Big week: OpenAI unveiled GPT-6 Astra, its most capable model yet, with a major focus on agentic work, professional tasks, software engineering, and computer use.

But the bigger story is not just that the models are getting smarter.

AI is getting better at doing the work itself… navigating tools, handling longer workflows, running locally on powerful PCs, and staying active across more complex tasks.

At the same time, the infrastructure underneath AI is shifting fast. Nvidia is pushing harder into local AI on Windows and just acquired Hugging Face in a 12.9 billion deal, putting one of the biggest open-model platforms under the same roof as the company powering much of the AI boom.

Here’s what else is packed into this edition…

Skills to Pay the Bills: Build a Referral Engine Blueprint prompt that turns happy customers into a repeatable new-business machine without begging, awkward asks, or discount coupons no one uses.

Top AI News: The top fresh stories from the week covering important news for business owners.

Let’s jump right in.

Here’s a stat that will annoy you…

Your best next customer probably knows one of your current customers by name.

The reason you are not booking them?

You never systematically ask for the introduction.

You wait for it.
You hope for it.
You cross your fingers and call it “word of mouth.”

Meanwhile, a competitor with a mediocre product and a strong referral program takes the deal.

Referrals do not need to be awkward.

And you do not need to hand out random gift cards to earn them.

What you need is a system:

• a clear moment to ask
• a message that makes it easy to forward
• a small thank you that resonates with people
• a simple way to track who introduced whom

You do not need a growth agency to build this.

You need a smart second brain that can look at your offer, your best customers, and your delivery workflow, then turn it into a referral engine you can actually run this quarter.

Paste this into ChatGPT, Claude, Grok, Gemini, or your AI tool of choice.

Feed it your offer, a snapshot of your happiest customer, and where they live inside your process.

Get back a referral program you can turn on before Friday.

You are the Referral Engine Blueprint, a strategic growth partner for small business owners who have loyal customers but no systematic way to turn those relationships into new business.

Your job is to take my offer, a description of my ideal customer, my typical delivery process, and any current referral activity, then turn it into a clear, low-effort referral system I can execute in the next 30 days.

Work like a smart operator who has built ten referral programs across service businesses, agencies, and product companies.

Push back when my current referral approach is passive, awkward, or built around discounts my customers do not care about.

Tell me the truth about what my customers are likely to say yes to and what will just create silence.

First, orient yourself.

Ask me:

• what I sell
• my typical deal size
• what a wildly happy customer looks like
• the exact moment in my process where they usually say something like “this is amazing” or “I love working with you”

Do not proceed until I share these.

Then analyze in four passes:

1. Trigger Moment
Identify the one point in my customer’s experience where they feel the most value and are most likely to help me. That moment is where the ask lives.

2. Ask Language
Draft the exact words I should use in the trigger moment, spoken and in writing, so the request feels natural instead of transactional.

3. Forward-Ready Asset
Write a short, forwardable message my customer can copy into a text, email, or DM to a friend without editing it. It should sound like them, not like marketing.

4. Thank You Loop
Design a simple thank you that acknowledges the referral before the deal even closes, so my customer feels rewarded for the introduction, not just the outcome.

Then recommend a lightweight tracking system I can run inside a spreadsheet, CRM, or Notion doc.

Include:

• three fields I need
• three fields I should skip
• how often I should review it

Suggest one paid incentive and one non-paid incentive.

Then tell me which one my specific customer profile is more likely to respond to and why.

Close with a 30-day rollout plan:

• which customer to ask first
• what to say
• what to send
• how to follow up
• how to review results at day 30

Do not recommend a generic “10% off your next order” incentive unless my customer profile cares about discounts.

Do not build a program that requires an app, portal, or login for the customer.

It has to work over text and email.

Do not assume I have a large marketing team.

Assume everything is done by me or one part-time helper.

If my inputs are too vague to give confident advice, ask exactly one clarifying question at the end of your reply.

Section 1: What I Heard
Give me three bullets that prove you understood my business.

Section 2: The Four Passes
Break down Trigger Moment, Ask Language, Forward-Ready Asset, and Thank You Loop.

Section 3: Tracking Setup
Give me the simplest way to track referrals without overcomplicating it.

Section 4: Incentive Recommendation
Recommend one paid and one non-paid incentive, then choose the better fit.

Section 5: 30-Day Rollout Plan
Give me a week-by-week plan to launch the referral system.

To start, tell me what you sell, your typical deal size, what a wildly happy customer looks like, and the exact moment in your process where they usually say something enthusiastic.

Open ChatGPT, Claude, Grok, Gemini, or your preferred AI tool.

Copy the full prompt above.

Before you paste it in, gather:

• what you sell
• your typical deal size
• your best customer type
• a few examples of happy customers
• when customers usually feel the most value
• any referrals you have received before
• how customers usually communicate with you
• whether you prefer paid or non-paid thank-yous

Paste the prompt into your AI tool.

Answer the questions.

Then review the referral system it creates.

Look closely at:

• the exact moment it tells you to ask
• the words it gives you to use
• the forward-ready message
• the thank you loop
• the tracking fields
• the 30-day rollout plan

Then pick your first five customers.

Send the ask.

Track the results.

Improve the wording.

Repeat.

That is how word of mouth becomes a system instead of a lucky accident.

Run this once, then rerun it every time you add a new offer or change your delivery process.

Referral triggers are not permanent.

They move as your business grows.

The owners who quietly refresh their referral program twice a year almost always beat the ones who set it once and forget it.

And remember…

A referral ask should not feel like begging.

It should feel like making it easy for happy customers to help the people they already care about.

Here’s Why Experts Say to Ditch Your Old Cutting Boards

Before you prep another meal, here’s what you need to know:

Here’s why over 50,000 pros and home chefs have already switched to TIBO titanium.

Get 40% OFF today.

OpenAI unveiled GPT-6 Astra, describing it as a major step forward for agentic work, professional tasks, software engineering, scientific research, and computer use. The company says Astra can complete complex tasks using fewer total tokens in some scenarios, although API pricing starts at 10 bucks per million input tokens and 50 bucks per million output tokens. OpenAI has also added stronger cybersecurity safeguards after Astra reached the company’s “critical” threshold for cyber capabilities. One challenge remains: OpenAI says Astra’s internal reasoning is harder to monitor than GPT-5.6 Sol’s, making model oversight an increasingly important research priority.

In other news…

NVIDIA has reportedly agreed to acquire Hugging Face for 12.9 billion, according to The Information as reported by Reuters. Hugging Face is one of the most important hubs for open-source AI models, tools, and developer workflows, so this is a huge move for the AI ecosystem. For small businesses, the practical takeaway is simple: the tools your team uses tomorrow may be shaped by whoever controls the model marketplace today. If you rely on open models, custom AI workflows, or developer tools, keep an eye on how access, pricing, and integrations change after this deal.

Vizard Agent can turn scripts, uploads, raw footage, or ideas into finished videos with captions, structure, and editing decisions handled by AI For SMBs trying to publish more video. Video is still one of the hardest content formats to produce consistently, and tools like this reduce the blank-page problem. The smart move is to use it for drafts and repeatable content, then add a human review before publishing anything brand-sensitive.

BetterClaw positions itself as a no-code AI agent builder that can help users deploy agents quickly. For small businesses, the appeal is obvious: build simple agents without needing a developer for every workflow. The use case is starting with one repeatable process, like lead follow-up, customer routing, internal research, or status updates, and seeing what AI can handle safely. Start small, review everything, and expand only when the workflow proves itself.

A big shoutout to Jacquie for trusting us on her journey!

Until next week,