Hey, it's Cindy 🌱 These are the five projects from the reel, with the full build path for each.
btw the waitlist for AI System Club is now open! If you're keen on using AI to automate your content, video editing, job search, business or build AI agents, feel free to sign up (only if you want to hehe). Ps. the waitlist gets first dibs:)
💡 One weekend, one project. The title says five because you should pick from five, not build five. One finished project beats five half-started ones, and each of these is scoped for a genuine two-day build with time left to make it look good.
project one
An AI clone of you
Beginner friendly, about 4 to 6 hours. A web page where visitors talk to "you": your cloned voice answering questions from your own writing. The most shareable thing you can build, and it quietly proves voice AI, a knowledge base, and app deployment in one project.
🎙️ Clone your voice. ElevenLabs' instant voice clone needs about a minute of clean audio. Record yourself reading naturally, no music, no echo. Starter plan is around $5 a month and includes cloning.
🧠 Give it your brain. Export your own writing (posts, notes, a bio, FAQs about you) and load it into a Chatbase bot, free tier. That becomes the knowledge base your clone answers from.
📄 Put it on a page. Ask Lovable to build a simple one-page site with your photo and the chat embedded. Wire the bot's replies through the ElevenLabs API so answers come back in your voice.
paste into Claude
I'm building an "AI clone of me" this weekend: a one-page site where
visitors ask questions and my cloned ElevenLabs voice answers from a
knowledge base of my own writing (Chatbase free tier).
My stack: ElevenLabs Starter, Chatbase, Lovable for the page.
1. Plan the build hour by hour across a weekend.
2. Write the knowledge-base structure: what documents about me to
include so answers sound like me and stay accurate.
3. Draft the system prompt for the bot: my tone, what it should
refuse to answer, and how to say "I don't know" gracefully.
project two
Your AI chief of staff
Intermediate, about 6 to 8 hours. An n8n agent that reads your Gmail and Calendar overnight, triages what matters, drafts replies to the important things, and sends you one prioritized morning brief. The project you'll still be using in month three, which is exactly what impresses people.
⚙️ Set up n8n. Free if you self-host, or start a cloud trial. Connect the Gmail and Google Calendar nodes with your account.
🕐 Build the nightly run. A scheduled workflow pulls unread email plus tomorrow's calendar, sends both to Claude with a triage prompt, and gets back: urgent items, drafts worth sending, and a ranked to-do list.
📬 Ship the brief. Format Claude's output into one email to yourself at 7am. Keep drafts as Gmail drafts, never auto-send.
📌 Never let it send. The agent drafts, you send. That single rule keeps this safe enough to run on your real inbox from day one.
paste into Claude
I'm building a morning-brief agent in n8n this weekend. It reads my
unread Gmail and tomorrow's Google Calendar every night and emails me
a prioritized brief at 7am, with reply drafts saved as Gmail drafts.
1. Design the n8n workflow node by node, in order, with what each
node does and its key settings.
2. Write the triage prompt that turns raw emails + events into:
3 urgent items, 3 drafts worth writing, and a ranked to-do list.
3. Give me a test plan: how to run it safely on my real inbox
without it ever sending anything itself.
project three
A second brain you can chat with
Intermediate, about 5 to 7 hours. A chatbot grounded in your own notes, PDFs, and saved articles that answers with citations pointing back to your files. This is retrieval-augmented generation, the exact pattern behind "chat with our company docs", which is the most requested AI build in business right now.
📚 Gather the corpus. Export your Notion pages, PDFs, and notes into one folder. Quality in, quality out: cut anything outdated before you start.
🗄️ Store it as vectors. Supabase's free tier includes a vector store. Chunk your documents, embed them, and load them in. Claude can write this entire script for you.
💬 Chat with citations. A simple Streamlit page: your question goes in, the top matching chunks come back, Claude answers using only those chunks and cites which file each claim came from.
paste into Claude
Build a "chat with my own notes" RAG app with me this weekend.
Stack: Supabase vector store (free tier), Claude API, Streamlit.
My documents: a folder of PDFs and markdown notes.
1. Write the ingestion script: chunking strategy, embeddings, and
loading into Supabase, with comments a beginner can follow.
2. Write the Streamlit app: question in, retrieved chunks shown,
answer with a citation to the source file for every claim.
3. Tell me how to evaluate it: 10 test questions I should try and
what a good vs bad answer looks like.
project four
A micro SaaS, actually shipped
Ambitious beginner, about 8 to 12 hours. A real web app with accounts, a database, and a Stripe payment link, live on a real URL, solving one niche problem you actually understand. A deployed product a stranger could pay for beats any certificate on a resume.
🎯 Pick a one-job problem. Something from your own work: a rate calculator, a brief generator, an audit checklist. If you can describe it in one sentence, it's the right size.
🏗️ Let Lovable build v1. Describe the app conversationally; Lovable generates the working app with Supabase behind it for auth and data. Free tier gets you a long way, Pro is about $25 if you run out of credits.
💳 Charge for it. A Stripe payment link is the whole monetization stack: no code, just a link on your pricing page. Ship it, post it, and see if one person pays.
paste into Claude
I want to ship a micro SaaS this weekend with Lovable + Supabase +
a Stripe payment link. I work in [your field].
1. Give me 5 one-job app ideas from my field, each described in one
sentence, ranked by how likely a stranger would pay $9 for it.
2. For the one I pick, write the exact build brief to paste into
Lovable: pages, features, data model, and what stays out of v1.
3. Write my launch checklist for Sunday night: domain, Stripe link,
a simple pricing page, and the one place I should post it.
project five
An AI data dashboard
Comfortable with a little code, about 6 to 10 hours. A live, clickable dashboard analyzing a dataset you actually care about: job postings in your field, local rent prices, your own spending. Claude writes the analysis app, you publish it free on Streamlit Community Cloud, and hiring managers get a link instead of a static screenshot.
📦 Get a real dataset. Scrape it with Claude's help or grab a public one. The analysis only lands if the question is real, so pick data you're genuinely curious about.
📈 Let Claude write the app. Hand it the CSV and the three questions you want answered. Ask for a Streamlit app with filters, charts, and a written "so what" section per chart.
🌐 Publish it. Streamlit Community Cloud hosts it free from a GitHub repo. The live URL goes on your LinkedIn, not in a drawer.
paste into Claude
I have a dataset about [topic] as a CSV and I want a live analysis
dashboard by Sunday. Stack: Python, Streamlit, free hosting on
Streamlit Community Cloud.
1. Interview me: ask the 5 questions you need about the data and
what decisions it should inform.
2. Then write the full Streamlit app: sidebar filters, 4 charts
that each answer a real question, and a short written insight
under every chart.
3. Walk me through publishing it from GitHub to Streamlit Cloud,
step by step, assuming I've never used either.
before you start
How to actually finish by Sunday
Pick Friday night Choose the project and set up accounts before the weekend starts. Setup friction on Saturday morning is where most weekend projects die.
Build ugly first Get the thing working end to end before making anything pretty. A working ugly demo Saturday night means Sunday is polish, not panic.
Let Claude drive Every project here assumes Claude writes the code and the configs while you make the decisions. Paste errors straight back into the chat; that's the workflow.
Ship with proof Record a 60-second screen demo the moment it works. The demo is what you post, show in interviews, and send to clients.
🌱 Honest expectation. "Complete this weekend" means a working v1 you can demo, not a polished product. That is genuinely enough: the person who shows up Monday with a working agent is in a different category from the person who watched tutorials.
Want to build all five with support instead of alone? Inside AI System Club I teach people to learn AI from scratch and build real systems like these, step by step. Doors open soon and founding pricing dies at close. 🌱 Join the waitlist (link in my bio too).
the waitlist is now open
AI System Club is launching soon
Build AI systems with me through recorded lessons you can follow at your own pace. Join the waitlist for first access and the founding price.
btw this list gets first dibs the moment doors open :)