Comment by jackfranklyn
1 day ago
This hits close to home. I've been building tools for bookkeepers and accountants as a side project, and the calculus you're describing - where a subscription becomes a weekend obligation - is exactly why I've tried to keep things genuinely useful rather than sticky.
The cynical approach would be to make the product hard to leave. But that just means you've built a trap, not something people actually want. Eventually they escape and hate you for it.
The test I use: would people recommend this to colleagues even if there's no referral incentive? If the answer is no, I'm probably building something people tolerate rather than something they value.
I doubt LLM-generated software is going to replace more traditional software any time soon, especially when accuracy is pretty important (such as accounting). One thing I learned from years as a PM in a very data-centric organization is understanding data, how it is generated/stored/cut/etc. is very important to getting accurate results.
Where I could see some really interesting results is the marriage of the two. For example, you have a solid data structure that an LLM can generate infinite custom views from.
i think the same, i think backend where data is more prominent is not going anywhere soon. llms produce very bad data structures.
but from good apis, good data, good interface they can generate quite nice frontends.
i guess, frontend as job is going to have a hard time.
also, writing code is not cognitive load, its always reading code. and llms just increase that. so i mostly try to avoid using them.
but i do like researching with them. context free. like googles ai mode, etc. not from my code editor cause then they get biased and suggest stupid sh8t all the time.
https://www.databricks.com is doing this already with data as well as multiple other companies
And I have first hand knowledge of well-known companies building their own tooling because the SaaS offerings have a bad price/feature ratio.
You can pivot your knowledge into building bespoke tools for the same people, just a LOT faster.
The recommendation thing is a nice benchmark, but if you're building hyper-specific tools - why would people recommend them to anyone? If you build a tool for an accountant that does some very niche thing only they're bothered by, why would they recommend to the analyst or receptionist in the company?