Start with Meh, aim for Wow, be happy with Okay

Last updated at June 30, 2025•💊 product-thinking ⚡ no-code

A few weeks ago, I watched a great talk from a Polish tech conference Jakub Mrugalski - A gdyby tak zrobić startup w weekend? . This post summarizes insights from Jakub Mrugalski’s talk. I’ve adapted, rephrased, and expanded on them here in English.

Quantity over quality

Peter Levels built 70 projects. Only 5 are successful. Success is a result of statistical luck, not perfect planning.

Every two weekends — one new project.

I need 6 weeks just to pick a name 🐒

Don’t ask friends for feedback

Your friends will lie to not hurt your feelings. Your enemies will lie so you fail. Your mom will just say: ‘Get a real job.’

Building a small audience — like on Twitter or Instagram — with people from your niche works much better. The speaker surrounds himself with devs and sysadmins, and asks them if his idea makes sense.

Step	Beginner’s Mistake	What to Do Instead
Idea	Start coding immediately	Test interest with a landing page
Landing Page	Build it by hand	Use a builder. Carrd ($19/year) is cheap and fast. Mobirise has an AI module that generates a page from your prompt
Email Form	Pick tools with free limits. Free plans like MailerLite’s cap you at 1,000 contacts — not ideal when you’re testing dozens of ideas.	Pick budget-friendly tools — try Sendy. Don’t code it yourself — waste of time
Email Collection	No criteria for stopping	Set limits — e.g., 1 month and 300 signups
Project Decision	Keep going “because I’ve already started”	Kill it ruthlessly if the audience doesn’t care

I first collect 300 signups. Only then do I scale with ads on Facebook, Twitter, or other platforms. But if I can’t get even 300 from my own audience — better to kill the project. It hurts, yes — but it saves you from wasting time and life.

What seems like a great idea to you might be rejected by real people — ‘We won’t pay for that — it’s stupid.’ Better to find that out early.

Prototyping = Input + Output + Memory + Magic

Tally – tool of choice for user Input

A supercharged version of Google Forms
Unlimited forms, full customization, logic (if-conditions, calculations)
Submissions go to a webhook for further processing

Airtable for Output and Memory

Simple REST API
Automation-friendly (webhooks, auto-deploy)
Looks like Excel, works like a database
Free up to 1,000 rows — or 500k with referrals

Perfect choice for catalogs:

..I wanted to gather all participants on a single page. The original creators did it in a week. I did it in 30 minutes. Mine was ugly — but no one wanted to use theirs. Being first matters.

Magic Glue aka Backend automation

IFTTT: Super easy, good for cheap hacks
Make: Great for business workflows
n8n: Great for devs, self-hostable, powerful
Not Zapier: It’s too expensive

For payments

EasyCart (for beginners)
Sellix (for pros)

Use Rapid API or RapidHub

70% cheaper than official APIs
Unified interface = faster learning
Great for throwaway MVPs

No-code usage examples from Jakub

Jakub wanted to reward users who tweeted about his hosting service. The initial idea was to automate this via Twitter’s official API, but Elon played with API prices and it’s become not a budget-friendly for such task. So, instead, it was done with help of no-code tool that supports Twitter applets for $2.92/month. Jakub created a trigger that watches for a specific hashtag on his account and sends matching tweets to a webhook.

RSS → GPT → Dropbox

To avoid information overload from traking hundreds of RSS feeds via Feedly, a flow was setup where GPT-4 filters for relevance and saves only the most useful articles to Dropbox. This reduced daily reading load from ~1,000 headlines to ~100.

Reddit → Raindrop, filtered by AI

For Reddit, he uses n8n to scan his favorite subreddits every few hours. GPT-4 helps clean up the feed, removing junk and spam. The relevant posts are saved straight to Raindrop.

Typos? Outsourced

His draft texts are sent to GPT-4 via Make, which corrects grammar, punctuation, and tone before publishing.

Auto-posting to 14 platforms

The most ambitious automation was probably a pipeline for publishing news updates:

Pull content from RSS
Generate a preview image using Renderform
Save archive copy in Dropbox
Auto-post to 14 platforms: Facebook, Discord, Twitter, Telegram, LinkedIn, Mastodon, Hey (Polish alt-social), GitHub, and more

The entire thing took under an hour to build.