear picture of where AI search is headed and how to get ahead of it.

Download the free report

First redesign in 25 years. The new bar takes longer conversational queries and pulls shopping results inline. Feed quality is now the visibility lever, not blue-link SEO.

via Adweek CPG/Grocery • Read more

Abandoned cart, checkout, and browse SMS flows now live inside Shopify Messaging with pre-built templates and a spending threshold. Klaviyo and Postscript just lost their easiest upsell.

via Shopify Changelog • Read more

Google and Samsung's smart glasses will ship co-branded with Warby Parker and Gentle Monster. DTC eyewear is the chosen taste layer for big tech's next wearable, not the other way around.

via Business of Fashion • Read more