Summary

This last weekend Lovable hosted an impromptu hackathon called the AI Showdown.

I made an entry for the $25K competition.

Its rules are simple:

• Compare head to head the performance of OpenAI vs Anthropic vs Gemini.

So I decided to create a very simple submission that doubles down as an evaluation for future models and other platforms as well.

Namely, create a 2D Pokémon style game called Kanjemon, where the goal is to capture all 25 Kanji (Japanese scripture characters)

Motivation

I make language learning apps.

I also wanted a very simple comparison.

Results

Honestly, it's possible to 1-shot prompt an amazing MVP on Lovable -- no auth, just pure gameplay.

My personal favorite is Anthropic (better style taste), but OpenAI was incredible too (followed all instructions)

I wrote a lengthier write-up here, where I include the prompts and logic behind the submission.

YouTube demo

This flow was born out of the success of its predecessor -- a simple n8n automation that invoked the Fal AI API.

I wanted to go deeper in Veo3's capabilities.

I have to say it's great at generating ads.

Check this (Adidas) ad out !

In a nutshell the workflow is divided into three main phases: 

1. Create Veo3 Video 
2. Wait for Video Processing 
3. Post on YouTube 

Check out the full demo here.

Let me know if you build anything -- link it below !

I have been really active on YouTube in the last 2 months.

My niche is mostly automations using nocode tools, specifically n8n.

Yesterday, I got notice that Veo3 is accessible by API.

So I decided to make a video about : Veo3 Video creation using n8n

I posted on YouTube and Medium.

The result ?

Top of the ranks on Google Search for 'n8n veo3 tutorial'

This wouldn't have happened, if I didn't hustle for it.

Keep making.

Today I got notice that the Veo3 API is publicly available through fal ai.

So I made a video that walks you through.

It's an impressive model BUT quite expensive. It's ~1$ per video second.

Here's the demo!

A bit about the flow.

How It Works

1. A user submits a video prompt through a form.

2. An AI agent (using OpenRouter's Google Gemini 2.5 Flash model) refines the user's prompt into a detailed "script-to-screen" format suitable for video generation.

3. The refined prompt is sent to the FalAI Veo 3 model via an HTTP request to generate the video.

4. Details of the video request, including the date, the refined prompt, and the request URL, are stored in a Google Sheet.

Setup

To run this workflow, you need to set up credentials in n8n for:

* OpenRouter: Generate API key from your OpenRouter account.

* Fal AI: Generate API key from your FalAI account

* Google Sheets: Uses OAuth 2.0. Connect by authenticating your Google account.

Hey guys,

This tutorial walks you through along every step of the way. It's a tedious process, but you can do it.

Reason I made this is because WhatsApp has 1+ Billion users. Lots of clients have asked me to help them with a WhatsApp chatbot.

In summary what I did was:

Sett up the WhatsApp Trigger in n8n

Configure the Facebook/Meta Side

Connect n8n and Setting up the AI

Configure the WhatsApp Send Message Node

Finalize and Test the n8n Workflow

Activate and Live Test the Full System

Here's a full length write up on Medium.

Here's the YT Tutorial.

Given that we use LLM virtually daily, and some of us, have to relay those costs to clients/colleagues, I figured I create a tool for the purpose.

How I made it

1. Crafted a prompt and fed it into Google Gemini 2.5 Pro Research.
2. Created a mockup on Gemini.
3. Revised it using Lovable.
4. Ported the code into VS Code and integrated it into my project.
5. Deploy!

Check it out here.

Also, a small YT demo.

Hoping this story resonates with beginners out there.

In the last 2+ years, I have been freelancing in full-stack design and development.

Having earned ~$8k in that time span, I’d be lying if I said “no money no problems”.

May of 2025 though brought a change in perspective.

I turned to automation out of necessity — working on an intelligent social media listening tool, which demanded a lot backend development.

So, I decided to learn n8n and create YouTube tutorials along the way.

Creating a brand is a marketing goal for 2025.

(Marketing is harder than coding, right?)

Results?

- $114 in Gumroad revenue with a 4.4% conversion in May

- YT Channel subs went from 42 -> 94

Market validation? Is this marketing channel sustainable ?

Absolutely and here’s why.

• I get market feedback on what people want.

Save time in lead generation. SEO — specifically Schema Markups (for rich snippets in Google or AI searches).

• Note that the cost of running an agent can be anywhere between $0.007 to $1.5 per run depending on the complexity.

Outreach is more expensive because of NLP tasks, whereas SEO tends to be cheaper.

• My network and credibility are growing.

As a thank-you to the community, I’m sharing a beginner oriented tutorial on YouTube channel SEO.

It updates the keywords and description to your YouTube channel — to help boost discoverability.

Here’s the flow for the visual folks.

YouTube Channel SEO Workflow

I created an automation for performing marketing outreach using n8n.

My target audience are all the makers who have a digital presence and want to appear on AI searches.

They need to use a schema markup generator tool, but they don't know it yet.

So I made an employee that :

• fetch the top Product Hunt performers,
• scrape their website,
• find target keywords (for Reddit Lead search),
• Propose a schema markup to improve your search engine optimization
• Send me a draft
• I approve (or decline) and then it sends (or reworks it)

Then it checks out for the day.

Check it out here.

Alright, in this YT tutorial, I show you how to extract emails using MindStudio.AI.

• It's a browser extension and meant for low-tech professionals.
• They allow for up to 1000 free executions / month

What it is

If you’ve ever googled a recipe and seen a clean card with the image, cook time, maybe even reviews—all showing before you even hit a website—that’s JSON-LD.

Same thing when a video just shows up as an actual video block in search instead of a plain blue link. That’s structured data doing its job.

JSON-LD is basically a way to label your content so search engines understand what they’re looking at. It doesn’t change how your page looks to users, but it makes a big difference to Google.

Why it's important

This stuff matters because it helps your content get shown better—not just ranked higher, but actually presented better. Cleaner, more useful, more clickable.

Free Tools

Here’s a quick video that breaks it down and shows how to test your Schema for free

If you want to test or create your own Schema Markup, I built this.

Automation

So I decided to use n8n to automate its generation. Here's a tutorial.