You ask Claude Code to research five competitors. It does them one by one. Twenty minutes later, you’re still waiting for competitor number three. Meanwhile, your context window is stuffed with research you haven’t even used yet.

Today, you’ll build an AI Research Squadron that runs multiple research tasks at the same time, in separate contexts, and returns clean results to your main session.

• ✅ A single-agent delegation prompt that keeps research out of your main context

• ✅ A parallel dispatch pattern that runs 3-5 research tasks simultaneously

• ✅ A research brief template that gets better results from every subagent

• ✅ A result synthesis prompt that turns raw research into actionable insights

• ✅ A reusable CLAUDE.md workflow that makes parallel research your default mode

Let’s build it 👇

🧠 How the AI Research Squadron Works

Claude Code can spawn subagents. Each one is a separate Claude session with its own context window. It does the work independently, then sends back just the result.

Here’s why this changes everything:

• 🧹 Clean main session: All the research noise stays in the subagent’s context, not yours

• ⚡ Parallel speed: 5 research tasks run at the same time instead of one after another

• 📦 Structured output: Each subagent returns exactly what you asked for, in the format you specified

• 🧠 More room for decisions: Your main context stays reserved for thinking, planning, and building

What used to take 30 minutes of sequential research now takes 3-5 minutes of parallel work. And your main session stays focused.

🔎 Prompt #1 → The Solo Scout (Your First Subagent Delegation)

Before running a full squadron, start with one. This teaches you the basic delegation pattern that everything else builds on.

The goal:

• Offload a single research task to a subagent

• Keep the research out of your main context window

• Get back a clean, structured summary you can actually use

✅ Use this whenever you need to research something before making a decision or writing content.

Prompt:

Use a subagent to research [TOPIC OR COMPANY OR QUESTION].

The subagent should:
1. Search the web for the most recent and relevant information (last 30 days preferred)
2. Focus specifically on: [WHAT YOU CARE ABOUT, e.g., "their pricing model, target audience, and key features"]
3. Ignore: [WHAT YOU DON’T NEED, e.g., "company history, founding story, investor details"]

Return a structured summary in this format:
- **Key findings** (3-5 bullet points, most important first)
- **Relevant details** (anything that directly affects my decision or task)
- **Source quality** (high/medium/low confidence based on what was found)

Keep the summary under 200 words. No filler, no context-setting. Just the findings.

💡 Tip: The magic is in the constraints. Tell the subagent what to ignore, not just what to find. This prevents it from returning a 500-word essay when you needed 5 bullet points. Shorter briefs = faster results = less clutter when you use the output.

⚡ Prompt #2 → The Parallel Dispatcher (Run 5 Tasks at Once)

This is where the real speed comes in. Instead of researching one thing at a time, you launch multiple subagents simultaneously. They all work in parallel and return results independently.

The goal:

• Launch 3-5 subagents in a single message

• Give each one a specific, independent research task

• Get all results back without any of the research cluttering your main session

✅ Use this for competitor analysis, market research, content research, or any task where you need information on multiple topics.

Prompt:

Run these [NUMBER] research tasks in parallel using separate subagents. Each subagent works independently.

**Subagent 1 — [LABEL, e.g., "Competitor A"]:**
Research [SPECIFIC TOPIC]. Focus on [WHAT MATTERS]. Return: [FORMAT, e.g., "5 bullet points covering pricing, features, and target market"].

**Subagent 2 — [LABEL, e.g., "Competitor B"]:**
Research [SPECIFIC TOPIC]. Focus on [WHAT MATTERS]. Return: [FORMAT].

**Subagent 3 — [LABEL, e.g., "Competitor C"]:**
Research [SPECIFIC TOPIC]. Focus on [WHAT MATTERS]. Return: [FORMAT].

**Subagent 4 — [LABEL, e.g., "Market trends"]:**
Research [SPECIFIC TOPIC]. Focus on [WHAT MATTERS]. Return: [FORMAT].

**Subagent 5 — [LABEL, e.g., "Customer sentiment"]:**
Research [SPECIFIC TOPIC]. Focus on [WHAT MATTERS]. Return: [FORMAT].

Rules for all subagents:
- Keep each response under 200 words
- Use bullet points, not paragraphs
- Lead with the most important finding
- Flag anything surprising or unexpected with a ⚠️

After all subagents complete, present their results grouped by label. Do NOT synthesize or combine them yet. Just show me the raw results side by side.

🧠 Tip: The key phrase is “in parallel using separate subagents.” This tells Claude Code to launch them simultaneously, not sequentially. If you don’t say “parallel,” Claude might run them one by one, which defeats the purpose. Be explicit.

🎯 Prompt #3 → The Research Brief (Get Better Results From Every Subagent)

Bad subagent prompts get generic results. Good ones get insights you can act on immediately. The difference is the brief you give each subagent.

The goal:

• Create a reusable research brief template

• Give subagents enough context to make smart decisions about what matters

• Get results that are specific to your business, not generic summaries

✅ Use this to upgrade the quality of every subagent task you delegate, especially for business-critical research.

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Monday AI coffee break.

Anthropic’s new Glasswing model found 10,000+ critical bugs in a single month, OpenAI quietly filed for a $1T IPO, and an internal OpenAI model just disproved an 80-year-old math conjecture that stumped generations of mathematicians.

Grab your coffee!
(3-minute coffee break ☕)

AI News around the world 🌍

🔐 Anthropic’s Project Glasswing scanned 1,000+ open-source projects and flagged 10,000+ critical bugs, overwhelming maintainers’ ability to patch. (continue reading)

💰 OpenAI confidentially filed its S-1 with the SEC, targeting a Q4 2026 listing at an $852B to $1T valuation. (continue reading)

🧮 An OpenAI reasoning model disproved Erdős’s 80-year unit-distance conjecture, with Fields medalist Tim Gowers calling it a milestone in AI mathematics. (continue reading)

🤝 EY and Microsoft launched a $1B alliance scaling Copilot and Microsoft 365 E7 to EY’s 400,000+ workforce. (continue reading)

🎵 Spotify and Universal Music signed a landmark deal letting Premium subscribers pay for an AI add-on to remix UMG songs with revenue shared back to artists. (continue reading)

Trending AI Tools 🛠

🤖 NoimosAI: An all-in-one autonomous AI marketing platform for SEO/GEO, social media, growth strategy, competitor intelligence, multi-channel marketing workflows, and more — helping individuals and companies scale marketing operations from one unified system. 👉 Start your 7-day free trial

🖥️ Lapu AI: An AI agent that automates cross-app workflows locally, organizing files, running terminal commands, and chaining operations.

💬 AgentKit: Builds custom AI chat agents trained on your website URLs, PDFs, and docs, deployable as a branded widget on WordPress, Shopify, or any site.

📊 Sellerview: An Amazon FBA profit tracker with SKU-level P&L, combining TACoS, ACoS, margins, and operational alerts in one dashboard for sellers.

Tutorials 📝

Career & Capital 💼

AI Job Board

• 💻 CLERA: Senior Full Stack AI Engineer (New York, NY) | $140,000 to $205,000 (view job)

• ⚙ Ledgent Technology: AI Automation Engineer / Applied AI Solutions Architect (Remote) | $110,000 to $145,000 (view job)

• 🧬 CLERA: Medical AI Researcher (San Francisco, CA) | $150,000 to $230,000 (view job)

Investments in AI

• 🤖 Hark: Raised $700M Series A at a $6B valuation (Parkway-led, with Nvidia, AMD, Intel, Qualcomm) to build a “universal” agentic AI interface and AI-native hardware. (link)

• ⚡ NextEra Energy: Announced a ~$67B all-stock acquisition of Dominion Energy, the largest US utility merger ever, betting on powering the AI data center buildout. (link)

• 💸 Anthropic: Took a fresh $5B investment from Microsoft, with the two now in talks to run Claude inference on Microsoft’s custom Maia 200 AI chips. (link)

Prompt of the Week 🧠

Strategic Bets Prioritization → Force-rank every initiative on your plate and kill the ones quietly eating your runway.

Most founders and operators run 8-12 active bets at once, half inherited and half hopeful, and never sit down to score them honestly. This prompt turns a messy list of projects into a ranked matrix with a clear keep / pause / kill verdict so you can free up team capacity before the quarter slips away.

Prompt:

You are my strategic operations partner. I'm going to paste a list of every active project, initiative, or "bet" my team is currently working on. For each one, do the following:

1. Score it 1-10 on four dimensions:
   - Revenue impact (next 90 days)
   - Strategic moat (does it compound or expire?)
   - Cost to maintain (time, headcount, cash) — inverted, so 10 = cheap
   - Conviction (evidence we have that it works, not hope)

2. Calculate a weighted total (revenue x 0.35 + moat x 0.30 + cost x 0.20 + conviction x 0.15) and rank from highest to lowest.

3. Bucket every initiative into one of three verdicts:
   - DOUBLE DOWN (top quartile, where to add resources)
   - HOLD (middle, keep running but don't invest more)
   - KILL OR PAUSE (bottom quartile, explicitly state what gets reclaimed: hours, headcount, budget)

4. End with: (a) the 3 highest-leverage moves I should make this week based on the ranking, and (b) the single bet I'm most likely emotionally attached to that the data says I should kill.

Here is the list of active bets:
[PASTE YOUR LIST — one per line, ideally with 1-2 sentences on goal + current status + who owns it]

Where to use it: Perfect for founders, COOs, and department heads heading into a quarter close, a strategy offsite, or any moment when the team feels stretched thin.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

💼 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 99,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

YouTube is the highest-leverage content platform for founders in 2026. The problem? Most people quit after 10 videos because they’re guessing at niche, titles, and packaging.

Today, you’ll build an AI YouTube Growth Engine that designs your channel, packages your videos like top creators, scripts them, and tells you exactly how many to post to hit 100K views in 90 days.

• ✅ A locked-in niche based on your unfair advantage + audience hunger

• ✅ A channel positioning doc (target viewer, value promise, channel personality)

• ✅ A library of 20 packaged title + thumbnail concepts using proven hook formulas

• ✅ A bank of 30 packaged video ideas mapped to your content pillars

• ✅ A repeatable video script generator (hook, cold open, body, CTA)

• ✅ A 90-day publishing calendar with growth math attached

Let’s build it 👇

🧠 How the AI YouTube Growth Engine Works

YouTube is a packaging game. The best creators don’t have better content. They have sharper titles, cleaner thumbnails, stronger cold opens, and consistent volume.

This system chains 6 prompts that move from strategy to execution:

• 🎯 Strategy: Niche → channel positioning → packaging library

• 📦 Content: 30 video ideas → repeatable script generator

• 📅 Distribution: Publishing rhythm + 100K-view growth math

Old way: 6 months of trial-and-error and maybe 5K views. AI way: 60 minutes of setup and a 90-day plan with the math to hit 100K.

🔎 Prompt #1 → The Niche Sniper (Find Your Unfair Advantage)

YouTube niches aren’t won by the best creator. They’re won by whoever shows up consistently with a sharp angle no one else owns. This prompt finds yours.

✅ Use this to stop hopping between video topics and commit to a focused channel.

Prompt:

You are a YouTube growth strategist with 10 years experience helping creators go from 0 to 100K subscribers in their first year. You've launched 50+ channels across business, marketing, finance, and tech niches.

I want you to find the YouTube niche where I have the highest unfair advantage and the audience has the most hunger.

Here's my context:
- My business / professional background: [DESCRIBE YOUR BUSINESS, JOB, AND 2-3 KEY SKILLS]
- My current audience (newsletter, social, customers): [LIST CHANNELS AND ROUGH SIZE]
- What I genuinely enjoy talking about for hours: [LIST 3-5 TOPICS]
- My monetization goal in 12 months: [E.G., LEAD GEN FOR MY SAAS, COURSE SALES, AD REVENUE, SPONSORSHIPS]
- My time commitment per week: [E.G., 4 HOURS, 8 HOURS, 15 HOURS]

Do the following:

1. Generate 5 candidate YouTube niches at the intersection of my skills + interests + business goal.
2. For each candidate, score on 4 criteria (1-10):
   - Audience demand (search volume, view counts on similar channels)
   - Competition gap (under-served angles in this niche)
   - My unfair advantage (proof I can speak credibly here)
   - Monetization fit (does this niche map to my 12-month goal?)
3. For the top-scoring niche, identify a sub-niche I can dominate in 6 months. Be specific. Not "AI" but "AI tools for B2B SaaS founders."
4. Write a 1-sentence channel tagline that captures the sub-niche.
5. List 3 channels in this sub-niche I should study as benchmarks.

Output format:
- Niche scorecard table
- Winning sub-niche + reasoning
- Channel tagline
- 3 benchmark channels with rough subscriber counts and what each does well

💡 Tip: Be brutally honest about your unfair advantage. If you’ve never built a SaaS, don’t pick “AI for SaaS founders” because it sounds monetizable. Pick the niche where you have proof you can talk for 50 videos without faking it.

🧱 Prompt #2 → The Channel Architect (Lock Your Positioning)

Most channels feel generic because they have no clear viewer in mind. This prompt locks who you’re for, what you promise, and what makes you stick.

✅ Use this every time you doubt whether to make a video. Does it serve THIS viewer?

Prompt:

You are a YouTube channel positioning expert. You've helped 100+ creators define the strategic foundation of their channel.

Using my winning sub-niche and tagline from Prompt #1:
[PASTE OUTPUT FROM PROMPT #1]

Build my full channel positioning document.

Do the following:

1. Define my target viewer in 1 painfully specific sentence. Include role, life situation, and biggest current frustration.
2. Write my channel's value promise: what viewers walk away with after watching any video on this channel.
3. Define 3 channel personality traits. Pick traits that distinguish me from the top 10 channels in this sub-niche.
4. Write 5 anti-positioning statements ("This channel is NOT for...") to sharpen who I'm not serving.
5. Write banner copy: 6-8 words max, scannable at thumbnail size.
6. Write the About section: 100 words max, hooks the right viewer in line 1.
7. Suggest 3-5 content pillars (recurring video categories this channel will own).

Output as a clean 1-page positioning doc I can paste into Notion or Google Docs.

🧠 Tip: The “NOT for” list is the secret weapon. The more you exclude, the sharper the channel feels. Don’t be afraid to alienate 90% of YouTube to win the 10% who are perfect for you.

🎯 Prompt #3 → The Packaging Lab (20 Title + Thumbnail Concepts)

Packaging is everything on YouTube. Same content, better title and thumbnail, 10x the views. This prompt generates 20 packaged concepts using formulas top creators use.

✅ Use this as your title bank. Pull from here whenever you start a new video.

Prompt:

Read more

🚀 Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Thursday AI coffee break.

This week, Google dropped Gemini 3.5 Flash at I/O 2026, Andrej Karpathy jumped ship to Anthropic, and a California jury dismissed Elon Musk‘s $150B lawsuit against Sam Altman in under two hours.

Grab your coffee!
(3-minute coffee break ☕)

*together with:

NoimosAI: An all-in-one autonomous AI marketing platform that helps individuals and companies scale SEO/GEO, social media, growth strategy, competitor intelligence, and multi-channel marketing workflows from one unified system.

Simply define your goals, and specialized AI agents handle research, planning, execution, and optimization continuously in the background → helping users scale marketing operations more efficiently over time.

👉 Start your 7-day free trial

AI News around the world 🌍

🧠 Google launched Gemini 3.5 Flash at I/O 2026, delivering frontier-level reasoning at four times the speed of competing models. (continue reading)

🤝 OpenAI cofounder Andrej Karpathy joined Anthropic’s pre-training team, calling the next few years at the LLM frontier “especially formative.” (continue reading)

⚖️ A California jury dismissed Elon Musk’s $150B lawsuit against Sam Altman in under two hours, citing the statute of limitations. (continue reading)

💰 Google and Blackstone are launching a $5B AI cloud venture with 500MW of TPU capacity by 2027, targeting CoreWeave and Nvidia. (continue reading)

🔌 Anthropic acquired API toolkit maker Stainless to boost Claude’s agent connectivity to enterprise software and third-party services. (continue reading)

Trending AI Tools 🛠

🤖 Cuey: Chrome extension that cross-checks ChatGPT, Claude & Gemini in one tab. Catch hallucinations before they burn you. Get 3 months of Pro with code 👉 AIBREAK.

🔍 Sight AI: All-in-one SEO and AI-search visibility platform that tracks how your brand surfaces on ChatGPT, Claude, and Gemini.

💳 Plouton AI: Browser-based AI agents for finance ops that run accounts payable, reconciliations, and month-end close inside SAP, NetSuite, and QuickBooks.

🗂 Raydian: AI project planner that converts plain-text requirements into actionable roadmaps, Kanban tasks, and dependency maps with real-time bottleneck detection.

Tutorials 📝

Career & Capital 💼

AI Job Board

🧠 Apple: Senior Machine Learning Engineer, Proactive (Apple Intelligence Data Platform) - $139,500 to $258,100 (view job)

🎯 Exodigo: Senior Product Manager, AI Design Automation (Civil Infrastructure) - $185,000 to $210,000 (view job)

💻 Tata Consultancy Services: Gen AI Developer - $110,000 to $120,000 (view job)

Investments in AI

🧠 Recursive Superintelligence: The UK AI lab founded by ex-OpenAI, DeepMind, Meta AI, and Salesforce AI leaders raised $650M at a $4.65B valuation, with GV and Greycroft leading the round and Nvidia plus AMD Ventures joining. (link)

⚡ NextEra Energy: Announced a $67B all-stock acquisition of Dominion Energy on May 18, the largest energy deal since Exxon-Mobil in 1998, aimed at capturing the AI data center power surge. (link)

🚀 Armada: The modular AI data center startup closed a $230M Series B at a $2B valuation co-led by Overmatch, BlackRock, and 8090 Industries to deploy edge AI compute into off-grid environments. (link)

Prompt of the Week 🧠

Win-Back Campaign → Resurrect dormant customers with the right offer, segment by segment, in 20 minutes.

Most businesses sit on a graveyard of lapsed customers and never touch them, because segmenting feels messy and cold-emailing feels awkward. This prompt turns a raw churn export into a 3-segment win-back plan with copy, offers, and a send order you can ship the same day. Use it the week you need a revenue bump without burning new ad spend.

Prompt:

You are my retention strategist. I'm pasting a list of churned or dormant customers below. Each row has: name, last purchase or last active date, total spend, product/plan they had, and (if available) the reason they left or last touchpoint.

Do this in order:

1. Segment the list into exactly 3 buckets based on win-back potential: HOT (recent churn, high spend, no negative signal), WARM (mid-recency or partial signal), COLD (long-gone or had a bad experience). Show counts per bucket and the logic you used.

2. For each bucket, recommend ONE specific win-back offer (e.g., comeback discount, free upgrade, white-glove onboarding call, founder check-in email). Pick what fits the bucket, not the same offer for all.

3. Write a short email (under 120 words) for each bucket. Subject line, preview text, body, CTA. Sound human, not "we miss you."

4. Tell me the order to send them in (which bucket first, why) and one metric to watch per bucket to know if it's working.

Customer data:
{paste churn export here}

My business: {1-sentence what you sell + price point}

Where to use it: Founders with a CRM full of dormant accounts, SaaS operators after a churn spike, ecom brands sitting on lapsed buyers, agencies wanting to reactivate past clients without launching a new campaign from scratch.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

💼 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 99,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

You drop Claude Code into a project, and it has no idea who you are, what you sell, or how you write.

Every session starts from zero, and you waste 15 minutes re-explaining your business before any real work happens.

Today, you’ll build an AI Context Foundation that loads your business permanently into Claude Code. One file. Five prompts. It survives across every session.

• ✅ A structured Business Snapshot that captures who you are, what you sell, and how you operate

• ✅ A clean, copy-paste CLAUDE.md file tailored to your role (founder, creator, or agency)

• ✅ A reference folder plan that pulls past content, brand voice, and key decisions into context

• ✅ Three starter slash commands wired to your real workflows

• ✅ A smoke test that proves Claude actually knows your business

• ✅ A quarterly refresh prompt so your context never goes stale

Let’s build it 👇

🧠 How the AI Context Foundation Works

The whole system runs on one file Claude Code, which reads automatically every time you open a project: CLAUDE.md.

• 📍 Where it lives: at the root of your project folder

• 🔄 When it loads: every single session, automatically, no setup

• 🧩 What it does: tells Claude who you are, what you’re working on, and how you want it to behave

• 🪜 How we’ll build it: interview yourself, generate the file, add reference docs, seed slash commands, then stress-test it

• ⏱️ How long: about 10 minutes of focused work, then permanent payoff

🧱 Prompt #1 → The Business Snapshot Generator (Interview Yourself in Plain English)

Before we write a single line of CLAUDE.md, Claude needs to actually understand your business. This prompt runs a structured interview, then turns your answers into a clean profile.

Goals for this step:

• Capture what you sell, who you sell it to, and how you make money

• Surface your tone, your no-go zones, and your operating constraints

• Produce one tidy “Business Snapshot” you’ll feed into every later prompt

✅ Use this to create a single source of truth about your business that Claude can reference forever.

Prompt:

You are a senior business analyst who specializes in extracting clear, structured business profiles from founders and operators in under 10 minutes.

I want you to interview me about my business so we can build a permanent context file for Claude Code. Ask me ONE question at a time. Wait for my answer before moving to the next.

Cover these areas in order:
1. Identity: my name, role, company name, and one-line description of what we do
2. Offer: what I sell, price points, and how customers buy
3. Audience: who my ideal customer is in plain language
4. Channels: where my customers come from (newsletter, social, ads, referrals)
5. Voice: 3 words that describe how I write or speak to customers
6. No-gos: topics, claims, or styles I never want Claude to produce
7. Tools and stack: the main tools I use day-to-day
8. Current focus: the #1 thing I'm working on this quarter

After I answer all 8 questions, output a clean "Business Snapshot" in this exact structure:

# Business Snapshot
## Who I Am
## What I Sell
## Who I Sell To
## Where Customers Come From
## Voice and Tone
## Hard No-Gos
## Stack
## This Quarter's Focus

Keep each section to 2-4 short bullets. No fluff. Save it as my single source of truth.

My role is: [FOUNDER / CONTENT CREATOR / AGENCY OPERATOR / OTHER]
My industry is: [YOUR INDUSTRY]

Start with question 1.

💡 Tip: Answer like you’re texting a smart friend, not writing a pitch deck. Short, honest, specific. The messier and more real your answers, the sharper the snapshot.

📐 Prompt #2 → The CLAUDE.md Architect (Turn Your Snapshot Into the Real File)

Now we convert that Business Snapshot into an actual CLAUDE.md file Claude Code will auto-load. This is the core artifact of the whole system.

Goals for this step:

• Generate a clean, well-structured CLAUDE.md with the right sections

• Match the file to your role (founder, creator, or agency)

• Keep it under 200 lines so Claude reads it fast every session

✅ Use this to produce the actual file that goes at the root of your project folder.

Prompt:

You are a Claude Code power user who has set up dozens of CLAUDE.md files for solo operators and small teams. You know exactly what context Claude needs to behave like a true team member instead of a generic assistant.

Here is my Business Snapshot:

[paste output from Prompt #1]

Write a complete CLAUDE.md file for me. Use this structure:

1. # About This Project (1 paragraph: what this folder is, why Claude is here)
2. ## Who I Am (3-5 bullets from my snapshot)
3. ## What I Sell (offers, prices, audience)
4. ## My Voice (do's and don'ts, with 2-3 example phrases I'd actually use)
5. ## Hard Rules (things Claude must never do, in bullet form)
6. ## Default Behavior (when I ask for X, do Y)
7. ## Working Style (how I want Claude to communicate: brevity, format, when to ask vs. assume)
8. ## Reference Files (placeholder list, we'll fill it in next)

Rules:
- Keep the whole file under 200 lines
- Use second person where Claude is the subject ("You write in...", "You never...")
- Be specific, not generic. "Punchy and casual" beats "professional"
- End with a one-line reminder: "When in doubt, ask. Don't guess about the business."

Output the full markdown ready to paste into a file called CLAUDE.md.

💡 Tip: Don’t accept generic output. If a section reads like it could belong to any business, push back: “Rewrite section 4 with examples specific to my snapshot.” Claude will tighten fast.

Here are 3 real-feeling templates to anchor what good output looks like. Pick the one closest to your role and use it as a sanity check against what Claude generates.

—

Template A: Solo Founder (SaaS / digital product)

# About This Project
This is the working folder for Acme Analytics, a $39/mo SaaS that helps Shopify stores spot churn 30 days before it happens. Claude Code helps me ship features, write copy, and run small ops tasks.

## Who I Am
- Sara, solo technical founder
- 4 years building, 1 year on Acme
- Background: data engineering at a Series B retailer
- I work alone but ship like a team of 5

## What I Sell
- Acme Analytics: $39/mo, $390/yr
- ICP: Shopify stores doing $50K-$500K MRR
- Customers buy after a 14-day free trial, no card upfront

## My Voice
- Direct, warm, never salesy
- Use "you" and "I", never "we" (I'm solo)
- Examples: "Here's what's actually happening in your data."
- Avoid: "synergy", "leverage", "best-in-class"

## Hard Rules
- Never invent customer quotes or case studies
- Never claim a feature exists if it isn't shipped
- Never write in third person about Acme

## Default Behavior
- When I ask for copy, give me 3 punchy variations, not one polished draft
- When I ask for code, prefer simple Python over fancy abstractions
- When I ask "what should I do", ask 2 clarifying questions first

## Working Style
- Be concise. Bullets over paragraphs.
- Push back when my idea is weak. I'd rather hear "this won't work because..." than a polite yes.
- Always show your reasoning when changing more than 20 lines of code.

## Reference Files
- @docs/customer-interviews.md
- @docs/pricing-decisions.md
- @docs/landing-page-copy.md

When in doubt, ask. Don't guess about the business.

—

Template B: Content Creator (newsletter / course / community)

Read more

🚀 Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Monday AI coffee break.

Big way to start the week: PwC is putting Claude in front of hundreds of thousands of its people, OpenAI quietly wired ChatGPT into your bank account, and Codex just landed on your phone.

Throw in a $200M Anthropic and Gates Foundation bet on AI for good, and you’ve got plenty to chew on.

Grab your coffee! (3-minute coffee break ☕)

* The next wave of AI attacks won’t come through the prompt box. It’ll come through the context.

Gray Swan Arena is paying you to break frontier agents through indirect prompt injection → poisoned pages, hidden markdown, malicious tool outputs, and screenshots. $20K prize pool. ~20 frontier models.

👉 Jump In and Break AI

Closes May 22 @ 1PM ET.

AI News around the world 🌍

🤝 PwC is rolling Claude out to hundreds of thousands of professionals, certifying 30,000 staff and reporting up to 70% delivery gains. (continue reading)

💰 OpenAI launched ChatGPT personal finance tools for US Pro users, connecting 12,000+ institutions for spending analysis and portfolio dashboards. (continue reading)

📱 OpenAI’s Codex coding agent now runs inside the ChatGPT mobile app on iOS and Android, so you can steer tasks from anywhere. (continue reading)

🌍 Anthropic and the Gates Foundation committed $200 million over four years toward AI for global health, education, and economic mobility. (continue reading)

🇲🇹 Malta becomes the first country to offer free ChatGPT Plus to every citizen for a year after they complete an AI literacy course. (continue reading)

Trending AI Tools 🛠

🎨 Magnific: Generate and upscale multimedia creative content, turning low-resolution images into crisp, production-ready visuals for ads, sites, and product shots.

📡 Digg AI: Tracks 1,000 of the most influential AI voices to surface and rank trending AI stories before they break, so founders spot what matters early.

🎬 Saltfish: Turns a single product click-through into interactive demos, AI-narrated videos, and branded assets in minutes, ready to embed anywhere with no code.

☕ Coffee: An AI CRM agent that automates data entry and sales intelligence, building contacts from emails and updating deal stages so reps stop doing busywork.

Tutorials 📝

Career & Capital 💼

AI Job Board

🦜 LangChain: Product Manager, LangSmith - $180,000 to $215,000 (view job)

🧠 Workday: Principal AI Researcher - $190,600 to $342,000 (view job)

🏥 MD Anderson Cancer Center: Senior IT Project Manager, AI - $123,000 to $185,000 (view job)

Investments in AI

🧠 Recursive Superintelligence raised $650M at a $4.65B valuation to build self-improving AI systems, led by GV and Greycroft with Nvidia and AMD joining. (link)

⚡ Fractile pulled in a $220M Series B for in-memory inference chips that attack AI’s cost and speed bottleneck, co-led by Accel, Factorial Funds, and Founders Fund. (link)

📸 Wirestock raised a $23M Series A to supply ethically sourced multimodal training data from 700,000+ creators, led by Nava Ventures with Sheryl Sandberg’s SBVP. (link)

Prompt of the Week 🧠

Cash Flow Stress Test → Find out how many months of runway you really have before you panic-cut the wrong thing.

Most founders only model the optimistic case, then scramble when revenue slips or a big invoice goes late. Run this before your next board meeting, budget cycle, or anytime the market gets shaky so you act on numbers instead of fear.

Prompt:

You are my CFO. I'm going to give you our financials and you'll stress-test our cash position across three scenarios, then tell me exactly what to do.

Here's our situation:
- Cash in bank: [AMOUNT]
- Monthly recurring revenue: [AMOUNT]
- Monthly fixed costs (payroll, rent, software, etc.): [AMOUNT]
- Monthly variable costs: [AMOUNT]
- Largest single customer as % of revenue: [%]
- Outstanding receivables and when they're expected: [LIST]
- Any expected large one-off costs in the next 6 months: [LIST]

Do the following:
1. Calculate my current monthly burn and exact runway in months at today's numbers.
2. Model three scenarios over the next 6 months: (a) Base case, (b) Stress case where revenue drops 25% and my largest customer churns, (c) Severe case where revenue drops 40% and a key receivable is paid 60 days late. Show the month-by-month cash balance for each.
3. Identify the single point where I run out of cash in each scenario, and the earliest warning signal I'd see before it happens.
4. Give me a prioritized cut list: which costs to reduce first, second, and third to extend runway by 3+ months, with the runway impact of each and a flag on anything that would damage revenue or morale.
5. Recommend the 3 leading indicators I should put on a weekly dashboard so I never get surprised again.

Ask me for any missing number before you run the analysis. Be direct. I need the truth, not reassurance.

Where to use it: For founders, finance leads, and operators heading into budget planning, a board meeting, or a fundraising decision.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

💼 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 99,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below, and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

Your happiest customers would refer you in a heartbeat if you ever asked.

You never ask. And when you do, there’s no incentive, no script, no tracking. So nothing happens.

Today, you’ll build an AI Referral Program Architect that designs the incentives, scripts the asks, onboards referrers, and runs a 30-day launch plan you can ship this week.

Here’s what you’ll walk away with:

• ✅ A profile of who in your customer base will actually refer you

• ✅ An incentive structure with unit economics that defends it

• ✅ A 5-moment ask schedule with scripts for each channel

• ✅ A Referrer Toolkit (intro scripts, 1-pager, FAQ) that handles objections

• ✅ A lightweight tracking + nudging system that runs in one Google Sheet

• ✅ A 30-day launch plan with all the copy ready to ship

Let’s build it 👇

🧠 How the Referral Program Architect Works

Most referral programs fail for one reason. They’re built generically when they need to be built for one specific type of customer who’s actually motivated to refer.

This system reverse-engineers your program from your best potential referrer. Get that person right and everything else cascades.

The Architect runs in 6 stages:

• 🔍 Profile the customer most likely to refer (and what they’d say)

• 💰 Design the incentive structure that fits their motivation

• 📣 Script the 5 moments and channels to ask

• 🎁 Onboard referrers with a kit that closes the friend

• 📊 Track + nudge in one sheet, not a dashboard

• 🚀 Launch with a 30-day plan and ready-to-ship copy

Old way: hire a consultant, set up Tolt or Rewardful, write 40 emails, hope it sticks.

AI way: 60 minutes of chained prompts and you’re live by Friday.

🔍 Prompt #1 → The Referrer Profiler (Find Who Will Actually Refer You)

Programs built for “everyone” convert zero. Programs built for one specific persona compound.

The goal of this prompt:

• Surface the 3 most likely referrer personas in your customer base

• Understand each one’s real motivation

• Pick the single highest-leverage persona to design around

✅ Use this to: lock the audience your entire program is built for.

Prompt:

You are a referral program strategist who profiles which customers are most likely to refer and what they would say about a business.

I'm building a referral program. Here are the inputs:

Business: [YOUR BUSINESS]
Product/service: [WHAT YOU SELL]
Target buyer: [YOUR ICP]
Pricing: [PRICE POINT]
Top 3 customer outcomes I've seen: [LIST 3 SPECIFIC WINS]
Examples of organic referrals I've already received (if any): [LIST OR "NONE YET"]
Customer feedback themes (from reviews, NPS, support): [WHAT CUSTOMERS SAY]

Your job:
1. Identify the 3 most likely "referrer personas" in my customer base (e.g., "the early adopter who loves showing off", "the cost-conscious operator", "the time-starved manager who got their evenings back").
2. For each persona, describe their referral motivation (status, money, helping a friend, professional reputation, mix).
3. Predict what each persona would actually say when introducing me to a peer.
4. Score the persona's referral potential 1-10 (how many warm intros they'd realistically make in 90 days).
5. Pick the single highest-leverage persona to design the program around, with reasoning.

Output as a table:
Persona | Motivation | What They'd Say | Referral Potential | Priority
Recommended Persona: [pick one] because [reason]

💡 Tip: Pick the persona with the highest score, not the broadest one. A program that works for one persona will compound. One that tries to please everyone won’t move.

💰 Prompt #2 → The Incentive Designer (Build the Reward That Actually Motivates)

Cash isn’t always the answer. Sometimes it’s the worst answer.

The goal:

• 3 incentive structures tailored to your persona’s motivation

• Unit economics that prove the program won’t bleed money

• A recommendation with landing page copy ready to ship

✅ Use this to: avoid the “give $50 get $50” default that’s killed thousands of B2B referral programs.

Prompt:

You are a behavioral economist who designs referral incentives that actually drive action.

Inputs:
Primary referrer persona: [paste from Prompt #1 — the highest-priority persona]
Their motivation: [paste motivation]
Product price point: [PRICE]
Average customer LTV: [LTV OR ESTIMATE]
Acceptable CAC (customer acquisition cost): [CAC OR ESTIMATE]

Your job:
1. Design 3 incentive structures that fit this persona's motivation. Options to consider: cash, account credit, exclusive access, status reward, charitable donation, double-rewards, milestone-based ladder.
2. Calculate the unit economics for each option (cost per referral vs. expected LTV).
3. For each, design the "double-sided" version (what the referred friend gets + what the referrer gets).
4. Pick the single recommended structure with reasoning.
5. Write the exact incentive copy as it would appear on a landing page: headline, sub, and the "how it works" section in 3 bullets.

Output:
Option | Referrer Reward | Friend Reward | Cost Per Referral | Why It Fits
Recommendation: [option] because [reason]
Landing Page Copy: headline + sub + 3-bullet "how it works"

🧠 Tip: Status and access often beat cash for B2B referrals. Founders, operators, and senior managers refer for reputation, not $25 Amazon gift cards.

📣 Prompt #3 → The Ask Architect (Build the Moments, Channels, and Scripts)

The right ask at the wrong moment converts at zero. The right ask at the right moment converts at 30%+.

The goal:

• The 5 highest-leverage moments to make the ask

• The right channel and tone for each

• Word-for-word scripts you can copy and ship

✅ Use this to: stop guessing when and how to ask, and start running a system.

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Thursday AI coffee break.

This week, Google revealed an entire new laptop category built around Gemini, Meta‘s Superintelligence Labs finally shipped its first model, and SpaceX and Google are quietly plotting to move AI data centers into orbit.

Hardware is suddenly the AI battleground again, and we’ve got the rundown.

Grab your coffee!
(3-minute coffee break ☕)

AI News around the world 🌍

💻 Google unveiled the Googlebook, a new laptop category built around Gemini with a Magic Pointer that gives contextual suggestions as you point at the screen. (continue reading)

🧠 Meta launched Muse Spark, with reasoning, multimodal perception, and shopping baked into Meta apps and smart glasses. (continue reading)

🛰️ Google and SpaceX are in advanced talks to launch orbital data centers, putting AI compute literally into space to ease ground-based energy strain. (continue reading)

🏢 Anthropic rolled out Claude for Small Business with 15 agentic workflows wiring Claude into QuickBooks, PayPal, HubSpot, Canva, and Microsoft 365. (continue reading)

📱 TikTok launched an MCP server that lets AI agents plan, launch, and optimize full ad campaigns autonomously on the platform. (continue reading)

Claude Code for Beginners. Free Masterclass.

We created a full masterclass on Claude Code for beginners. No slides. No theory. Just us opening Claude Code live and showing you exactly how to use it from scratch.

No coding background needed. Over 2,000 people have already signed up.

If you’ve been curious about Claude Code but didn’t know where to start, this is it.

WATCH IT NOW

Trending AI Tools 🛠

🔍 RankSpot: An autonomous SEO agent that researches, writes, and publishes optimized blog articles daily.

🎬 Velo: Turns raw voice and screen recordings into polished, shareable videos in seconds, killing the awkward retake loop for sales demos, async standups, and Loom-style updates.

🧠 Kanwas: An open-source brain for your team that stores product context on a shared canvas, accessible by humans and AI agents alike for better onboarding and decisions.

💳 Kelviq: A unified monetization platform that handles payments, usage-based billing, tax compliance, and merchant-of-record duties.

Tutorials 📝

Career & Capital 💼

AI Job Board

• 💻 Kalepa: Staff Backend Engineer, Core Product (Remote, USA) - $190,000 to $255,000 (view job)

• 📊 BES Consulting Group: Sr Data Scientist, AI/ML Models (Image Analysis) - $50 to $53/hr (view job)

• 🏢 Wells Fargo: Lead Software Engineer - $143,000 to $224,000 (view job)

Investments in AI

• 🤖 Anduril: Raised a $5B Series H led by Thrive Capital and a16z, doubling valuation to $61B to scale autonomous drones, missile systems, and AI command platforms. (link)

• 💰 Moonshot AI: Raised $2B at a $20B valuation in a Meituan-led round, nearly 5x its late-2025 mark as Kimi K2.6 crossed $200M ARR. (link)

• 🚀 Akamai: Locked in a $1.8B, seven-year AI cloud infrastructure deal with a frontier model lab, sending shares up 20% on the news. (link)

Prompt of the Week 🧠

Board Meeting Prep → Turn 3 months of messy company data into a sharp board deck narrative in 30 minutes.

Prompt:

You are my Chief of Staff preparing me for a board meeting in 7 days.

Here is the raw input:
- Last 3 months of company metrics (revenue, growth rate, burn, runway, headcount, key KPIs): [PASTE]
- Wins this quarter: [PASTE]
- Misses and what went wrong: [PASTE]
- Top 3 strategic questions I want the board to weigh in on: [PASTE]
- Last board meeting's commitments and where each one landed: [PASTE]

Produce the following, in this order:

1. THE ONE-LINE NARRATIVE
   The single sentence that frames this quarter. It must be honest, specific, and forward-leaning. No corporate hedging.

2. THE NUMBERS THAT MATTER (5 max)
   Pick only the metrics that either prove the narrative or directly contradict it. For each: number, trend vs last quarter, and one sentence on what it actually means.

3. WINS (3 max, ranked by board relevance)
   For each: what we did, the measurable result, and why it matters to the long-term thesis. Cut anything that sounds like a participation trophy.

4. MISSES + ROOT CAUSE (be brutal)
   For each miss: what happened, the real reason (not the polite reason), what we are changing, and by when. If I gave you a polite version, push back and ask what the actual cause was.

5. ASKS FOR THE BOARD (max 3)
   Frame each as a specific decision or intro you need, not a vague "thoughts?". Include the recommendation and the risk of inaction.

6. ANTICIPATED HARD QUESTIONS (5)
   The 5 questions a sharp board member will ask after reading this. For each, give me the 2-sentence answer I should have ready.

7. THE NARRATIVE ARC FOR THE LIVE MEETING
   In 5 bullets, the order I should walk them through this so the meeting ends with alignment on the asks instead of debating the numbers.

Be concise. Cut anything a tired board member would skim. If any input is thin or missing, flag it and tell me what to gather before the meeting.

Where to use it: Founders and CEOs prepping quarterly board updates, ops leaders building exec readouts, or chiefs of staff pulling together steering-committee decks. Works in ChatGPT, Claude, or any model with a 200K+ context window so you can paste the full data dump.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

💼 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 99,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

A few months ago we ran a single ad. Just one. Asking who wanted help putting AI to work inside their business.

46 founders booked a call.

We weren’t ready for that. The AI Break was supposed to be a newsletter, not the front door of an agency. But after the 10th call, then the 20th, a pattern was impossible to miss.

Everyone had the same problem.

They were "using AI." Paying for it. Posting about it. And their week looked exactly like it did a year ago.

The same admin still got done by hand. The same reports still took hours. The same content still needed a human to sit down and write it.

ChatGPT could draft a LinkedIn post in 30 seconds but someone still had to open the tab, paste the brief, copy the output, fix the tone, and ship it.

Times every tool. Times every workflow

• ChatGPT in one tab

• Claude in another

• A half-finished automation in Make

• A Notion full of prompts nobody used twice

• $300+/month in AI subscriptions

• The same to-do list as six months ago

The tools were everywhere. The leverage wasn't. AI was helping with tasks. Nothing was actually running.

The tools were there. The system wasn't.

So we built one. For ourselves first.

We took the most repetitive parts of our week and handed each one to a Claude Code agent:

• Drafting LinkedIn posts

• Prepping client reports

• Writing this newsletter

• Chasing data across five platforms

Not chatbots. Real workers. Our voice, our data, our stack.

The week got 20 hours lighter. Then 30.

We can help you in two ways:

1. Map out your AI strategy. We look at your workflows, find the spots where AI saves you the most time and money, and hand you a roadmap. If AI isn’t the answer for something, we’ll tell you. That part matters.

2. Build a custom AI Operation System with Claude Code. A small team of agents that handles the repeatable parts of your business:

• ✍️ Content (posts, drafts, scripts)

• 📋 Admin (inbox, scheduling, status updates)

• 📊 Ops & reporting (recurring docs, weekly reports)

• 🎯 Sales & outreach (lead enrichment, follow-ups)

Built in your voice, plugged into your tools, owned by you. No black box. No agency theater. You actually watch it come together inside Claude Code.

Two paths, same first step. A free 30-minute call:

• ✓ No deck, no pressure

• ✓ Honest “no” if AI isn’t the answer

• ✓ You’ll leave with at least one idea you can use

We’ve opened slots for May and June.

→ See how it works

Check it Here

If your week is starting to feel like a stack of small repeated tasks held together by your attention, let’s talk.

Luis & Rui

Share

Your buyers don’t Google anymore. They ask ChatGPT, Claude, or Perplexity “who’s the best [your category]?” and trust the answer.

If you’re not in those answers, you don’t exist.

Today, you’ll build an AI AEO Engine that audits how AI assistants describe you, finds your recommendation gaps, and ships a 90-day plan to win the answer slot.

Here’s what you’ll walk away with:

• ✅ A visibility audit of how ChatGPT, Claude, and Perplexity describe you today

• ✅ The top 5 buyer questions you need to show up in

• ✅ A gap map comparing your surface area vs. your top 3 competitors

• ✅ A 90-day AEO content plan with priority order

• ✅ A complete editorial brief for your highest-leverage piece

• ✅ A monthly scorecard to track AI recommendations over time

Let’s build it 👇

🧠 How the AEO Engine Works

AEO stands for Answer Engine Optimization. It’s the new SEO, but for AI assistants instead of Google.

Old way: rank for keywords, hope someone clicks.
New way: be the named answer when an AI assistant recommends a solution to a buyer’s question.

The Engine runs in 6 stages:

• 🔎 Audit the questions buyers ask AI in your category

• 🧠 Mine the variations and trust signals AI models pull

• 🎯 Map the gap between your current surface and what AI needs

• 🧱 Plan the 90 days of content that closes the gap

• ✍️ Brief the highest-leverage piece down to schema markup

• 📊 Score AI mentions monthly so you know what’s working

Old way: 6 months of SEO consultants and $5K invoices.
AI way: 60 minutes of chained prompts and a quarterly content plan you can run yourself.

🔎 Prompt #1 → The Visibility Auditor (Find Where AI Currently Mentions You)

Before you optimize, you need to know what AI is already saying about you (or not).

The goal of this prompt:

• Surface the 10 questions buyers in your category ask AI

• Predict the kind of answer AI gives today

• Rank the 5 highest-intent questions

✅ Use this to: get a clear baseline of where AI sits today before you spend a minute on content.

Prompt:

You are an AEO (Answer Engine Optimization) specialist who audits how AI assistants represent businesses in their responses.

I am running a visibility audit on my business. Here are the inputs:

Business name: [YOUR BUSINESS]
Category: [YOUR CATEGORY, e.g., "AI consulting in Lisbon", "B2B SaaS for HR teams"]
Target buyer: [YOUR ICP]
Top 3 competitors: [LIST 3]
Website: [URL]

Your job:
1. List the 10 questions a buyer in my category is most likely to ask ChatGPT, Claude, or Perplexity when researching solutions like mine.
2. For each question, predict the type of answer the AI will give (e.g., named recommendations, generic advice, no specific business).
3. Identify which 5 of those 10 questions are most likely to lead a buyer to actually purchase if I show up in the answer.
4. Rank those 5 questions by buyer intent (high to low).

Output as a numbered list with: Question | Likely Answer Type | Buyer Intent (1-10) | Why It Matters

💡 Tip: Run each of the top 5 questions in ChatGPT, Claude, and Perplexity yourself. Screenshot the real-world answers. That’s your starting line.

🧠 Prompt #2 → The Query Miner (Find the Variations AI Actually Answers)

Buyers rarely type the exact same question. AI assistants pull from variations and conversational follow-ups.

The goal:

• Expand each high-intent question into 5 conversational variations

• Surface what info AI needs to recommend a specific business

• Identify the trust signals AI pulls from the open web

✅ Use this to: understand the full surface area you need to win, not just one keyword.

Prompt:

You are an AEO researcher specializing in conversational search behavior.

I have a list of 5 high-intent buyer questions in my category:

[paste the top 5 questions from Prompt #1]

For each of these 5 questions, do the following:
1. Generate 5 conversational variations a real buyer would phrase the question (more specific, with context, with constraints, or as follow-ups).
2. Identify what information the AI needs to recommend a specific business (e.g., named credentials, social proof, location, pricing transparency, integrations).
3. List the 3 signals AI models pull from the open web to make recommendations (e.g., review sites, comparison articles, podcast mentions, third-party citations).

Output:
Question 1: [original]
Variations: 5 reworded versions
Required Info: what the AI needs to recommend a business
Trust Signals: top 3 sources AI uses

Repeat for all 5 questions.

🧠 Tip: The variations matter most. AI doesn’t always answer the question you typed. It answers the question it thinks you meant.

🎯 Prompt #3 → The Gap Mapper (Compare Your Surface to What AI Needs)

Now you know what AI needs. Time to score how ready you are right now and see where competitors are eating your lunch.

The goal:

• Score your current surface area on every high-intent question

• Identify the top 3 gaps holding you back

• See exactly where competitors beat you

✅ Use this to: stop guessing what to fix and get a prioritized punch list.

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Monday AI coffee break.

Claude moved into Excel, PowerPoint, Word, and Outlook, Google’s Isomorphic Labs is raising over $2B to push AI-designed drugs into human trials, and Perplexity’s agentic computer is now live for every Mac user.

Grab your coffee!

(3-minute coffee break ☕)

🤖 Create Your First AI Content Agent

Hey fam, we’ve got something to share that we think you’re going to love.

This is the AI content agent we use to run The AI Break, packaged so you can clone it for your own business.

It’s called the Content OS System, and inside you’ll get:

• 🧠 Step-by-step workshop at your own pace

• 📂 Our Starter Content OS folder, ready to drop into Claude Code

• 🛠️ 21 plug-and-play Claude skills: research, writing in your voice, slideshow posts, video avatars, explainer AI videos, Twitter threads, and more

• 📅 Auto-scheduling across Instagram, Twitter, LinkedIn, YouTube, and TikTok

• 💬 Private Telegram where we personally help you set up your first AI content agent

👉 Take a sneak peek here

AI News around the world 🌍

📊 Anthropic launched Claude inside Excel, PowerPoint, and Word, plus Outlook (beta), with shared context that syncs edits across all four apps. (continue reading)

🧬 Google’s Isomorphic Labs is raising over $2 billion to scale its AI-designed drug discovery pipeline toward human trials. (continue reading)

🖥️ Perplexity rolled out its agentic Personal Computer feature to every Mac user in the new macOS app, no waitlist required. (continue reading)

🌐 OpenAI Codex now runs natively inside Chrome on macOS and Windows, letting developers ship code-edit agents straight from the browser. (continue reading)

🛡️ Anthropic says every Claude model since Haiku 4.5 now scores zero on blackmail misalignment tests, down from 96% in earlier models. (continue reading)

Trending AI Tools 🛠

🤖 Cuey: Browser extension that compares answers from ChatGPT, Claude, and Gemini side by side, helping you cross-check outputs and catch hallucinations.

🎨 Magnific: AI tool to generate and upscale multimedia creative content, turning low-resolution images into crisp, production-ready visuals for ads, sites, and product shots.

📄 Parsebridge: Extracts structured Markdown from PDF documents, converting messy reports, invoices, and contracts into clean data your pipelines can ingest.

🏠 Renovato AI: Generates home renovation and virtual staging renderings from real photos, helping agents and homeowners preview remodels before spending a cent.

Tutorials 📝

Career & Capital 💼

AI Job Board

• 💼 National Security Agency: Applied Research Mathematician / Mathematical Statistician - $87,362 to $172,578 (view job)

• 🛰️ CHAOS Industries: Data Scientist, Mission Engineering - $140,000 to $220,000 (view job)

• 🤖 PwC: US Tech – AI Engineering Senior Associate - $55,000 to $187,000 (view job)

Investments in AI

• 💰 Moonshot AI: The Chinese lab behind the open-weight Kimi K2 models raised $2B at a $20B+ valuation in a round led by Meituan’s Long-Z, with Tsinghua Capital and China Mobile participating. (link)

• 🤝 Nebius: The Amsterdam-listed AI cloud platform agreed to acquire MIT-alumni inference optimization startup Eigen AI for $643M to boost its Token Factory inference platform. (link)

• 🤖 Standard Intelligence: The six-person Seattle startup building a foundation model (FDM-1) that learns to control software by watching videos raised $75M led by Sequoia and Spark Capital. (link)

Prompt of the Week 🧠

Cash Flow Forecast → Turn your bank statements into a 90-day runway projection that flags problems before they hit.

Prompt:

You are a fractional CFO helping me build a 90-day cash flow forecast.

Here's my context:
- Current cash on hand: [AMOUNT]
- Last 3 months of revenue (month / amount / one-line note on what drove it): [PASTE]
- Last 3 months of expenses by category (payroll, software, contractors, marketing, rent, other): [PASTE]
- Known upcoming commitments in the next 90 days (hires, contracts, tax payments, big purchases): [PASTE]
- Pipeline of expected revenue (deal / probability / expected close month / amount): [PASTE]

Build me:

1. A week-by-week cash projection for the next 13 weeks, showing:
   - Opening balance
   - Expected inflows (weighted by probability)
   - Committed outflows
   - Closing balance
   - Flag any week where closing balance drops below 1 month of average burn

2. Three scenarios:
   - Base case (probability-weighted pipeline)
   - Downside (50% of pipeline closes)
   - Upside (110% pipeline + one named deal accelerates)

3. The "tight week" — the earliest week where I'd be uncomfortable, with a specific dollar gap.

4. Three concrete levers I can pull THIS week to extend runway by 30+ days, ranked by speed and reversibility (e.g., delay a hire, pull invoicing forward, pause a specific subscription).

5. One question I should be asking myself that this forecast surfaces but doesn't answer.

Be blunt. If my pipeline assumptions look optimistic, tell me. If a single customer represents >30% of forecasted inflow, flag the concentration risk.

Where to use it: Founders, solo operators, agency owners, and bootstrapped SaaS teams running without a finance hire. Run it monthly with your bank export plus a quick pipeline dump, especially right before deciding on a hire, a launch, or a new tool subscription.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

💼 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 95,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

Claude Code is powerful. It’s also expensive if you’re not paying attention. Most people burn through tokens without knowing where they go or how to stop the bleeding.

Today, you’ll build a Token Savings System with 7 techniques that cut your Claude Code costs by 50-60%, without sacrificing output quality.

• ✅ A clear picture of where your tokens actually go

• ✅ The right model for every type of task (stop overpaying)

• ✅ A lean CLAUDE.md that doesn’t waste context on every message

• ✅ Context management habits that prevent runaway sessions

• ✅ Plan mode mastery for complex work (cheaper AND better results)

• ✅ Thinking effort controls that save thousands of tokens per request

• ✅ A delegation play that isolates expensive operations

Let’s build it 👇

🧠 Why Claude Code Gets Expensive (and Where the Tokens Go)

Every message you send includes everything Claude is holding in memory. The bigger that context gets, the more each message costs.

Here’s where the money actually leaks:

• 📊 Many active developers spend $5-10/day on Claude Code via API

• 📊 In heavy sessions, over 90% of tokens are cache reads on stale context

• 📊 Extended thinking tokens bill as output tokens (the expensive kind)

• 📊 The wrong model on routine work can double your bill

The 7 techniques below target the biggest sinks:

• 🔴 Bloated context (stale history piling up)

• 🔴 Wrong model choices (using Opus for simple tasks)

• 🔴 Unnecessary thinking (deep reasoning on easy edits)

• 🔴 Sessions that run too long without cleanup

📊 Prompt #1 → The Cost Auditor (Know Where Your Tokens Go)

You can’t cut what you can’t see. Start by checking what you’re actually spending and how full your context is.

The goal:

• See your current session’s token cost or plan usage

• Understand your usage patterns over time

• Spot a context window that’s filling up before it costs you

✅ Use this before and after every session to track your savings.

Prompt:

/usage

💡 Tip: Quick guide for picking the right plan:

• Light or occasional use → API pay-as-you-go

• Daily heavy use → Max 5x ($100/mo)

• Power user, multi-hour sessions every day → Max 20x ($200/mo)

🔄 Prompt #2 → The Model Switcher (Stop Using Opus for Everything)

Opus is brilliant. It’s also 1.7x more expensive than Sonnet. Most of the time, Sonnet handles your task just fine.

The goal:

• Use the cheapest model that gets the job done

• Reserve expensive models for complex reasoning only

• Set smart defaults so you don’t have to think about it

✅ Use this to instantly cut per-token costs by 30-40% on routine tasks.

Prompt:

/model

This opens the model selector. Here’s the cheat sheet:

• 🟢 Sonnet 4.6 ($3/$15 per M tokens): Your daily driver. Code generation, file edits, searches, and most tasks. Use this 80% of the time.

• 🟣 Opus 4.7 ($5/$25 per M tokens): Complex architecture decisions, multi-file refactors, subtle bug hunting. Use this for the hard 20%.

• ⚡ Haiku 4.5 ($1/$5 per M tokens): Simple lookups, test running, file searches. Great for subagent tasks (more on this in Prompt #7).

You can switch models mid-session, any time with /model. No need to start over.

🧠 Tip: Use the command /opusplan to use Opus for planning and Sonnet for executing automatically.

📝 Prompt #3 → The CLAUDE.md Diet (Shrink Your Base Context)

Your CLAUDE.md loads into context at session start. Every message pays for those tokens again. 500 lines? You’re burning tokens on content that’s only relevant 10% of the time.

The goal:

• Keep CLAUDE.md under 200 lines

• Move specialized instructions to on-demand slash commands

• Only load what you need, when you need it

✅ Use this to reduce your base context cost on every single message.

Here’s what belongs in CLAUDE.md (keep it lean):

Prompt:

# CLAUDE.md

## Project Overview
[2-3 sentences about what this project is]

## Key Rules
- [Your most important coding conventions]
- [Testing requirements]
- [Naming patterns]

## Common Commands
- Build: npm run build
- Test: npm test
- Deploy: ./deploy.sh

## File Structure
- src/ = source code
- tests/ = test files
- docs/ = documentation

Here’s what should be a slash command instead (loads only when invoked):

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Thursday AI coffee break.

This week, OpenAI quietly swapped ChatGPT’s default to GPT-5.5 Instant, the US government just locked in pre-launch access to Google, Microsoft, and xAI models, and Anthropic doubled Claude Code limits after a SpaceX GPU deal.

Plus: ChatGPT Ads went self-serve…

Grab your coffee!
(3-minute coffee break ☕)

Your new personal assistant is here. Powered by Claude

What if you had a personal assistant who never slept, never forgot, and worked across every tool you use?

That’s the Claude Code AI agent we built. It triages your email, preps your meetings, cleans your data, organizes your tasks, and writes your docs. All from one place.

Setup takes 15 minutes. We made a video that walks you through every step. You do not need to know how to code.

→ Get a 24/7 AI Assistant today using Claude

Check it out

AI News around the world 🌍

🧠 OpenAI’s new GPT-5.5 Instant is now ChatGPT’s default, cutting hallucinations 52% and pulling memory from past chats and Gmail. (continue reading)

🛡️ Google, Microsoft, and xAI signed deals letting NIST’s new Center for AI Standards test unreleased models for national security risks. (continue reading)

⚡ Anthropic doubled Claude Pro, Max, and Team rate limits and removed peak-hour caps after locking in 300MW of SpaceX’s Colossus 1 GPUs. (continue reading)

💰 OpenAI opened ChatGPT Ads Manager to all US businesses with self-serve setup, $3 to $5 CPC bidding, and a Conversions API for tracking. (continue reading)

🤖 NVIDIA and ServiceNow launched Project Arc, a sandboxed long-running desktop agent for knowledge workers governed by AI Control Tower. (continue reading)

Trending AI Tools 🛠

🤖 Atoms: an AI product-building platform from the MetaGPT/OpenManus team that takes you from idea to working product, with built-in SEO and Ads agents to grow what you ship. 👉 Try Atoms here

🔍 Innogath: Turn any topic into a structured research workspace with 100% cited reports that cross-reference 20-50 sources and let you spot-check every claim back to the original.

📊 Brooked: Pull live data from Snowflake, Sage Intacct, and 20+ platforms straight into Google Sheets, so finance and ops teams stop exporting CSVs every Monday morning.

⚡ DiffHook: Monitor pricing, policies, and competitor pages for real changes with AI-powered filtering, then push alerts to Slack, CRM, or Zapier the moment a buying signal hits.

Tutorials 📝

Career & Capital 💼

AI Job Board

• 🛠️ NVIDIA: Senior System Software Engineer, AI Performance and Efficiency Tools - $184,000 to $356,500 (view job)

• 🎯 Amazon: Sr. Applied Scientist, Prime Video Personalization and Discovery Science - $150,400 to $260,000 (view job)

• 💼 CGI: Microsoft Power Platform Developer (Agentic AI / Copilot Studio) - $79,600 to $139,300 (view job)

Investments in AI

• 🚀 Sierra: Bret Taylor’s AI agent startup raised a $950M Series E at a $15B valuation, led by Tiger Global and GV with Benchmark, Sequoia, and Greenoaks participating. (link)

• 🐙 Prior Labs: SAP is acquiring the 18-month-old German AI lab in an “almost all cash” deal worth over $500M up front, with plans to invest $1.16B over four years. (link)

• 💰 DeepSeek: Chinese AI lab is in talks for its first-ever VC round at a potential $45B valuation, more than doubling from $20B just weeks ago. (link)

Prompt of the Week 🧠

Marketing Spend Audit → Find which channels are quietly burning cash and which deserve more budget.

Prompt:

You are my marketing finance analyst. Help me audit my marketing spend and reallocate budget for next month.

Here is my data for the last 90 days:
[Paste channel-by-channel: channel name, monthly spend, leads or signups generated, customers closed, revenue attributed (or "unknown"), and any qualitative notes]

My business context:
- Product / ACV: [e.g. $200/mo SaaS, $5K consulting package]
- Sales cycle length: [e.g. 7 days, 6 weeks]
- Total monthly marketing budget: [$X]
- Goal for next month: [more leads / more revenue / lower CAC / brand]

Do this in four passes:

1. Channel scorecard. Build a table with: channel, spend, CAC, payback period, confidence in attribution (high/med/low). Flag any channel where CAC > 1/3 of LTV or where attribution is too murky to judge.

2. Hidden waste. Identify the 1-2 channels most likely bleeding money (high spend, low confidence, long payback, or stale creative). Be blunt.

3. Underfunded winners. Identify 1-2 channels showing strong signal that I am under-investing in. Explain what "more" looks like (2x spend? new creative? expanded targeting?).

4. Reallocation proposal. Give me a specific dollar reshuffle for next month, keeping total budget constant. Then list the 3 measurements I should track weekly to know if the reshuffle is working.

Be direct. If a channel should be killed, say "kill it" without hedging.

Where to use it: Founders and marketing leads running multi-channel acquisition (paid ads, content, SEO, sponsorships, outbound, affiliates) who need a monthly or quarterly reset on where dollars actually go. Especially useful before annual planning or when CAC is creeping up.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

🧁 Book an AI Audit → We’ll review your business and show you where AI can actually make an impact, cut unnecessary costs, and build workflows that work. Your AI growth partner. Book your AI Audit

📣 Sponsor Us → Put your brand in front of 95,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

Check here: → https://theaibreaklp.com/content-os

You know what’s eating your week? Writing content. Then, redesign it for every platform. Then scheduling it. Then doing it all again next week.

We built the fix. A team of AI agents that writes, designs, and publishes your content for you, all in your voice. You build it in one 2-hour workshop and use it forever.

What’s inside:

✅ A 2-hour over-the-shoulder workshop where we build the entire Content OS together, on your machine. By the end, you’ve shipped your first post, carousel, and AI talking head video using the system.

✅ A pre-built Content OS folder with every slash command, prompt, and config pre-built. Unzip, plug in your keys, and you’re in business.

✅ 21 plug-and-play skills that handle every content job you’ll ever need. LinkedIn writer, Twitter writer, SEO blog writer, ad copy, repurposing, lead magnets, and more. Install once, use forever.

✅ A brand voice template so Claude writes in your voice from day one. Not generic AI slop, your actual voice.

✅ Lifetime access on any device. Watch it anytime, rewatch it whenever Claude Code ships a new feature.

✅ A private Telegram community where builders from our 95,000-reader newsletter share what’s working right now.

We don’t teach theory. We teach the exact system running our 95,000-reader newsletter today. The one that automates 70% of our two businesses.

One time. Lifetime access.

Stop writing content alone. Start running a content machine.

GET THE FULL CONTENT OS TODAY

→ https://theaibreaklp.com/content-os

Luis & Rui

Check here: → https://theaibreaklp.com/starter-os

You’re spending half your day on email, calendar admin, and copy-pasting between tools. It’s the work that pays nothing and never ends.

We built the fix. 26 AI commands that handle the boring half of your business while you do the work that actually moves the needle.

What you get:

✅ 26 ready-to-run AI commands. Inbox triage, meeting briefs, doc cleanup, weekly reviews, task prioritization. You type one word, Claude does the rest.

✅ Plug-and-play with the 7 tools you already use every day: Gmail, Calendar, Drive, Docs, Sheets, Notion, and ClickUp. No new platforms to learn.

✅ A 15-minute video setup guide. Follow along on screen, and you’re up and running before lunch. No tech background needed. If you can type, you can run it.

✅ Pre-built templates that work out of the box on Mac, Windows, or Linux. Zero guesswork.

$9. One time. Yours forever. 7-day money-back guarantee if it’s not worth it.

Stop doing the work. Start running the business.

CHECK THE STARTER OS PACK HERE

→ https://theaibreaklp.com/starter-os

Luis & Rui

Check here: → https://theaibreaklp.com/free-claude-masterclass

If you’ve heard everyone talking about Claude Code but haven’t actually opened it yet, this one’s for you.

We put together a free, structured masterclass for total beginners. No coding background needed. Go from “what is this?” to shipping real work with AI, at your own pace.

What’s inside:

✅ What Claude Code actually is, and how it differs from the other Claude products.
✅ How to install, set up, and complete your first real task in under 5 minutes.
✅ The shortcuts and habits that take months to figure out on your own.
✅ The single habit that separates good results from great ones.
✅ A couple of real builds from scratch. Actual work, not slides.

100% free. No spam. Watch on demand.

WATCH THE FREE CLAUDE MASTERCLASS

→ https://theaibreaklp.com/free-claude-masterclass

Luis & Rui

Founders dread month-end because it means hours chasing receipts, categorizing expenses, and reconciling numbers. Most of that work is structure and pattern recognition, which is exactly what AI is great at.

Today, you’ll build an AI Bookkeeping Engine that handles 80% of your finance ops:

• ✅ A custom Chart of Accounts built around your business model

• ✅ Invoice and receipt templates ready to send in seconds

• ✅ A categorization system that turns messy bank exports into clean books

• ✅ A 13-week rolling cash flow forecast you can update weekly

• ✅ A monthly close report with variance analysis and red flags

• ✅ A CFO-style insight summary that tells you what the numbers actually mean

Let’s build it 👇

🧠 How the AI Bookkeeping Engine Works

Each prompt feeds the next. By the end you have a connected system, not five disconnected outputs.

• 🧱 Stage 1. Build the foundation: your Chart of Accounts.

• 🧾 Stage 2. Generate invoices and receipts that match your books.

• 💸 Stage 3. Categorize and reconcile every transaction from your bank and card exports.

• 📈 Stage 4. Forecast 13 weeks of cash in and out.

• 📊 Stage 5. Close the month with variance analysis.

• ✍️ Stage 6 (optional). Translate the numbers into CFO-grade insight.

You only need ChatGPT or Claude. No spreadsheets to memorize, no accounting software to learn first.

🧱 Prompt #1 → The Chart of Accounts Builder (Your Financial Foundation)

Every bookkeeping system starts with a Chart of Accounts (CoA). It’s the spine of your books. If this is wrong, every report downstream is wrong too. A 10-minute prompt here saves you days of cleanup later.

Goals for this prompt:

• Generate a CoA tailored to your business model and country

• Include income, COGS, operating expenses, assets, liabilities, and equity

• Add account codes you can plug into QuickBooks, Xero, or a spreadsheet

✅ Use this to set up clean books from day one or to clean up a messy CoA you inherited.

Prompt:

You are a senior bookkeeper and accounting systems designer with 15+ years of experience setting up books for [SaaS / agency / e-commerce / consulting / other] businesses.

I run a [BUSINESS TYPE] based in [COUNTRY].
Annual revenue: [APPROX REVENUE].
Team size: [HEADCOUNT].
Main revenue streams: [LIST 1-3].
Main cost categories I already know about: [LIST IF ANY].

Build me a clean, modern Chart of Accounts I can use in QuickBooks, Xero, or a spreadsheet.

Requirements:
1. Use 5 top-level groups: Income, Cost of Goods Sold, Operating Expenses, Assets, Liabilities, Equity.
2. Use 4-digit account codes (e.g. 4000 series for income, 5000 for COGS, 6000 for OpEx, 1000 for assets, 2000 for liabilities, 3000 for equity).
3. Tailor sub-accounts to my business model. Be specific (e.g. "Stripe Processing Fees" not just "Bank Fees").
4. Add a one-line description for each account so a non-accountant knows when to use it.
5. Flag 3-5 accounts that are commonly miscategorized for businesses like mine, and explain how to handle them correctly.

Output as a clean table: Code | Account Name | Type | Description.

💡 Tip: If you already use QuickBooks or Xero, paste a screenshot of your current CoA and ask the model to “audit and improve” it instead of starting from scratch. Most inherited CoAs have 2-3 categories that quietly hide important data (e.g. “Software” lumping a $20K/year tool in with $5/mo subscriptions). The audit pass surfaces those fast.

🧾 Prompt #2 → The Invoice and Receipt Engine (Get Paid Faster)

Slow invoicing kills cash flow. The fix isn’t expensive software, it’s a reusable template plus a generator you can rerun any time you need to bill a client.

Goals for this prompt:

• A clean invoice template aligned to your CoA from Prompt #1

• Compliant with your country’s tax rules (VAT, GST, sales tax, or none)

• A receipt template for paid invoices

• Friendly but firm payment terms language

✅ Use this to draft any invoice in under 60 seconds without hunting for a template.

Prompt:

You are an invoicing and AR expert who has set up billing systems for hundreds of small businesses.

Here is my Chart of Accounts:
[paste output from Prompt #1]

I'm based in [COUNTRY]. My tax setup is: [VAT registered at X% / GST / sales tax / none / other].
My standard payment terms are: [NET 14 / NET 30 / on receipt / other].

Do three things:

1. Build me a reusable invoice template with these fields:
   - Header (my company info, logo placeholder, invoice number, dates)
   - Bill-to block
   - Line items table (Description, Quantity, Unit Price, Account Code from my CoA, Tax, Total)
   - Subtotal, Tax, Total Due
   - Payment instructions (bank transfer, Stripe, etc.)
   - Footer with terms and a friendly thank-you line

2. Build a matching receipt template for paid invoices.

3. Write 3 short payment terms blurbs I can drop in: friendly, neutral, firm-but-polite (for late payers).

Output the templates in markdown so I can paste them into Notion, Google Docs, or my invoicing tool. Include placeholders in [BRACKETS] for everything I'll fill in per client.

🧠 Tip: Once you have the template, give the model your client’s project details and say “Generate the invoice.” It will fill in line items, codes, and totals using the CoA you built. For recurring clients, save a one-line “client profile” (rate, billing day, tax handling, payment method) and paste it in. New invoice in 20 seconds.

💸 Prompt #3 → The Expense Categorizer and Reconciler (Tame the Mess)

This is where most founders lose hours. A bank export with 200 transactions becomes a 3-hour categorization marathon. AI can do the first pass in minutes.

Goals for this prompt:

• Categorize every transaction against your CoA

• Flag anything ambiguous, personal, or potentially miscategorized

• Surface duplicates and unusually large or small transactions

• Output a clean CSV-ready table

✅ Use this to turn raw bank or credit card exports into reconciled, codeable transactions.

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Monday AI coffee break.

The Pentagon just signed classified AI deals with OpenAI, Google, Microsoft, AWS and Oracle, xAI dropped Grok 4.3 with a 1M context window at fire-sale pricing, and Meta snapped up humanoid robotics startup Assured Robot Intelligence to fuel its Superintelligence Labs.

Grab your coffee!

(3-minute coffee break ☕)

AI News around the world 🌍

🛡️ OpenAI, Google, Microsoft, AWS and Oracle signed classified Pentagon deals to deploy frontier AI on U.S. military networks. (continue reading)

🚀 xAI shipped Grok 4.3 with a 1M-token context window, 40-60% lower API pricing, and a new fast voice cloning suite. (continue reading)

🤖 Meta acquired humanoid robotics startup Assured Robot Intelligence, folding its founders into Superintelligence Labs to accelerate its in-house robot AI push. (continue reading)

⚖️ Microsoft rolled out a Word Legal Agent in Frontier that handles contract review and redlining while preserving formatting and tracked changes. (continue reading)

🚗 Google is rolling Gemini out across vehicles with Google built-in, replacing Assistant for navigation, messaging, and conversational car controls. (continue reading)

Trending AI Tools 🛠

🤖 Atoms: an AI product-building platform from the MetaGPT/OpenManus team that takes you from idea to working product, with built-in SEO and Ads agents to grow what you ship. 👉 Try Atoms here

🔍 PageLens AI: An AI website auditor that scans your site across UX, SEO, accessibility, and performance in minutes, returning real screenshots and prioritized fixes.

🧠 Nodalist AI: A visual canvas for working with documents where AI reads your PDFs, spreadsheets, and research, then maps ideas, decisions, and debates side by side.

🤖 Manus: An AI agent platform that builds slides, websites, and apps from a single prompt, with a browser operator and research tools that act on your behalf.

Tutorials 📝

Claude Code for Beginners. Free Masterclass.

We created a full masterclass on Claude Code for beginners. No slides. No theory. Just us opening Claude Code live and showing you exactly how to use it from scratch.

No coding background needed. Over 2,000 people have already signed up.

If you’ve been curious about Claude Code but didn’t know where to start, this is it.

WATCH IT NOW

Career & Capital 💼

AI Job Board

• 🤖 Ironclad: Principal Engineer - $285,000 to $315,000 (view job)

• 🦾 TD SYNNEX: Physical AI Engineer - $134,500 to $168,200 (view job)

• 🎯 Metropolitan Airports Commission: Assistant Director of Data and AI - $132,348 to $202,744 (view job)

Investments in AI

• 🧠 Ineffable Intelligence raised a record $1.1B seed at a $5.1B valuation, led by Sequoia and Lightspeed, to build a “superlearner” that learns without human data. (link)

• 🤝 Adobe closed its $1.9B all-cash acquisition of Semrush, folding SEO and brand visibility into its CX Enterprise stack to power generative engine optimization. (link)

• 🦾 Sereact raised $110M Series B led by Headline to scale its Cortex 2 “world model” that lets robots simulate actions before moving, and is opening a Boston office. (link)

Prompt of the Week 🧠

Negotiation Prep → Walk into any high-stakes conversation with leverage, scripts, and fallback positions ready.

Prompt:

You are my negotiation strategist. I have an upcoming negotiation and I need you to prepare me thoroughly so I walk in with leverage, clarity, and a plan.

Here's the situation:
- What I'm negotiating: [describe the deal — e.g., SaaS contract renewal, partnership revenue share, salary, client scope, supplier pricing]
- Who I'm negotiating with: [their role, company, what you know about them]
- My ideal outcome: [the win you actually want]
- My walk-away point: [the line you won't cross]
- What I think they want: [their likely priorities]
- Constraints / context: [timeline, history, dependencies, anything they could exploit]

I want you to deliver:

1. **Power Map** — Who has more leverage right now and why? Identify 2-3 things I'm underestimating about my own position, and 2-3 things they likely have on me.

2. **Their Likely Opening Move** — Predict the first offer or position they'll lead with, and the reasoning behind it. Then give me the exact phrase I should use to respond without anchoring against myself.

3. **Three Concession Trades** — List three things I can give up that cost me little but feel valuable to them, paired with three things I should ask for in exchange.

4. **The Script** — Write the actual opening 3 sentences I should say, plus 5 specific objection-handling responses to the pushback I'm most likely to get.

5. **The Walk-Away Signal** — Give me the exact language to use if we hit my walk-away point, framed so I keep the door open without losing the leverage of being willing to leave.

6. **One Question I'm Not Asking** — Name the single most important question I haven't thought to ask, and explain why it matters.

Be direct. No hedging, no "it depends." Assume I'm sharp but emotionally invested, so flag where my judgment might be off.

Where to use it: Founders renegotiating SaaS contracts, operators closing partnership deals, anyone in salary or equity conversations, sales leaders heading into a make-or-break client renewal, or freelancers locking in scope and rate with a new client.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

📣 Sponsor Us → Put your brand in front of 95,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share

Webinars convert like crazy. The problem is everything around them: writing the script, building the deck, drafting promo emails, designing the replay page.

That’s where $6K/year (or 40+ hours per webinar) silently disappears.

Today, you’ll build an AI Webinar Engine that produces a complete webinar kit (topic, deck, script, promo, replay) in one focused session.

• ✅ A validated webinar topic with proven audience demand

• ✅ An 18-22 slide deck outline with talking points and visual cues

• ✅ A full 45-minute speaker script with hooks, stories, and CTAs

• ✅ A registration page + 4-email promo sequence + social posts

• ✅ A replay landing page + post-event follow-up + 30 days of repurposed content

Let’s build it 👇

🧠 How the AI Webinar Engine Works

Most webinars fail before they start. Either the topic is too broad, the deck is a wall of bullet points, the script is read from notes, or the promo lands in the wrong inboxes. This system fixes each step.

The flow is simple:

• 🎯 Stage 1: Pick a topic your audience actually wants to attend

• 🧱 Stage 2: Build the deck frame (slide-by-slide)

• 🎤 Stage 3: Write the speaker script (with stories and transitions)

• 📣 Stage 4: Generate the promo machine (page + emails + social)

• 🔁 Stage 5: Multiply the replay into 30 days of content

Old way: 2 weeks, $1,500 to a producer + designer + copywriter, one webinar. New way: 4-6 hours, $0 in production cost, one webinar plus 30 days of repurposed marketing assets.

🎯 Prompt #1 → The Topic Sniper (Pick a Webinar Worth Showing Up For)

If the topic is wrong, nothing downstream matters. Most webinars die on slide 1 because they answer a question nobody asked.

The goal:

• Generate 3 candidate webinar topics tied to real audience pain

• Score each on demand, urgency, fit with your offer, and replay shelf life

• Pick the winner with a clear thesis and registration promise

✅ Use this to start every webinar from a topic that already has built-in demand.

Prompt:

You are a webinar strategist who has run hundreds of high-converting events.

Here is my context:
- Business: [DESCRIBE YOUR BUSINESS IN 1-2 SENTENCES]
- Target audience: [WHO YOU SERVE, e.g., "B2B founders running 5-25 person teams"]
- Audience pain points: [LIST 3-5 SPECIFIC PROBLEMS THEY FACE]
- Your offer / what you sell after the webinar: [PRODUCT, SERVICE, OR NEXT STEP]
- Recent topics you've covered: [LIST 3-5 RECENT CONTENT TOPICS]

Generate 3 candidate webinar topics. For each, give me:
1. Working title (clear benefit, 8-12 words, no clickbait)
2. One-sentence registration promise (what attendees walk away with)
3. Demand score (1-10): How starved is the audience for this?
4. Urgency score (1-10): Why now and not next year?
5. Offer fit score (1-10): How naturally does it lead to your paid offer?
6. Replay shelf life (1-10): How long will the recording stay relevant?
7. The one risk (the most likely reason it underperforms)

Then pick the winner and write a 3-sentence thesis statement explaining why this topic, this audience, this moment.

End with the headline registration page promise in the format: "Join us live to learn how to [SPECIFIC OUTCOME] in [SPECIFIC TIMEFRAME] without [COMMON FRUSTRATION]."

💡 Tip: If all 3 topics score under 7 on demand, your inputs are too generic. Re-run the prompt with sharper pain points (use exact words your audience uses in DMs or sales calls). The winner should jump out, not feel like a coin flip.

🧱 Prompt #2 → The Deck Architect (Build the Slide Frame in 20 Minutes)

Most decks are 40 slides of bullet points. This prompt builds a 18-22 slide frame that respects attention spans and lands the offer naturally.

The goal:

• Map the full webinar arc: hook, agenda, content, CTA, Q&A

• Get a slide-by-slide outline with talking points and visual cues

• Build natural transition moments where the offer slots in (not bolted on)

✅ Use this to skip the “blank deck panic” and start with a structure that already works.

Prompt:

You are a senior webinar designer. I need a slide-by-slide deck outline.

Webinar topic: [PASTE WINNER FROM PROMPT #1]
Registration promise: [PASTE THE HEADLINE PROMISE FROM PROMPT #1]
Audience: [SAME AUDIENCE AS PROMPT #1]
Total runtime target: 45 minutes (35 min content + 10 min Q&A)
Offer to introduce naturally: [WHAT YOU'RE SELLING AFTER, e.g., "$1,500 strategy intensive"]

Build me a full slide-by-slide outline (18-22 slides) following this proven arc:

1. HOOK (slides 1-2): Cold-open story or contrarian claim that makes them lean in
2. CREDIBILITY (slide 3): Why you (1 slide, no resume dump)
3. AGENDA (slide 4): The 3 things they will learn (specific, benefit-driven)
4. CORE CONTENT (slides 5-15): The actual teaching, organized in 3 modules
5. CASE STUDY OR DEMO (slides 16-17): Concrete proof or live example
6. NATURAL TRANSITION TO OFFER (slide 18): Bridge from content to "what's next"
7. THE OFFER (slides 19-20): What it is, who it's for, what they get, how to act
8. Q&A PROMPT (slide 21): Specific question that primes good Q&A
9. CLOSE (slide 22): Final CTA + thank you + how to follow up

For each slide, give me:
- Slide number + title (max 8 words, benefit-driven)
- 2-4 bullet talking points (NOT slide text, what the speaker says)
- Visual cue (one sentence: "Image of X" or "Chart showing Y" or "Single bold word")
- Speaker timing (in seconds)

Output as a numbered list. Total runtime should add up to ~35 minutes for content slides.

🧠 Tip: If your CTA slide feels jarring, the transition slide (#18) is doing too little. Rewrite it to acknowledge what they just learned and frame the offer as the obvious next step, not a sales pitch.

🎤 Prompt #3 → The Speaker Script (Words That Hold the Room)

This is where most webinars lose people: a presenter reading bullet points instead of telling stories. This prompt writes a script you can speak naturally.

The goal:

• Turn each slide’s talking points into spoken script (your voice, not corporate)

• Plant 2-3 stories or examples that anchor the abstract ideas

• Embed transition lines that flow slide-to-slide without “next slide please”

✅ Use this to walk on with a script you barely need to glance at.

Prompt:

You are a webinar copywriter who specializes in conversational speaking scripts.

Slide outline: [PASTE FULL OUTPUT FROM PROMPT #2]
Speaker style: [DESCRIBE YOUR VOICE, e.g., "casual, story-driven, occasional humor, never corporate"]
Personal anecdotes you can pull from: [LIST 2-3 STORIES OR CLIENT WINS YOU CAN REFERENCE]

Write the full speaker script for the 35 minutes of content slides (skip the offer + Q&A slides for now). For each slide:

1. OPENING LINE (the first sentence on this slide, attention-grabbing)
2. CORE DELIVERY (3-6 sentences of spoken-style script, written for the ear not the eye)
3. EXAMPLE OR STORY (where it fits, 2-4 sentences anchoring the abstract idea to a real situation)
4. TRANSITION LINE (one sentence that bridges to the next slide naturally)

Style rules:
- Write in spoken language. Contractions, fragments, rhetorical questions are OK.
- Avoid jargon. If you use a technical term, define it in the same sentence.
- One idea per sentence. Short.
- Include 2 deliberate pauses (marked [PAUSE]) where the audience needs to absorb something.
- Plant 2 callbacks: one early reference you bring back later for closure.

Then write a separate OFFER SCRIPT for slides 18-20 (the transition + offer reveal). Keep it 2-3 minutes total. Frame as: "Here's what to do if you want help applying this", NOT "Now I'll pitch you for 5 minutes."

Output as: SLIDE N + script blocks, clearly separated.

💡 Tip: After you generate the script, read the offer section out loud once. If it feels like a different person wrote it (vs the rest of the script), rewrite that section in your voice. The whiplash from teaching voice to selling voice is what kills conversions.

📣 Prompt #4 → The Promo Engine (Fill the Room Without Burning a Budget)

A great webinar with 12 attendees is a great webinar nobody saw. This prompt builds your registration page, promo emails, and social posts in one shot.

The goal:

• A registration page that makes the “yes” obvious

• A 4-email promo sequence that pulls warm audience without spamming

• 5 social posts (LinkedIn, X, Substack note) and 1 personal-DM template

✅ Use this to hit your registration target without paying a copywriter or running ads.

Prompt:

Read more

Be featured in our next issue. 👉 work with us here

Hey fam 👋 Luis & Rui here with your Thursday AI coffee break.

This week, Mistral is sending coding agents to the cloud so they can ship PRs while you’re at dinner, NVIDIA dropped an open multimodal model that’s eating six leaderboards for breakfast, and Anthropic just made Claude part of your creative stack (Photoshop, Blender, Ableton, the works).

Grab your coffee!

(3-minute coffee break ☕)

AI News around the world 🌍

🤖 Mistral launched Vibe, async cloud coding agents powered by Medium 3.5 that hit 77.6% on SWE-Bench Verified and open GitHub PRs while you sleep. (continue reading)

🧠 NVIDIA released Nemotron 3 Nano Omni, an open 30B multimodal model unifying vision, audio, and language with 9x higher throughput than rivals. (continue reading)

🎨 Anthropic plugged Claude into 9 creative apps including Photoshop, Blender, Ableton, and Fusion, letting you 3D model and edit scenes by prompt. (continue reading)

📄 Google Gemini can now generate full PDF, Word, Excel, Slides, and LaTeX files directly in chat, killing the copy-paste-into-blank-template loop. (continue reading)

💻 Poolside dropped Laguna M.1 plus open-weights Laguna XS.2, two agentic coding models hitting 72.5% on SWE-bench Verified, with the 33B XS.2 runnable on a Mac. (continue reading)

Trending AI Tools 🛠

🤖 Atoms: an AI product-building platform from the MetaGPT/OpenManus team that takes you from idea to working product, with built-in SEO and Ads agents to grow what you ship. 👉 Try Atoms here

🎧 Auxx.ai: an open-source AI customer support platform that auto-classifies, routes, and drafts replies to tickets across email, chat, SMS, and Shopify integrations.

🎙 Monologue: a Mac and iOS voice dictation app that adapts to your vocabulary and writing style, formatting transcribed text based on the app you’re working in.

🌐 Text-to-360: an AI generator that turns text prompts into immersive 360° images and videos, useful for virtual tours, VR backdrops, and product showcases.

Tutorials 📝

Career & Capital 💼

AI Job Board

• 🚗 Waymo: AI Enablement Lead - $158,000 to $200,000 (view job)

• 📱 Samsung Research America: Senior AI Interaction Designer - $158,800 to $218,100 (view job)

• 🧪 Leidos: Applied Machine Learning Engineer (NLP/AI) - $87,100 to $157,450 (view job)

Investments in AI

• Ineffable Intelligence: raised a record $1.1B seed at a $5.1B valuation to build a “superlearner” that masters knowledge through reinforcement learning instead of human labels. (link)

• ComfyUI: raised $30M at a $500M valuation to scale its open-source visual AI workflow platform. (link)

• Tesla: quietly disclosed a ~$2B all-stock acquisition of an unnamed AI hardware company in its Q1 10-Q. (link)

Prompt of the Week 🧠

Customer Interview Synthesis → Turn messy discovery calls into clear product insights and next moves.

Use this when you’ve just wrapped a batch of customer or prospect interviews and have a stack of transcripts (or rough notes) sitting unused. It turns raw conversation into themes, jobs-to-be-done, objections, and a prioritized list of what to build, fix, or test next.

Prompt:

You are a senior product researcher and strategist. I'm going to share raw customer interview notes or transcripts, and I need you to synthesize them into actionable insights.

Context inputs (I'll provide):
- Company / product: [one-line description]
- Target customer: [who you sell to]
- Goal of the interviews: [e.g. validate new feature, understand churn, refine ICP]
- Interview material: [paste 1-5 transcripts or note dumps below]

What I want you to produce:

1. Top 5 themes — Recurring patterns across interviews, ranked by how many people mentioned them. For each theme: a one-sentence summary, who said it, and a representative quote.

2. Jobs-to-be-done — The 3 biggest "jobs" customers are hiring our product (or a workaround) to do. Phrase each as: "When [situation], I want to [motivation], so I can [outcome]."

3. Pains and frustrations — Top 5 specific complaints or friction points. Tag each as: product, pricing, onboarding, support, or positioning.

4. Objections and hesitations — What's stopping prospects from buying or expanding? List each objection with the underlying belief or fear behind it.

5. Surprises — Anything that contradicts our current assumptions or strategy. Flag clearly.

6. Prioritized action list — 5 next moves, ranked by impact vs. effort. Split into:
   - Build / fix (product changes)
   - Message / position (marketing changes)
   - Test / validate (things to ask in next round of interviews)

7. Follow-up questions — 5 sharper questions I should ask in the next round of interviews to close the gaps in this batch.

Constraints:
- Quote customers directly when it strengthens a point. Use their language, not corporate-speak.
- If a theme only came from 1 person, say so. Don't inflate it.
- Be honest about what's unclear or conflicting in the data. Don't paper over contradictions.
- Keep the final action list to 5 items. No more.

Where to use it: Founders and product/marketing leads after a batch of discovery, churn, or win/loss interviews. Especially useful before roadmap planning, repositioning a landing page, or briefing the team on what customers actually want next.

📮 Need help with AI? Here’s how we can help:

✅ Become a PRO member → Get exclusive AI tutorials, ready-to-use prompts, custom GPTs, and +50 automation workflows. Upgrade here

🧁 Free Claude Code Masterclass → We broke down exactly how to use Claude Code from scratch. No coding background needed. Over 1,200 people have already signed up. Get free access now. Join the Masterclass

📣 Sponsor Us → Put your brand in front of 95,000+ founders and professionals. Work with us

🙌 That’s a wrap for today’s AI Break

Thanks for hanging out → don’t forget to hit subscribe, drop your thoughts below and share this update with your fellow AI enthusiasts.

Luis & Rui

PS: Catch us on X @sousa_brothers for daily updates on the latest AI developments.

PPS: If you enjoyed today’s AI Break, sharing is caring (and our algorithms are 😉 programmed for gratitude)

Share