Personal software is not a thing, and won't be for a while

"I vibe coded my own Lovable alternative in 2 hours", "I just killed Slack", "SaaS is dead, I just made my own alternative in 2 days".

I see so many people on Twitter saying that they "vibe coded" their own alternative to popular software like Slack, ChatGPT, or any real popular SaaS. Many are calling it the age of personal software. There are even entire companies based on this premise, like Wabi.

Whilst I love the idea of "personal software", the reality is much more complicated than that. There's a reason that ChatGPT(just the site, not the models) has 100+ people working on it. You can't just replicate that with a 3 hour ralph loop. The result may be close, but it will never have the stability or the little things that ChatGPT may have.

This has been especially hard, if you actually code your own apps. Even with the best AI models, it still takes tons of time to go from idea → working app. A great example is all of these Lovable clones that were popping up on Twitter a few months ago.

When building Clovr, I initially tried to fork the open source Lovable clones, specifically Open Lovable by Firecrawl. I was just appalled at the state of the app. It was slow, buggy, and generations failed more than 50% of the time. The other alternatives either didn't work, or didn't meet our criteria.

So I decided to build our own custom solution, and so much stuff went wrong. Firstly, the agent harness was terrible. Claude would just keep messing up on the most basic things, and this went on for multiple days and multiple runs. We ended up with a 1,000 line hotfix file.

Even with my side-project Portal, an open-source Slack alternative, the chat rendering engine caused so many problems. Getting scrolling to work properly took 3 days of straight work on it(with Claude of course), just to get it to behave properly. As well as markdown rendering, even with an external library, it caused so many problems.

The point I'm trying to make is that software companies are nowhere near dead. Software is getting more and more complex each day, and vibe coding your own inferior alternative for $300 in LLM costs instead of paying $80/year is not worth it.

Software isn't dead, and won't be for a long, long time.