Advertise to 40,878 E-Commerce Businesses
Promote your tool, service or upcoming event - Reach Founders
Good morning!
In today’s newsletter,
How a skincare brand with 2 employees beat L'Oréal at trust
How to post on Instagram 4 times a week without burning out
Is a drop model right for your business?
Beehiiv's Summer Release Event: The future of audience growth
While You Were Building
1.ChatGPT opens ad inventory to retailers
2.Walmart, Target, Aldi pull artificial dyes
3.Gap and American Eagle cut forecasts
AI Launch Codes: Claude Opus 4.8 : The most honest AI?
This issue takes 2 minutes to read.
Check out our DTC tool stack here
Let’s dive into it👇
Free webinar
How a skincare brand with 2 employees beat L'Oréal at trust (and how to increase your brand’s AI recommendations)
Are you still burning cash on ads instead of leveraging AI?
Look at your metrics. Are your ad costs creeping up and conversion rates staying persistently low?
Not surprising if you’re still investing in traditional aesthetic branding, hoping a prettier video or punchier headline will solve a trust problem.
It won’t. The market has changed.
This Thursday, we are laying out the architectural blueprint showing how they achieved this:
Free Live Webinar in 3 days👇
How a Skincare Brand With 2 Employees Beat L'Oréal at Trust: Why Verifiable Brands Will Win the Next Decade of Commerce.
Thursday, 4th June | 6pm UK / 1pm ET / 10am PT · Live workshop with Ankur Modi (ex-Meta, Ex-Amazon systems architect, & NASDAQ IPO veteran). 60-minute session
You’ll learn:
How to make your product claims an assets that wins you organic conversions from AI engines
AI infrastructure that makes your brand AI’s default recommendation for the next decade
How a micro-team structures business data to automate trust signals
The step-by-step roadmap to becoming the default AI recommendationin your niche.
⚠️ Do Not Wait For A Replay: This session features Live Store Audits from a former FAANG architect. If you want your technical infrastructure and conversion leaks reviewed live, you must be in the room.
Organic Content
How to post on Instagram 4 times a week without burning out
Brands that post 3-5 times a week grow followers twice as fast as brands that post once or twice. That's from Buffer's analysis of 52 million posts.
Most founders already know they should post more. The problem is lack of time and ideas.
Here's a system that solves both.
One idea, four posts
Think about 1 post idea.
A product benefit. A customer question you answered this week. A mistake you made. Then cut it four ways.
A 60-second Reel.
A carousel breaking it into steps.
A single image quote pulled from the Reel.
A story asking your audience if they've experienced the same thing.
Schedule everything in one sitting. Forget it until next week.
Action summary:
Write down one thing you know that your customer doesn't — a tip, a mistake, a lesson
Cut it into a Reel, a carousel, a single image, and a story poll
Schedule all four before the week starts
Repeat next week with one new idea
Credit: Buffer
Brand Strategy
Is a drop model right for your business?
A drop model means your store opens for a short window, sells a limited amount of product, then closes.
Fluff, an Australian beauty brand, does this four times a year for seven days each time. Three years in, it's working.
But it won't work for every business. Here's the one question that tells you whether it's worth trying.
What does your customer buy after the first purchase?
Fluff's compact is a one-time buy. The refills are not. So when the store is closed, refill orders keep revenue coming in.
If your product has no natural follow-on purchase, a drop model leaves you with four revenue events a year and nothing in between.
What a follow-on purchase looks like across different businesses
It doesn't have to be a literal refill. Coffee, supplements, candles, skincare, pet food — anything that runs out and gets reordered works. A subscription tier works. A lower-priced consumable that pairs with your hero product works.
What doesn't work: one-time purchases with no logical next buy. Furniture, most apparel, custom products.
Why a drop model is worth exploring
Scarcity creates urgency in a way that a permanent storefront never does. When customers know the window closes in seven days, they don't wait.
Four focused selling events a year is operationally simpler than managing a store 365 days. Less time selling means more time building the brand.
Action summary:
Write down what your customer buys from you after their first purchase
If nothing comes to mind, a drop model will hurt your cash flow more than help it
If something does, map your production lead time and see how many drops per year are actually realistic
Credit: Shopify Masters
Free AI Guide
100+ ChatGPT Prompts to Revolutionize Your Day by HubSpot
Discover how you can leverage ChatGPT to boost efficiency, streamline tasks, and stay ahead in your industry. Supercharge your productivity with HubSpot's comprehensive guide.
While You Were Building
1.ChatGPT opens ad inventory to retailers
OpenAI partnered with Skai to route commerce advertisers into ChatGPT, following its earlier Criteo deal. Paid placement inside the chat surface now runs through ad stacks brands already use.
via Digiday • Read more
2.Walmart, Target, Aldi pull artificial dyes
The big three grocers are stripping synthetic colors from cereal and frozen aisles, opening shelf space for natural-dye CPG startups that were previously stuck in specialty.
via Modern Retail • Read more
3.Gap and American Eagle cut forecasts
Both apparel retailers slashed full-year guidance, citing curbed discretionary spending. Two simultaneous downward guides from mass apparel mark a category demand-side shift.
via Business of Fashion • Read more
Claude Opus 4.8 : The most honest AI?
Your AI has been confidently wrong, and you had no way to tell.
It handed you a clean, certain answer. You acted on it.
And then later you found out it was wrong after it already mattered.
Opus 4.8 just shipped, and the headline is honesty. It now flags when it isn't sure instead of bluffing through it.
Opus 4.8
On May 28, Anthropic announced Claude Opus 4.8. Their new flagship, and the most capable model the public can use right now. Same price as 4.7, and just 41 days after it.
And like every model release, we got a benchmark report.
Yes, it's better. It leads Opus 4.7, GPT-5.5, and Gemini 3.1 Pro on almost every test. On knowledge work, the row that matters if you run a store instead of a codebase, it isn't close: 1890 against GPT-5.5's 1769.
But the benchmarks aren't what makes this release worth your time.
What makes Opus 4.8 different is honesty.
Anthropic trains its models to be honest, meaning, in their words, to "avoid making claims they can't support."
But every model before this one had the same flaw. It jumped to conclusions, and it sounded exactly as confident guessing as it did when it was right. That confidence is what got you to ship the wrong answer.
So Anthropic measured how often each model slips into misaligned behavior. Things like deception, or just telling you what you want to hear. Lower is better.
Opus 4.8 is 4 times less likely than 4.7 to let a flaw in its own work slip by without flagging it.
This means fewer clean-looking answers that are quietly wrong. When it isn't sure, it tells you. When it is sure, you have more reason to trust it.
What launched alongside the model
Effort control
There's a new control in Claude now, right next to where you pick the model. It lets you set how much effort Claude puts into an answer.
Turn it up and Claude thinks longer and harder before it responds. Better answers on the tough stuff, but slower, and it burns through your usage limits faster. Turn it down and it answers quickly and barely touches them.
It runs across five settings: low, medium, high, extra, and max. Opus 4.8 sits on high by default, which Anthropic calls the best balance of quality and speed.
And it's on every plan, including the one you're already paying for.
Here's the catch.
Opus 4.8 is the best model you can use right now.
But that doesn't mean you should use it for everything. And it doesn't mean every task needs max effort.
This is where most people get it wrong. I get emails almost every week from founders saying the same thing. They keep running out of Claude. They hit the limit, then sit locked out for hours, usually right when they need it most.
The reason is almost always the same. They run every task on the flagship model, at high effort, in a chat bloated with context it never needed. A one-line email gets the same firepower as a financial model.
Most tasks don't need deep reasoning. Most tasks don't need the flagship. When you spend that power on low-value work, you burn your limit on the stuff that didn't need it, and you've got nothing left for the work that does.
Today we break that down, task by task, so you stop hitting the wall.
→ The full breakdown, which model and effort level to use for each task, and how to feed Claude lean context, is below for AI Launch Codes subscribers.
Here is the framework to effectively use Claude
Step 1: Selecting the model
This post is for paid subscribers
We built AI Launch Codes to help you scale smarter. We test AI tools, build workflows, and write prompts—so you don’t have to. But testing takes time, money, and effort. This paid upgrade helps us keep experimenting, while you get the winning playbooks delivered straight to your inbox.
UpgradeWhy You’ll Want It:
- Use AI to scale faster, cut costs, and boost efficiency
- Ready to use workflows, automations and prompts
- Real world use cases, no theory