14 Comments
User's avatar
Garrett MacDonald's avatar

Someone will make a public version of the tool, it’ll get scraped up into the next round of AIs’ training data, then they’ll just know how to write the whole program for the next person with that problem immediately

Expand full comment
Scott Werner's avatar

Do you think there are any reasons at all that someone would choose something off the shelf or pre-made over making it themselves?

Expand full comment
Garrett MacDonald's avatar

I think I know what you’re getting at: the costs of building vs buying have to be assessed and you should go with the lower cost. What I’m grasping at is that the costs of building for everyone will continue to fall.

Expand full comment
Scott Werner's avatar

Yeah...I think there are two things going on here:

1) I agree that the costs will continue to fall, but for certain things they'll fall a lot faster as more people are able to do them. I suspect there will probably be some aesthetic dimension to the decision-making as well even beyond just the build vs buy. I can choose to record myself playing a song by my favorite band, I can go see a cover band, or I can go see the band themselves live. Or why do you choose to buy one style of dinner plate over another? Or one type of tile for your bathroom over another?

2) I suspect there will be a separate class of things that the costs fall slower for that you still need expertise to do (maybe larger scale? more complicated?), or expertise to make the costs fall faster than someone untrained. For example, if you look at some of the processes people have invented for working through todo lists with Claude Code vs just naively using it.

Expand full comment
Stef's avatar

Damn...then LLMs just become the OmniApp. It has seen all problems and all solutions. Where does that lead?

Expand full comment
Jamel Eddine Lassoued's avatar

>Now? You need to visualize logs in a specific way for debugging your main product. Four hours with Claude, and you have LogThing. Another hour to add a "pay what you want" dialog. Ship it. Someone else needs exactly this.

I've never been nerdsniped so hard LMAO. I built exactly this thing a few weeks ago for our company's supremely inconsistent log formats and called it "logi". Figured wrapping BigQuery would be easier than convincing management to dedicate time for consolidating logging formats.

Expand full comment
Michael Steamweed's avatar

As with VS Code and a vast market of extensions, curated by # downloads, average review scores, release date(s), etc...

So too with the LLM-written tools. Some emergent marketplace called Gibe Vibe [TM], or whatever. Most tools available for free but always accepting donations. Fraction of them with upfront pricetag.

Expand full comment
James Pember's avatar

I’ve been running a vibe coded tool for grooming my GitHub issues backlog locally for a month or so. Maybe I should just let it out?

Expand full comment
Scott Werner's avatar

Yeah! If you do, I'd love to check it out

Expand full comment
James Pember's avatar

Gimme a few days :)

Expand full comment
James Pember's avatar

https://github.com/jamespember/git-issue-flow Here it is!

Expand full comment
Scott Werner's avatar

Awesome!

Expand full comment
Zachary Conner's avatar

This model really works for desktop apps and various cli-based utilities. Hence why your natural marketing solution was shareware. Even in this world where the cost of software development has cratered, SaaS will still exist in scenarios where what you’re really paying for is hosting, storage, and compute.

Expand full comment
Scott Werner's avatar

I can see that, though I suspect we're going to see SaaS much more unbundled as the cost of software development goes down and it stop being the default. The local first movement is starting to pick up steam, and there are a lot of SaaS products that could be a desktop/mobile app with a simple sync engine for many small to medium sized teams.

Expand full comment