DEV Community: Danilo Caffaro

I Built a Free AI Tech Advisor That Recommends Products Better Than Any Salesperson

Danilo Caffaro — Sat, 04 Apr 2026 22:58:00 +0000

TL;DR: I built melhor.dev — a free AI-powered tech review site for the Brazilian market. After 6 months of testing, here are the exact prompts, workflows, and tools I used. Plus, I compiled everything into a free downloadable guide.

The Problem With Tech Reviews

Every time someone asks "should I buy Sony WH-1000XM6 or AirPods Pro 3?", they get:

Sponsored articles that rank by affiliate commission, not quality
Outdated reviews from 2023 with wrong prices
Generic specs copy-pasted from Amazon

I wanted something different: AI-analyzed, always current, zero BS.

What I Built: melhor.dev

tech.melhor.dev — 35 pages of structured tech reviews for the Brazilian market.

Every page has:

Answer box for AI search engines (GEO/AEO optimized)
Structured product data with schema markup
Honest pros/cons (no sponsored content)
Local prices in BRL

The whole system runs on Astro SSG + a few AI-powered workflows.

The AI Workflow Behind It

Step 1: Research Automation (n8n)

Trigger: New product category request
→ Web search (Perplexity API) for latest prices + reviews
→ GPT-4 analyzes: extract specs, generate verdict, write pros/cons
→ Output: structured JSON with product data
→ Astro page generation from template

Step 2: SEO + GEO Optimization

Each article gets:

Featured snippet-ready answer box (first 50 words)
FAQ schema (JSON-LD)
Product schema with real prices
robots.txt that explicitly allows GPTBot, ClaudeBot, PerplexityBot

Step 3: Multi-platform Distribution

IndexNow → instant Bing/DDG indexing
Google Search Console API → sitemap push
llms.txt → AI engine discovery

The Results So Far

35 pages live in 3 days
TTFB < 30ms (static site + CDN)
0 paid traffic — 100% organic
First GSC impressions appearing within 48h

Key Insight: The Gap in the Market

Brazilian tech review sites are terrible. Reveio, ReviewDosMelhores — generic, shallow, clearly SEO farmed.

The opportunity: authentic, structured content that actually helps people make decisions.

This is exactly what LLMs want to cite. Google AI Overview loves answer-first content. Perplexity loves structured data.

Free Downloads

I packaged everything into free resources:

1. Free: n8n AI Automation Workflows Pack

If you want to replicate this workflow setup, I published the exact n8n templates I use:
Get the free workflow pack →
(Pay what you want — including /usr/bin/bash)

2. Free/Paid: AI Agent Team Guide

The full system including 15+ agent templates, the content workflow, and the SEO automation setup:
AI Agent Team Guide → (from )

3. The Site Itself

Browse the full tech review site: tech.melhor.dev

Tech Stack (full)

Component	Tool	Cost
Framework	Astro SSG	Free
AI workflow	n8n (self-hosted)	~/mo VPS
Search	Perplexity API	~/mo
Analytics	GA4	Free
CDN/DNS	Cloudflare	Free
Hosting	DigitalOcean	4/mo

What's Next

Building the next 15 articles this week. Planning to add:

Price tracking (auto-update when prices change)
User-submitted reviews
WhatsApp bot that answers product questions

If you're building something similar or want to discuss the AI workflow, drop a comment — happy to share more details.

This is part of a series on building AI-powered web properties. Follow for updates.

How I Automated My Entire Business with n8n + AI (5 Workflows You Can Steal)

Danilo Caffaro — Sat, 07 Mar 2026 16:35:18 +0000

I was spending 20+ hours/week on repetitive tasks

Client onboarding emails. Social media posting. Lead follow-ups. Invoice reminders. Report generation.

Every week, the same tasks. Every week, I told myself: "I'll automate this eventually."

Then I found n8n — an open-source automation platform that changed everything. Combined with AI (ChatGPT/Claude APIs), I built workflows that now run my business on autopilot.

Here are 5 workflows you can steal today.

Workflow 1: AI-Powered Client Onboarding

Before: 45 minutes per new client (welcome email, Notion setup, Slack invite, CRM update)

After: 2 minutes (trigger form → n8n does everything)

Trigger: Typeform submission
→ Create Notion page with client details
→ Send personalized welcome email (AI-generated based on their industry)
→ Add to CRM pipeline
→ Create Slack channel
→ Schedule 7-day follow-up sequence

Time saved: ~3 hours/week

Workflow 2: Content Generation Pipeline

Before: 4 hours writing one blog post

After: 30 minutes editing an AI draft

Trigger: Schedule (every Monday 9am)
→ Fetch trending topics from Dev.to API
→ Claude generates outline + first draft
→ Save to Google Docs for editing
→ Post to Slack #content-review channel
→ After approval: publish to Dev.to + LinkedIn

Time saved: ~3.5 hours/post

Workflow 3: Lead Qualification Bot

Before: Manually reviewing every lead form submission

After: AI scores and routes leads automatically

Trigger: New form submission
→ AI analyzes: company size, budget, urgency, fit score
→ Score > 80: Direct to sales calendar (Calendly)
→ Score 50-80: Add to nurture email sequence
→ Score < 50: Send resources PDF, add to newsletter

Conversion improvement: 40% more qualified meetings

Workflow 4: Customer Support Triage

Before: Reading every support email to decide priority

After: AI categorizes and routes instantly

Trigger: New email to support@
→ AI classifies: billing, technical, feature-request, urgent
→ Urgent → Slack alert + auto-reply with ETA
→ Billing → Auto-resolve common questions (invoice links, plan details)
→ Technical → Create GitHub issue + assign to dev
→ Feature request → Add to Notion roadmap

Response time: From 4 hours → 12 minutes average

Workflow 5: Weekly Business Intelligence Report

Before: 2 hours pulling data from 6 different dashboards

After: Auto-generated report every Monday at 8am

Trigger: Schedule (Monday 8am)
→ Pull metrics: Stripe revenue, Google Analytics, social stats, email open rates
→ AI generates executive summary with trends + recommendations
→ Create beautiful HTML report
→ Send to team Slack + email

Time saved: 2 hours/week + better insights

Why n8n over Zapier or Make?

Feature	n8n	Zapier	Make
Self-hosted option	✅ Free	❌	❌
AI integration	Native	Limited	Basic
Pricing	Free tier + $20/mo	$20/mo (limited)	$9/mo (limited)
Code flexibility	Full JS/Python	No-code only	Limited
Community templates	1000+	6000+	1500+

The killer advantage: n8n is open source. You can self-host it for free, customize everything, and your data never leaves your server.

Total time saved: 15+ hours/week

Workflow	Hours saved/week
Client onboarding	3h
Content pipeline	3.5h
Lead qualification	2h
Support triage	4h
Weekly reports	2h
Total	14.5h

That's almost 2 full workdays back. Every single week.

Want the complete guide with all templates?

I wrote a comprehensive guide covering 15+ automation workflows with step-by-step setup instructions, AI prompt templates, and ready-to-import n8n JSON files.

👉 Get the Complete n8n + AI Automation Guide

Includes:

15 production-ready workflow templates
ChatGPT + Claude integration blueprints
WhatsApp, email, CRM automation setups
Sales pipeline automation
Content generation systems

7-day money-back guarantee.

If this helped you, drop a ❤️ and share with someone drowning in repetitive tasks!

n8n vs Zapier vs Make: Which Automation Tool Should You Use in 2026?

Danilo Caffaro — Mon, 02 Mar 2026 20:23:25 +0000

After building 50+ automation workflows across all three platforms, here's my definitive comparison.

Quick Answer

n8n → Best for developers and AI workflows (free self-hosted)
Zapier → Best for non-technical users (most integrations)
Make → Best value for complex workflows (visual builder)

The Full Comparison

Feature	n8n	Zapier	Make
Pricing	Free (self-hosted) / $20/mo cloud	Free / $20/mo	Free / $9/mo
Integrations	400+	6,000+	1,000+
AI Capabilities	⭐⭐⭐⭐⭐	⭐⭐⭐	⭐⭐⭐
Visual Builder	⭐⭐⭐⭐	⭐⭐⭐	⭐⭐⭐⭐⭐
Self-Hosting	✅	❌	❌
Code Nodes	✅ Full JS/Python	Limited	Limited
Learning Curve	Medium	Easy	Medium
Error Handling	Excellent	Basic	Good

When to Choose n8n

n8n is my go-to for anything involving AI. Here's why:

1. AI Agent Nodes

n8n has native AI agent nodes that let you build autonomous agents without code. You can:

Connect to OpenAI, Anthropic, Google, or local models
Give agents tools (web search, database, API calls)
Build multi-step reasoning chains

2. Self-Hosting = Unlimited

Self-host on a $5/month VPS and run unlimited workflows. No per-execution pricing.

3. Code When You Need It

Function nodes let you write JavaScript or Python when visual nodes aren't enough. This is a game-changer for complex data transformations.

Real n8n Workflow Example

I built a lead generation pipeline that:

Scrapes LinkedIn profiles (via API)
Enriches data with company info
Generates personalized emails with GPT-4
Sends via SendGrid with tracking
Follows up automatically based on opens

Total cost: $5/month (VPS) + API costs. Same workflow on Zapier would cost $100+/month.

When to Choose Zapier

Zapier wins in one area: simplicity and integrations.

6,000+ app integrations (3x more than competitors)
"If this, then that" logic is dead simple
Great for quick, one-off automations
Tables and Interfaces add database-like features

Best for: Non-technical users, simple workflows, quick setup.

When to Choose Make

Make (formerly Integromat) offers the best visual workflow builder:

Drag-and-drop interface is the most intuitive
Better pricing than Zapier for complex workflows
Excellent error handling and flow control
HTTP/JSON modules for custom API calls

Best for: Complex multi-step workflows on a budget.

My Recommendation

Building AI agents or complex automation? → n8n
Non-technical, need it working in 5 minutes? → Zapier
Complex workflows, budget-conscious? → Make

The $5,400 AI Agent Deal

Here's a real story: someone built a WhatsApp chatbot + Salesforce integration using n8n for a criminal defense lawyer. The deal closed for $5,400 in just 2 calls. The client saves $250K/year on manual intake processing.

This is the power of automation tools — not just saving your own time, but building solutions others will pay thousands for.

Ready to Build?

I've packaged my best automation workflows into ready-to-import templates:

👉 5 n8n AI Automation Workflows — Ready to import and customize.

Compare all automation tools at AI Tools Hub.

ChatGPT vs Claude in 2026: I Used Both for 6 Months — Here's My Honest Take

Danilo Caffaro — Mon, 02 Mar 2026 20:22:57 +0000

I've been using both ChatGPT and Claude daily since mid-2025. After 6 months of heavy usage across writing, coding, analysis, and creative work, here's my unfiltered comparison.

TL;DR

Feature	ChatGPT	Claude
Coding	⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Creative Writing	⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Research	⭐⭐⭐⭐⭐	⭐⭐⭐
Image Generation	⭐⭐⭐⭐⭐	❌
Long Documents	⭐⭐⭐	⭐⭐⭐⭐⭐
Following Instructions	⭐⭐⭐	⭐⭐⭐⭐⭐
Price	$20/mo	$20/mo

Coding: Claude Wins (Barely)

Both are excellent for coding. But Claude has a noticeable edge in:

Understanding large codebases — Claude's 200K context window means it can actually read your entire project
Following complex instructions — When I give multi-step refactoring tasks, Claude stays on track better
Code quality — Claude tends to write cleaner, more idiomatic code

ChatGPT fights back with:

Faster responses for quick questions
Better at explaining concepts to beginners
Plugins/GPTs ecosystem for specialized coding tasks

Real Example

I asked both to refactor a 500-line React component into smaller pieces. Claude identified 7 logical components and created clean interfaces between them. ChatGPT identified 5 and left some tight coupling.

Creative Writing: Claude by a Mile

This isn't even close. Claude writes more naturally, with better:

Voice consistency — it maintains tone throughout long pieces
Nuance — it doesn't default to corporate-speak
Structure — better at organizing long-form content

ChatGPT tends to:

Use the same transitional phrases repeatedly
Default to a "helpful assistant" tone
Over-explain simple concepts

Research: ChatGPT Wins

ChatGPT with web browsing is genuinely useful for research. Claude doesn't have web access (as of early 2026), which limits it for:

Current events and data
Fact-checking
Finding specific resources

Image Generation: Only ChatGPT

ChatGPT has DALL-E built in. Claude has nothing. If you need image generation, ChatGPT is your only option between these two.

The Context Window Factor

This is Claude's secret weapon. 200K tokens means you can:

Upload entire codebases for review
Analyze long legal documents
Process full books for summarization

ChatGPT's 128K is good, but Claude's extra context makes a real difference for professional work.

My Actual Setup

After 6 months, here's how I use each:

Claude for:

All coding tasks
Long-form writing
Document analysis
Complex reasoning tasks

ChatGPT for:

Quick research questions
Image generation
When I need web-connected answers
Voice conversations (mobile app)

The Verdict

If I could only keep one, I'd keep Claude — but it's close. The coding quality and writing ability edge it ahead for my workflow.

The truth is: you should probably use both. $40/month for both subscriptions is still cheaper than any single SaaS tool.

Want to Get More From AI Tools?

I put together a Complete AI Productivity Toolkit with templates, workflows, and prompt libraries for ChatGPT, Claude, and other AI tools.

For a deeper comparison of 50+ AI tools, check out AI Tools Hub — my free comparison site.

I Replaced My $5K/Month Developer With No-Code AI Tools (Here's My Stack)

Danilo Caffaro — Mon, 02 Mar 2026 03:30:50 +0000

The $60K/Year Problem

Last year, I was spending $5,000/month on a part-time developer for:

Building landing pages
Setting up automations
Integrating APIs
Data processing scripts

Good developer. Fair price. But I realized: 80% of these tasks could be automated with no-code AI tools.

Here's the exact stack I use now — and what it costs.

The Stack

1. n8n (Self-Hosted) — Automation Engine

Replaces: Zapier ($50-200/month), developer time for integrations
Cost: Free (self-hosted) or $20/month (cloud)

n8n is the backbone. It connects everything:

Email → AI → CRM pipeline: Incoming email gets classified by AI, routed to the right workflow
Content repurposing: Write once → n8n creates versions for 5 platforms
Lead qualification: New form submission → AI qualifies → books call or sends info

My top 5 automations:

Email classifier + auto-responder (saves 8hr/week)
Content repurposer (write once → 5 platforms)
Lead qualifier (scores incoming leads automatically)
Meeting notes → tasks (records → transcribes → creates Todoist tasks)
Competitive intel monitor (daily scrapes → weekly AI digest)

2. Claude/ChatGPT API — AI Brain

Replaces: Content writers, data analysts, research assistants
Cost: ~$20-50/month (API usage)

I don't use the chat interface for work. Everything goes through the API via n8n:

Structured prompts → consistent outputs
Batch processing (100 items in minutes)
No manual copy-pasting

3. Cursor/Bolt — When I Actually Need Code

Replaces: Developer for simple scripts and landing pages
Cost: $20/month

For the 20% that still needs code:

Landing pages (describe → deploy)
Data processing scripts
API integrations that n8n can't handle

4. Notion — Knowledge Base + Planning

Replaces: Project management tools, wikis, docs
Cost: $10/month

With AI prompts built into templates:

Daily planning (AI-generated priorities)
Decision journal (AI-analyzed patterns)
Content calendar (AI-suggested topics)

The Numbers

	Before	After
Developer	$5,000/mo	$0
Zapier	$200/mo	$0 (n8n self-hosted)
Various SaaS	$150/mo	$50/mo
AI APIs	$0	$50/mo
Total	$5,350/mo	$100/mo
Annual	$64,200	$1,200

Savings: $63,000/year. And honestly? The output quality is better because it's consistent and runs 24/7.

What I Can't Automate (Yet)

Complex business logic / custom features
Debugging weird edge cases
Design work (AI images are improving but not there yet)
Anything requiring human judgment on ambiguous decisions

I still hire developers for these — but now it's 5 hours/month, not 80.

How to Start

Week 1: Audit Your Tasks

List everything you or your team does. Mark each as:

🟢 Automate now — repetitive, rule-based, high-volume
🟡 AI-assist — needs human review but AI does 80% of work
🔴 Keep manual — complex, ambiguous, high-stakes

Week 2: Set Up n8n + First Automation

Install n8n (Docker or cloud)
Connect your email + AI API
Build the email classifier workflow
Watch it run for a week. Tweak.

Week 3: Add Content Automation

Build the content repurposer
Set up Notion with AI-powered templates
Create your first automated weekly report

Week 4: Scale

Add lead qualification
Build competitive intelligence monitor
Document everything (so it doesn't break)

Resources

I've packaged my exact workflows and templates:

🔧 n8n AI Automation Workflows — 5 ready-to-import n8n workflows (the ones described above)
📋 No-Code AI Starter Pack — Complete toolkit: agents + workflows + prompts + Notion templates
💰 Complete AI Toolkit Bundle — Everything above + more, save 60%

The no-code AI revolution isn't coming — it's here. The question is whether you'll ride it or compete against people who do.

What tasks are you automating? Share in the comments 👇

Building no-code AI systems. More at caffaro.gumroad.com

60 AI Prompts That Actually Convert: A Marketer's Cheat Sheet for ChatGPT & Claude

Danilo Caffaro — Mon, 02 Mar 2026 03:30:11 +0000

Stop Writing Prompts From Scratch

Every time you open ChatGPT and type "write me a blog post about..." you're wasting the most powerful tool in marketing.

Good prompts have structure. Great prompts have frameworks built in — so the AI produces consistently excellent output every time.

Here are 12 of my best marketing prompts, organized by funnel stage. Use them today.

TOFU (Top of Funnel) — Awareness

1. Blog Post Generator

Write a [length]-word blog post about [topic].
Target audience: [who]. SEO keyword: [keyword].

Structure:
- Hook opening (personal story or surprising stat)
- 3-5 subheadings with actionable content
- Include at least 2 data points or examples
- CTA at the end: [desired action]
- Tone: [conversational/professional/edgy]
- Reading level: 8th grade

2. Social Media Thread

Create a [X/LinkedIn/Reddit] thread about [topic].

Thread structure:
1. Hook tweet (max 240 chars, creates curiosity)
2. 5-7 value tweets (one key insight each)
3. Summary tweet
4. CTA tweet

Rules:
- Each tweet must stand alone
- Use specific numbers over vague claims
- No hashtags in the hook
- End with a question to drive engagement

3. SEO Meta Description

Write 5 meta descriptions for a page about [topic].
Target keyword: [keyword]. Max 155 characters each.

Requirements:
- Include the keyword naturally
- Create urgency or curiosity
- Include a benefit (not just a feature)
- One should use a number/statistic
- One should ask a question

4. YouTube Video Script Outline

Create a YouTube video outline about [topic].
Target length: [X] minutes. Style: [tutorial/vlog/review].

Include:
- 3 hook options (first 5 seconds — the make-or-break)
- Chapter breakdown with timestamps
- Key visual/B-roll suggestions for each section
- 2 mid-roll CTA placements (natural, not pushy)
- End screen CTA
- Suggested title + thumbnail text (max 6 words)

MOFU (Middle of Funnel) — Consideration

5. Email Nurture Sequence

Write a 5-email nurture sequence for [product/service].
Audience: [who signed up for what].

Sequence:
1. Welcome + immediate value (no selling)
2. Story/case study (social proof)
3. Educational content (teach something useful)
4. Objection handling (address top 3 concerns)
5. Soft pitch with deadline/incentive

Rules:
- Subject lines: max 6 words, curiosity-driven
- Each email: max 200 words
- One CTA per email
- P.S. line in emails 3-5

6. Comparison Page Content

Write a comparison between [your product] and [competitor 1, 2, 3].
Tone: Fair and objective (but we win on [key differentiator]).

Format:
- Feature comparison table (10 features)
- Prose section: "Who should choose [each option]"
- Honest cons of your own product
- Testimonial placement suggestions
- CTA: "Try [product] free"

Do NOT trash competitors. Acknowledge their strengths.

7. Lead Magnet Outline

Design a lead magnet for [audience] about [topic].
Format: [checklist/template/mini-course/swipe file].

Include:
- Title (benefit-driven, specific)
- 5-10 sections/items
- Quick wins (something they can implement in 5 min)
- One "aha moment" that builds trust
- Subtle product mention (max 1, natural context)
- Landing page headline + 3 bullet points

BOFU (Bottom of Funnel) — Conversion

8. Landing Page Copy

Write landing page copy for [product].
Price: [amount]. Target: [audience].

Structure:
- Hero: Headline (benefit) + subheadline (how) + CTA
- Problem section: 3 pain points (use "you" language)
- Solution: How [product] solves each pain point
- Social proof: Testimonial format suggestions
- Features → Benefits (table: feature | so what? | result)
- Objection handling: FAQ (top 5 questions)
- Final CTA with urgency element
- Guarantee statement

9. Sales Email (Cold Outreach)

Write a cold outreach email to [prospect type].
Goal: [book a call / get a reply / demo request].

Constraints:
- Max 100 words (seriously)
- First line: personalized observation about their business
- One specific result you've achieved for similar companies
- Question that's easy to answer (not "do you have 15 min?")
- No attachments, no links in first email
- Subject line: max 4 words, lowercase

10. Product Description

Write a product description for [product].
Platform: [Gumroad/Etsy/Amazon/website].

Include:
- Headline: [Specific Result] + [Timeframe] + [Method]
- Bullet points: 5 benefits (not features)
- "Who This Is For" section (3 ideal customer profiles)
- "Who This Is NOT For" (builds trust)
- What's included (itemized list with quantities)
- Social proof placeholder
- Risk reversal (guarantee)
- Price anchoring (compare to alternatives)

11. Testimonial Request

Write a message asking [customer name] for a testimonial.
Context: They [specific positive result/interaction].

Make it easy:
1. Reference their specific win
2. Ask 3 specific questions (not "can you write a testimonial?")
   - What was your situation before?
   - What specific result did you get?
   - What would you tell someone considering this?
3. Offer to draft something for their approval
4. Keep it under 100 words

12. A/B Test Variations

Generate 5 A/B test variations for [element].
Current version: [paste current copy].
Metric to improve: [CTR/conversion/open rate].

For each variation:
- The copy
- Psychological principle used (scarcity, social proof, curiosity, etc.)
- Hypothesis: "This should improve [metric] because..."
- Expected impact: low/medium/high

Also suggest: what to test next based on results.

The Full Collection

These 12 prompts are from my library of 60 marketing prompts — organized by funnel stage, tested across real campaigns.

The complete pack includes:

🎯 20 TOFU prompts (awareness + traffic)
🔄 20 MOFU prompts (nurture + consideration)
💰 20 BOFU prompts (conversion + sales)
Copy-paste ready for ChatGPT, Claude, or Gemini
Customization variables in every prompt

👉 60 AI Marketing Prompts Pack — Your complete marketing prompt library.

Want everything? Complete AI Toolkit Bundle — 5 products, save 60%.

Which prompt was most useful? Drop a comment — I'll expand on the ones people find most valuable.

Building AI marketing systems. More guides →

My Notion AI Productivity System: How I Use 20 Built-In Prompts to Run My Entire Business

Danilo Caffaro — Mon, 02 Mar 2026 03:29:20 +0000

Why Most Notion Setups Fail

You spend a weekend building the perfect Notion workspace. Databases for everything. Templates for every scenario. Color-coded tags. Relation properties linking to relation properties.

Two weeks later, you're back to sticky notes.

The problem isn't Notion — it's that most setups are passive filing systems. They store information but don't help you think.

Here's what changed everything for me: building AI prompts directly into my Notion workflow.

The System: 3 Core Components

1. AI-Powered Daily Planner

Every morning, I open my daily page and run this prompt:

Based on my current projects [list from database], pending tasks [list from to-do], 
and calendar events [today's schedule], create my optimal daily plan:

1. Top 3 priorities (what moves the needle most)
2. Time-blocked schedule (including breaks)
3. Tasks to delegate or defer
4. One decision I need to make today

Constraints:
- No more than 3 deep work sessions (90 min each)
- Protect 2pm-3pm for admin/email
- Flag anything that's been pending > 3 days

This takes my chaotic to-do list and turns it into a focused action plan in 30 seconds.

2. Decision Journal

Every important decision gets logged with this template:

## Decision: [What are you deciding?]

### Context
- What triggered this decision?
- What are the constraints?
- What do I know? What don't I know?

### Options Considered
1. Option A: [description] — Pros: / Cons:
2. Option B: [description] — Pros: / Cons:
3. Do nothing — Pros: / Cons:

### AI Analysis
[Run prompt]: "Given these options and constraints, what am I likely overlooking? 
What would a skeptical advisor say? Rate each option 1-10 on: 
reversibility, upside potential, downside risk, speed to implement."

### Decision Made: [Option chosen]
### Review Date: [When to revisit]

Six months in, this journal has become my most valuable business asset. I can see patterns in my decision-making, catch recurring mistakes, and learn from what worked.

3. Weekly Review Dashboard

Every Friday, one prompt summarizes my entire week:

Review my activity from this week:
- Tasks completed: [from database]
- Decisions made: [from journal]
- Metrics: [from tracking]

Generate:
1. Wins this week (what went well)
2. Lessons learned (what I'd do differently)
3. Next week priorities (based on current trajectory)
4. Health check: Am I spending time on what matters most?

5 More AI Prompts I Use Daily

Email Drafting

Draft a [professional/casual/firm] email to [recipient context].
Key points: [bullets]. Tone: [warm but direct]. 
Max length: 5 sentences. Include one clear CTA.

Meeting Prep

I have a meeting about [topic] with [who] in [time].
Based on [context/previous notes], prepare:
1. Three questions I should ask
2. Key points to communicate
3. Potential objections and responses
4. Desired outcome

Content Brainstorm

Generate 10 content ideas about [topic] for [platform].
Audience: [description]. Goal: [awareness/engagement/conversion].
For each idea: headline + 1-sentence hook + estimated effort (low/med/high).
Rank by potential impact.

Project Kickoff

I'm starting a new project: [description].
Break it into phases with milestones.
For Phase 1, give me:
- 5 specific tasks
- Estimated hours per task
- Dependencies
- Definition of "done"
- The one thing most likely to go wrong

Retrospective

Project [name] is complete. Results: [metrics].
Conduct a retrospective:
1. What worked? (Keep doing)
2. What didn't? (Stop doing)
3. What surprised us? (Learn from)
4. What would I do differently with 2x budget? 0.5x budget?
5. Three-sentence summary for future reference.

The Complete System

This article shows the framework. The full system includes:

20 optimized AI prompts built into Notion templates
Daily Planner with time-blocking and priority scoring
Decision Journal with pattern tracking
Weekly Review dashboard with trend analysis
Project templates with built-in AI checkpoints
Ready to duplicate — just copy to your Notion workspace

👉 Notion AI Productivity OS — Your AI-powered command center.

Or save 60% with the Complete AI Toolkit Bundle — includes this + 4 more products.

Start Simple

You don't need all 20 prompts on day one. Start with:

Morning planning prompt — 5 minutes to plan your day
Decision journal — log your next 3 important decisions
Friday review — 15 minutes to close the week

Three habits. Three prompts. That's the foundation.

What's your Notion setup like? I'd love to hear what works for you. 👇

Building AI-powered productivity systems. Follow for more practical guides.

How I Built a Personal AI Agent Team That Saves Me 20+ Hours/Week

Danilo Caffaro — Mon, 02 Mar 2026 02:34:16 +0000

I Was Using AI Wrong

For months, I was opening ChatGPT, asking a question, getting an answer, and closing the tab. Repeat 50 times a day.

That's like hiring a brilliant employee and only asking them to check the weather.

Then I discovered something that changed everything: AI agents — specialized assistants, each with a clear role, running pre-defined workflows.

Here's exactly how I set it up, and how you can too.

The Problem With "Chatting" With AI

Most people use AI as a search engine with personality. But AI's real power is in structured, repeatable workflows:

A Researcher that finds and summarizes information following a specific format
A Writer that produces content in your voice, every time
A Analyst that reviews data and spots patterns you'd miss
A Devil's Advocate that challenges your decisions before you commit

The difference? Each agent has:

A system prompt defining their expertise and personality
Input/output formats so results are consistent
Trigger conditions so they run at the right time

My Current AI Team (15 Agents)

Here's a quick overview of the agents I use daily:

Core Team

Agent	Role	Time Saved
🔬 Researcher	Deep-dives into any topic, returns structured summaries	3hr/week
✍️ Writer	Drafts content in my voice with specific frameworks	5hr/week
📊 Analyst	Reviews metrics, spots anomalies, suggests actions	2hr/week
😈 Devil's Advocate	Challenges every major decision	Priceless
📋 Planner	Breaks projects into tasks with timelines	2hr/week

Specialized Agents

Agent	Role	Time Saved
📧 Email Drafter	Writes professional emails from bullet points	2hr/week
🎯 SEO Optimizer	Optimizes content for search engines	1hr/week
🐛 Code Reviewer	Reviews code for bugs and improvements	3hr/week
💬 Customer Support	Drafts support responses	2hr/week
📱 Social Media	Creates platform-specific posts	2hr/week

How to Build Your First 3 Agents (30 Minutes)

Agent 1: The Researcher

System Prompt:
You are a senior research analyst. When given a topic:
1. Search for the most recent and authoritative sources
2. Identify key findings, statistics, and expert opinions
3. Present findings in a structured format:
   - Executive Summary (3 sentences)
   - Key Findings (bullet points)
   - Data Points (numbers, stats)
   - Contrarian Views (what the other side says)
   - Sources (with dates)
4. Flag any information that seems outdated or unreliable
5. Suggest 3 follow-up questions worth exploring

Agent 2: The Writer

System Prompt:
You are a content writer specializing in [your niche].
Voice characteristics:
- Conversational but authoritative
- Use concrete examples over abstract concepts
- Short paragraphs (2-3 sentences max)
- Include data points when possible
- Write at an 8th-grade reading level

Process:
1. Read the brief/research provided
2. Create an outline with hook, main points, and CTA
3. Write the first draft
4. Self-edit for clarity and engagement
5. Add formatting (headers, bold, lists)

Agent 3: The Devil's Advocate

System Prompt:
Your job is to challenge decisions and find weaknesses.
When presented with a plan or decision:
1. Identify the top 3 assumptions being made
2. For each assumption, explain how it could be wrong
3. Present the strongest counter-argument
4. Suggest what data would prove/disprove the plan
5. Rate confidence level (1-10) and explain why

Rules:
- Never agree just to be agreeable
- If the plan is actually good, say so — but still find edge cases
- Be specific, not vague ("what if it fails" is not helpful)

Results After 6 Months

20+ hours/week saved on manual tasks
Better decisions — the Devil's Advocate caught 3 bad calls
Zero "prompt fatigue" — every workflow is predefined
Consistent quality — refined prompts produce reliable output

Taking It Further

This article covers the basics. If you want the complete system:

All 15 agent templates (copy-paste ready)
YAML/JSON workflow configs
Integration guides for Notion, Google Workspace, and more
Decision framework for what to automate vs. keep manual

I've packaged everything into a guide: AI Agent Team Guide — 56 pages with ready-to-use templates.

Key Takeaways

Stop chatting, start delegating — give AI a role, not just a question
System prompts are everything — they turn generic AI into specialized agents
Start with 3 agents — Researcher, Writer, Devil's Advocate
Define inputs and outputs — consistency comes from structure
Iterate weekly — refine prompts based on results

What's your experience with AI agents? Drop a comment below — I'd love to hear what's working for you.

Building AI systems that actually work. Follow me for more practical AI guides.

I Built a Dietbox Alternative with AI Body Composition Analysis — Here's How

Danilo Caffaro — Fri, 27 Feb 2026 19:21:04 +0000

What if you could analyze a patient's body composition just from 2 photos?

That's the question I set out to answer when building NutriAssess — and the results surprised me.

As someone who's worked closely with nutritionists, I kept hearing the same frustration:

"I'm paying $30–80/month for tools like Dietbox, and half the features I don't even use."

So I built an alternative. And I used Gemini Vision AI to do something those tools can't.

🩺 The Problem: Nutritionists Are Overpaying

Platforms like Dietbox, Evofit, and NutriSoft are the industry standard in Brazil and growing globally. They do the job — but:

💸 $30–80/month per subscription
📋 Clunky interfaces built for desktop-first workflows
🚫 No AI-powered visual body composition analysis
🔒 Patient data locked in proprietary silos

For solo nutritionists, small clinics, or practitioners just starting out, this is a significant cost with limited flexibility.

🚀 Introducing NutriAssess

NutriAssess is a free-to-try, AI-powered nutritional assessment platform that gives nutritionists everything they need — without the bloated subscription.

🔗 Try it free → https://nutri-assess.vercel.app

💳 Get lifetime access for just $12 → https://caffaro.gumroad.com/l/nutri-assess

🤖 The Tech: Gemini Vision AI for Body Composition

Here's where it gets interesting.

Traditional tools require manual entry of skinfold measurements, circumferences, and other anthropometric data — collected by hand with calipers and tape measures.

NutriAssess adds a 360° AI Visual Assessment powered by Google's Gemini Vision AI. Here's how it works:

Patient uploads 2 photos (front view + side/back view)
Gemini Vision analyzes posture, body fat distribution, muscle definition, and symmetry
The AI generates a structured clinical report with:
- Estimated body fat percentage
- Fat distribution pattern (android/gynoid)
- Muscle mass visual assessment
- Postural observations
- Nutritional risk indicators
- Personalized dietary recommendations

This isn't a gimmick — it's a genuine clinical screening tool that saves 20–30 minutes per consultation.

// Simplified example of the AI analysis call
const response = await geminiVision.analyze({
  images: [frontPhoto, backPhoto],
  prompt: CLINICAL_ASSESSMENT_PROMPT,
  schema: BodyCompositionSchema
});

📊 Scientific Formulas Under the Hood

Beyond AI, NutriAssess implements validated scientific protocols:

Body Fat % — Jackson & Pollock (1978)

// 7-site skinfold formula (men)
Density = 1.112 - (0.00043499 × ΣSkinfolds) + (0.00000055 × ΣSkinfolds²) - (0.00028826 × Age)

// Siri equation (1956) — converts density to body fat %
BodyFat% = (495 / Density) - 450

Basal Metabolic Rate — Harris-Benedict (1919, revised 1984)

// Men:
BMR = 88.362 + (13.397 × weight_kg) + (4.799 × height_cm) - (5.677 × age)

// Women:
BMR = 447.593 + (9.247 × weight_kg) + (3.098 × height_cm) - (4.330 × age)

Total Energy Expenditure

TEE = BMR × Activity Factor
// Sedentary: 1.2 | Light: 1.375 | Moderate: 1.55 | Active: 1.725 | Very Active: 1.9

These are the gold-standard formulas used in clinical nutrition — the same ones Dietbox uses — but now accessible for free.

✨ Full Feature List

Feature	NutriAssess	Dietbox
AI Body Composition from Photos	✅	❌
Anthropometric Calculations	✅	✅
Anamnesis (Patient History)	✅	✅
Patient Portal	✅	✅
Dietary Assessment	✅	✅
Nutritional Plan Generator	✅	✅
Progress Tracking	✅	✅
Price	Free / $12 lifetime	$30–80/month

🔬 Core Modules

1. 360° AI Visual Assessment
Gemini Vision analyzes patient photos for body composition, posture, and fat distribution. Generates a complete clinical report in seconds.

2. Anthropometric Calculations
IMC, % body fat (Jackson & Pollock 3/7 sites), BMR (Harris-Benedict), TEE, ideal weight, lean mass, fat mass — all calculated automatically.

3. Anamnesis Module
Structured patient history: medical background, eating habits, physical activity, sleep, stress, digestive health, and dietary restrictions.

4. Patient Portal
Patients access their own assessments, progress charts, and nutritional recommendations through a dedicated login.

5. Dietary Assessment
Detailed food intake analysis with macros, micros, and meal timing recommendations.

6. Nutritional Planning
AI-assisted meal plan generation based on patient goals, restrictions, and metabolic data.

🏗️ Tech Stack

Frontend: Next.js 14 + TypeScript + Tailwind CSS
Backend: Next.js API Routes + Prisma ORM
Database: PostgreSQL (via Supabase)
AI: Google Gemini Vision API
Auth: NextAuth.js
Deployment: Vercel

📈 The Business Model

I wanted NutriAssess to be genuinely accessible:

Free tier: Core assessment tools, limited AI analyses
Lifetime deal: $12 one-time (no recurring fees, ever)

👉 Grab lifetime access for $12 → https://caffaro.gumroad.com/l/nutri-assess

For context, that's less than a single month of Dietbox's cheapest plan.

🎯 Who Is This For?

Solo nutritionists and dietitians
Nutrition students building their first client base
Small clinics looking to cut SaaS costs
Personal trainers who work with nutrition
Anyone curious about AI-powered health assessments

🔮 What's Next

[ ] Mobile app (React Native)
[ ] TACO/IBGE food database integration
[ ] PDF report generation
[ ] WhatsApp patient reminders
[ ] Multi-practitioner clinic management

Try It Yourself

The app is live and working right now:

🔗 https://nutri-assess.vercel.app

If you're a nutritionist, dietitian, or just curious about what AI can do in clinical settings, I'd love your feedback.

And if you find it valuable:

💳 Lifetime access for $12 → https://caffaro.gumroad.com/l/nutri-assess

Built with ❤️ using Next.js, Gemini Vision AI, and a genuine desire to make nutritional care more accessible.

Have questions about the tech? Drop a comment below — I read everything.

How I Built an AI Course Generator in a Weekend (DuoUniversal)

Danilo Caffaro — Fri, 27 Feb 2026 11:19:40 +0000

How I Built an AI Course Generator in a Weekend (DuoUniversal) Tags: #ai #gemini #react #microlearning Cover image: (to be generated in Phase 2) Canonical URL: (self-published) --- ## The Problem: Duolingo Only Works for Languages I love Duolingo. The gamification, the bite-sized lessons, the streaks — it's genuinely addictive in the best way. But what if I want to learn options trading with the same format? Or stoic philosophy? Or AWS certifications? The Duolingo formula works because it breaks complex topics into tiny, testable chunks with immediate feedback. But the platform is locked to languages. So I built DuoUniversal — Duolingo's format, but for any topic. Powered by AI. 🔗 Live demo: duouniversal.vercel.app 🔗 GitHub: github.com/danilocaffaro/duouniversal --- ## The Tech Stack (Keep It Simple) I wanted to ship fast, so I kept the stack minimal: - Frontend: Vite + React + Tailwind CSS - AI: Google Gemini 2.0 Flash (via API) - State: LocalStorage (no backend yet) - Deploy: Vercel (one-click) Total build time: ~12 hours across a weekend. --- ## How It Works (3-Step Pipeline) ### Step 1: User Describes a Topic Example: "I want to learn options trading" ### Step 2: AI Generates a Course Structure Gemini 2.0 Flash receives a prompt like this:

You are a course designer. Create a structured learning path for: "options trading" Output a JSON array of 10 lessons, ordered from beginner to advanced. Each lesson should have: - title (string) - description (2 sentences explaining what the learner will understand) - difficulty (beginner | intermediate | advanced) - estimatedMinutes (number, 3-7 min per lesson) Focus on concepts that can be taught in micro-lessons with multiple-choice questions.

Response example:

json [ { "title": "What Are Options? (Basics)", "description": "Understand the fundamental definition of stock options and the difference between calls and puts. Learn why traders use options instead of buying stocks directly.", "difficulty": "beginner", "estimatedMinutes": 5 }, { "title": "Strike Price and Expiration", "description": "Learn what a strike price is and how expiration dates affect option value. Understand the relationship between these two critical variables.", "difficulty": "beginner", "estimatedMinutes": 4 }, ... ]

### Step 3: AI Generates Lesson Content + Questions For each lesson, a second Gemini call generates: 1. Lesson content (3-5 short paragraphs, Duolingo-style) 2. 3-5 multiple-choice questions with: - Question text - 4 answer options - Correct answer index - Explanation of why the answer is correct Prompt:

Generate a micro-lesson for: "What Are Options? (Basics)" Include: 1. Lesson text (3-5 paragraphs, conversational tone, Duolingo-style simplicity) 2. 3 multiple-choice questions to test understanding Output as JSON.

Response:

json { "lessonText": "An option is a contract that gives you the *right* (but not the obligation) to buy or sell a stock at a specific price...", "questions": [ { "question": "What does owning a call option give you the right to do?", "options": [ "Sell a stock at any price", "Buy a stock at a predetermined price", "Borrow money from a broker", "Avoid paying taxes" ], "correctIndex": 1, "explanation": "A call option gives you the right to *buy* a stock at the strike price, not sell it." }, ... ] }

--- ## The UX: Duolingo's Formula Once the course is generated, the experience mirrors Duolingo: 1. Progress bar showing lesson completion 2. XP system — earn points for correct answers 3. Streaks (coming soon) — daily learning reminders 4. Levels — unlock harder lessons as you progress 5. Immediate feedback — see the correct answer + explanation after each question All progress is saved to LocalStorage (no login required for MVP). --- ## What Worked Well ### ✅ Gemini 2.0 Flash is Fast Course generation takes 2-5 seconds for a 10-lesson course. Fast enough that users don't get impatient. ### ✅ Structured Prompts = Consistent Output By enforcing JSON output with strict schemas, I avoided 90% of parsing issues. Gemini follows instructions well when you're specific. ### ✅ Vite + Tailwind = Rapid Prototyping I built the entire UI in ~4 hours. Tailwind's utility classes made it trivial to match Duolingo's clean aesthetic. --- ## What I'd Do Differently ### ❌ No Backend = No Analytics I can't track: - Which topics are most popular - Where users drop off - Conversion to premium (once I add it) Next step: Add a lightweight backend (Supabase or Firebase) just for analytics. ### ❌ Gemini Occasionally Generates Bad Questions Some questions are ambiguous or have typos. I need a curation layer where I manually review top topics (e.g., "Python", "Stock Trading") and lock those as "verified courses." ### ❌ No Mobile App The web app is responsive, but a native mobile app would be 10x stickier. Duolingo's streaks work because of push notifications. --- ## Monetization Strategy (The "First Dollar" Challenge) I'm running an experiment: Can I generate $1 from this in 7 days? Here's the plan: 1. Free tier with ads (Google AdSense) 2. Ko-fi "Buy Me a Coffee" button ($3 one-time donation) 3. Premium tier ($7.99/mo) — ad-free, unlimited courses, offline mode 4. Gumroad PWYW — "Support the project" with $0 minimum I'll report back on what actually converts. --- ## Open Source Plans The entire codebase will be open-sourced on GitHub: 🔗 Repo: github.com/danilocaffaro/duouniversal Why open source? - Educational value — Others can learn from the AI integration - Community contributions — Let people add features (e.g., flashcards, spaced repetition) - Credibility — "Here's the actual code" is more convincing than a landing page --- ## Try It Yourself 🔗 Live demo: duouniversal.vercel.app Suggested topics to try: - "Stoic philosophy" - "AWS Solutions Architect certification" - "Copywriting fundamentals" - "How to read financial statements" I'd love feedback — especially on: - Which topics generate good vs. bad lessons? - Would you pay $7.99/mo for ad-free + offline mode? - What features would make this a daily habit like Duolingo? --- ## Next Steps Week 1: - [ ] Add analytics (Vercel Analytics or Supabase) - [ ] Implement Ko-fi integration - [ ] Curate + verify 10 "premium" topics manually Week 2-4: - [ ] Launch on Product Hunt - [ ] Add user accounts (OAuth via Supabase) - [ ] Build streak/reminder system - [ ] Mobile app (React Native or PWA) --- Follow my progress: - Dev.to: @danilocaffaro - GitHub: @danilocaffaro - Gumroad: caffaro.gumroad.com If you build something with this idea, tag me — I want to see what you create! --- Published: 2026-02-27 | Read time: 5 min

5 AI Automation Workflows That Save Me 10+ Hours Per Week

Danilo Caffaro — Fri, 27 Feb 2026 03:35:41 +0000

I used to spend hours doing repetitive AI tasks manually: copying text into ChatGPT, formatting the output, pasting it into Notion, updating a spreadsheet.

Now I run 5 automated workflows that handle this end-to-end. Here's the exact setup.

The Stack

n8n (self-hosted automation) — the workflow engine
OpenAI API / Anthropic API — the AI layer
Notion API — where everything lands
Webhooks — triggers from email, forms, Slack

Total cost: ~$5/month in API calls. Everything else is self-hosted.

Workflow 1: Auto-Research Briefing

Trigger: New row in Notion "Research Queue" database

What it does:

Takes the topic from Notion
Sends it to GPT-4 with a structured research prompt
Returns: 5-bullet summary, 3 sources, key takeaways
Updates the Notion row with the research

Time saved: ~45 min/week (I queue 3-5 topics per week)

// Simplified flow
Notion Trigger → OpenAI Node → Notion Update

The key is the system prompt: "Research this topic. Return exactly 5 bullets, 3 credible sources with URLs, and one contrarian take. No fluff."

Workflow 2: Meeting Prep Generator

Trigger: Calendar event with "prep" tag (via Google Calendar trigger)

What it does:

Pulls meeting details (attendees, topic, context notes)
Generates a prep brief: key questions to ask, potential objections, relevant context
Sends to my Notion Daily Planner

Time saved: ~30 min/week

Workflow 3: Content First Draft Pipeline

Trigger: Webhook from a simple form

What it does:

I submit: topic + angle + target audience + word count
GPT generates a structured first draft
Draft lands in my Notion "Content" database with status "Draft"
I edit (usually takes 15 min vs. 60 min from scratch)

Time saved: ~2 hours/week

The system prompt is opinionated: "Write in a direct, conversational tone. No corporate jargon. Every paragraph must earn its place. Start with a hook, not a definition."

Workflow 4: Weekly Decision Review

Trigger: Every Friday at 4pm (cron)

What it does:

Queries my Notion Decision Journal for decisions due for review
For each: sends the original decision + prediction to GPT
GPT generates a review prompt: "Here's what you predicted. What did you learn?"
Creates a review task in my Daily Planner

Time saved: ~20 min/week (but the value is in the thinking it forces)

Workflow 5: Inbox Zero Processor

Trigger: New email matching certain filters

What it does:

Classifies email: actionable / informational / ignorable
For actionable: extracts the action item, due date, and context
Creates a task in Notion with all extracted info
For informational: summarizes in 2 sentences, archives

Time saved: ~3 hours/week (email is the biggest time sink)

Why This Beats Manual AI Use

The difference between using ChatGPT manually and running automated workflows is the same as the difference between carrying water in buckets and building a pipeline.

Manual AI use has three hidden costs:

Context switching — you leave your actual work to go prompt AI
Prompt fatigue — you spend mental energy figuring out what to ask
Output handling — you manually move the result where it needs to go

Workflows eliminate all three. The AI runs in the background. The output lands where you need it.

Getting Started

If you want to build something similar:

Start with ONE workflow — the one that saves you the most repetitive time
Use n8n (open source, self-hosted) or Make.com (hosted, free tier)
Keep prompts specific — vague prompts = vague output
Connect to where you work — Notion, Google Sheets, Slack, whatever

The Complete Setup

I've packaged my agent templates, prompts, and workflow configurations:

AI Agent Team Guide — 15+ agent templates, workflow configs, setup guide
Free templates on GitHub — 5 agent templates to get started

Both save you the weeks I spent figuring out what works.

Building AI-powered productivity systems. Writing about what actually works, not what sounds impressive.