I killed my SaaS after 2 months. Here's what replaced it. - Builder Créatif

Kill the SaaS, build the agent — illustration

I spent 2 months building a SaaS to automate my Reddit posts. Today I killed it. And replaced the whole thing with an AI agent that runs on its own.

Here’s why, and what I learned.

The original problem

I’m an indie dev. I have a portfolio of mobile apps, and a big part of my organic acquisition comes from Reddit. Authentic posts in the right communities, sharing my dev experience, mentioning my apps subtly.

The problem: writing those posts manually takes time. And I don’t have a marketer. So I had the logical idea: build a tool for it.

Growiapp was born. A SaaS that would:

Connect to Reddit via OAuth
Generate posts with the right voice for each subreddit
Manage a publishing calendar
Track performance

Simple, right?

The signal I ignored for weeks

The Reddit connection. Step number one. Create the Reddit app, configure OAuth, handle tokens, refresh tokens, scopes.

I was procrastinating on it spectacularly.

Not out of laziness. I coded 23 stories on Muse Otter during that time. Code doesn’t scare me. But this particular task repelled me. And when something repels you that much, it’s rarely a motivation problem - it’s a signal that the direction is wrong.

What my gut was telling me without me listening: I was building a SaaS to solve an ops problem.

The realization

The click happened when I discovered Upload-Post, a service that exposes a simple API to post on social media. A single CURL call to publish a Reddit post. No OAuth to manage, no tokens to maintain, no connections to monitor.

And then, the question that changed everything:

If I can post on Reddit with a simple CURL, why am I building an entire SaaS to do the same thing?

The answer: reflex. When you’re a dev, your default solution is to code. You have a problem? You build a tool. It’s like asking a surgeon what they recommend - they’ll say “surgery.”

But the problem here wasn’t technical. It was operational. I needed regular Reddit posts, not a SaaS product.

The solution: an agent, not a SaaS

Here’s what I set up in one morning:

A dedicated AI agent (I use a multi-agent system for my projects) with:

A complete Reddit profile: voice, tone, vocabulary, patterns that work per subreddit
Previous post samples as style reference
A “writing fingerprint” that ensures every post sounds human, not marketing
Anti-redundancy rules to never repeat an angle

A cron running twice daily:

9am: a casual post on a community subreddit (manga, gaming, tech) - zero promo, just organic participation
3pm: a discreet promotional post for my app, mentioned once, buried in valuable content

A publishing API (Upload-Post) that handles all the technical parts: Reddit auth, publishing, tracking.

A Telegram notification after each post with the link, so I can check.

That’s it. No dashboard, no database, no user management, no pending items to validate.

The comparison that hurts

	Growiapp (SaaS)	Agent + API
Setup time	2 months (and not done)	1 morning
Reddit connection	Custom OAuth, tokens, refresh	Handled by API
Maintenance	Banned accounts, dead connections, bugs	Zero
Cost	Hosting + dev time	~$10/month (API)
Scalability	Refactor for each new feature	Edit the prompt
Result	0 posts published	Automatic posts from day one

Two months of development versus one morning of configuration. And the result is identical. No, the result is better - because the agent actually runs.

What I learned

1. Don’t productize your internal workflows

If you’re the only user of a tool, you don’t need a SaaS. A script, a cron, an agent - anything that gets the job done without maintenance overhead.

SaaS becomes relevant when other people want to use your tool. Until then, it’s over-engineering.

2. Procrastination is a signal

When you procrastinate on a specific task while being productive everywhere else, the problem isn’t you. It’s the task. Or more precisely, it’s the direction that task represents.

I could have listened to myself earlier and saved 6 weeks.

3. AI agents change the game for workflows

A year ago, this solution didn’t exist. I had no credible alternative to the SaaS. Today, an AI agent can:

Generate content adapted to context
Call APIs
Maintain history to avoid repetition
Run autonomously via a cron

It’s a fundamental shift in how we think about automation. Before: I code a tool. Now: I configure an agent.

4. Kill fast, kill clean

The decision to kill Growiapp took 10 minutes. The execution (migrating intelligence to the agent, configuring crons, testing) took 2 hours.

The hardest part isn’t killing a project. It’s accepting that the invested time is a sunk cost. Those 2 months aren’t coming back. But the next months, I can use them better.

The new pipeline

Cron 9am  → Agent generates casual post → API publishes on Reddit → Telegram notification
Cron 3pm  → Agent generates promo post  → API publishes on Reddit → Telegram notification
Cron 10am → Marketing agent reviews yesterday's posts → Feedback loop

Zero human intervention. I get the notifications, I check from time to time, and I let the system run.

For indie devs who are hesitating

If you’re building an internal tool and you’ve been procrastinating for weeks on a technical aspect, ask yourself:

Am I building a product, or am I productizing a workflow?

If it’s a workflow, look for the simplest alternative. An agent, a script, a third-party API. The world has changed - we no longer need to code a SaaS for every automation problem.

And if you need to kill a project for that, do it fast. The time you free up is worth more than the code you abandon.

Growiapp is dead. My Reddit acquisition is better than ever.

Maybe that’s the most important lesson: sometimes the best feature is deleting the product.