I built and launched 9 startups in 1 year | Hacker News Reader

I built and launched 9 startups in 1 year

tzvc.me

30 pointstheo_champion40 comments

I’m more interested in the tech behind these products. I have a dozen or so “apps” which amount to a python script or localhost site, but deploying them online has been a complete pain in the ass. I will spend ~8 hours perfecting a python script only to spend the next weekend figuring out how to deploy it behind a url. It’s extremely frustrating.

Maybe it’s because I’m not an AWS expert anymore but frankly I don’t want to be. I don’t care. I want to spend my time figuring out the business logic of the app, not tuning obscure IAM configurations so that one part of my app can talk to another.

What I want: take my flask app or node app and my frontend files and run it from a temp url. Include some hosted db. If everything looks good, sell me a domain name. Bonus points if this can be done from terminal like surge.sh can for frontends. And let me cap my monthly spend.

I noticed that one of the apps uses Netlify, are there better options or is that generally considered the best?

I've been impressed by render so far (using them at work, no other affiliation) https://render.com/

Pretty straightforward infra-as-code -- this spec is basically all the documentation you need https://render.com/docs/blueprint-spec

I also felt this way about hosting simple projects on AWS. I tried https://fly.io and I think the experience there is exactly what you’re looking for. Super simple to deploy anything that is dockerized and setting up domains/ssl is the easiest I have experienced from any hosting provider.

I don't think it's that hard to a learn a little bit of AWS. What my "default" is

1. Host the complete React/Vue/static site of choice on netlify

2. Host the backend on AWS. I used to use ECS+Fargate, which means you just create a docker container, open some ports, and git push, but now it's even easier with AWS AppRunner. If you didn't want to touch AWS (it gets hairy fast), I would use something like PlanetScale or CockroachDB which have generous free tiers. From there you have a github action to push your container to your AWS ECR and AppRunner automatically redeploys your app. The only other platform that I think comes close to "I have my code just host it, and make sure it's redundant" is Fly.io

> I don't think it's that hard to a learn a little bit of AWS.

It's a lot more complicated than anything reasonable about deploying a webapp. Any shared hosting solution or deployment platform (think fly/shuttle) will be cheaper and easier to reason about. Even selfhosting on a rented VM or dedicated server and configuring auto-updates and maintenance windows will be much easier.

By going to netlify/AWS you are now tying the destiny of your app to two megacorps, doubling the risk of downtime (and cloud uptime is certainly not as good as dedicated) and other bugs.

Have you used AppRunner? I'm not talking about using AWS RDS, EC2 setting up IAM profiles or anything. AppRunner is fairly new and it's, IMO, simpler than Fly.io for webapps. Likewise, I don't think renting a VM or dedicated server is an option if you just want to deploy your app - if you don't want to think about deployment, networking, SSL provisioning, or failover and just your business logic, the appeal of Fly.io, netlify and Apprunner is just having an endpoint where you can 'git push' and having those platforms take care of everything.

I usually stay on the same stack

- frontend is ReactJs/ Typescript - Auto deployed on Netlify - backend in Express/Nodejs / Typescript - Deployed on Google Cloud Run - AI is usually in python running on Compute Engine VMs - Database auth is Supabase

I master these tools and more importantly, know their limitations. This way I can spin up a new project in days without surprises

CapRover app deployment on Digital Ocean is very cost effective.

It’s easy to deploy on a droplet using the DO scripts, and the droplets can be as cheap as you want them to be with the ability to easily scale up.

If you outgrown the biggest droplet size… congratulations! You made it!

DigitalOcean’s load balancer handles +10k requests per second, +10k simultaneous connections, +250 SSL connections per second for every $12. The ceiling is pretty high.

I love python, but that's the reason I eventually decided to build my services in php instead.

Hosting is dirt cheap, fixed cost, and it's just an rsync away from being deployed.

Vercel is the best option by far for everything you asked for. Only node.js though. Managed DB can use Planetscale or Supabase.

These are not startups, most aren’t even products.

I came to write the same thing. If anything, there is far too much working code to be startups /s.

It’s like most of the garbage on Product Hunt these days!

People selling Monday Workflows and AI ebooks :-/

That was my first reaction, but then on further consideration, some are start-ups, some are features.

How many people would have said Docusign, Okta, or even GitHub weren't start-ups, but just wrappers around existing technology.

> just wrappers around existing technology.

so 99% of SaaS?

I hate to nitpick, but when the https://chunk.run/ start button is typoed "Strat Now", I have to wonder how many corners were cut to launch 9 startups in a year.

“Strat Now” sounds like a button you would find on a guitar learning site. Maybe he reused the landing pages. ;-)

When I encounter typos like these, I assume people are just dyslexic, it’s not a question of “too many websites”. If I made 9 start-up POCs, I‘d never Strat Now. But I might Strat Eventually.

Nice catch! And as you can see, many corners haha

I'm struggling to think of a way that GitHunt service can be used in a way that isn't malicious. I say that because a common grey market is to sell email addresses to spammers, and that's the only target audience I can see using it. Not trying to be a dick, just worried that this service can be harmful for a lot of people.

Agreed, I searched for "elixir" and the first person returned on the list was a self-identified minor from Europe and the e-mail that was listed (partially redacted) was not the e-mail publicly listed on the persons GitHub page in the "Contact Me" section.

The list appears to be non-deterministic as it changed slightly on re-running the search query, but selling the e-mail addresses of people in GDPR protected regions seems like a not-great idea.

Seems like a cool idea with a risky implementation.

This is an idea I've been toying around with; this hard part is just coming up with and/or validating ideas. On top of just cranking things out to keep you sharp, I particularly enjoy just being in contact of people in different industries.

Debating something similar. Always loved the problem solving MVP and had a few small ones in the past but have been looking at launching a wave of micro-SaaS projects even if they are already out there to solve a single problem and stay sharp with new ideas and industries.

How can you attract users when most of these products are not available for trial?

No free trial or sample testing

Are those products or just individual features wrapped within a domain name?

Each one is a different product

You could make a nice living covertly selling the AI "startups" to wealthy folk who just want bragging rights about being founders.

Well their AI product hasn't sold for offers over 10k after >85 days, even when apparently earning 1k a month rev? https://news.ycombinator.com/item?id=35229988

Might be a danger for devs to pile into that space

And how'd it go? :)

This is definitely a entrepreneurial model that will work in the coming years. Establish a bunch of base businesses / MVPs, document them well, schedule recurring conversations with your AI to sync about the topic aka "conversational journaling/logging", then let all the agents get to work based on the new context provided.

Anything that's that easy to setup is going to have a profit trending to 0

You lost me at "schedule recurring conversations with your AI". What do you mean by that?

> about: Founder / CEO - uLog.ai - A conversational journal to help you create detailed timelines for any subject

It means he's got some product idea he's trying to slip in to the discussion.

I imagine each project having an AI agent assigned to them in the future with their own respective personalities that pop up to discuss things at regular intervals. AI can start conversations based on my preferences rather than the on-demand only style of most chat interfaces. Push AI, not pull-only AI.

@theo_champion nice list, but would love to see a write-up on your journey!

You can use dynamite to edit the "strat now" from chunk

Wouldn't trust it if i'm being honest haha

congrats on building/taking shows

1/ what were/are your goals with the projects? 2/ what are some of the biggest lessons?

would love to hear your take on 1/2 if open to sharing

Amazing achievement!

Reply on news.ycombinator.com