Show HN: AutoThink – Boosts local LLM performance with adaptive reasoning

Everyone’s had a project that failed —startup, side project, or ignored feature. What was your recent failure, and what's your take away? Not just "validate early," but some insights that changed your approach. Curious to hear and learn from your experiences.

Comments (13)

ben_w · 2d ago

That there is no correlation at all between (mobile app) code quality and business success.

Not an anti-correlation: I'm not saying bad code is a good business; I'm saying it's no correlation, that it doesn't make much difference either way.

The worst code you can imagine and then some? Business awards and acolades in one case, business went under in another. Latest design pattern, CI, code review? Too slow for market in one case, strong market position in another.

https://benwheatley.github.io/blog/2025/02/26-14.04.07.html

raydenvm · 1d ago

At some point, I also learned that code quality does not really matter. Product quality has an impact, but not code quality. With AI, it will likely matter even less because code quantity dramatically increases at the expense of quality.

cpach · 4h ago

I can see how this can be true for commercial products. Would you say it also applies to open source projects?

yamirghofran · 1d ago

Use dev containers and try to deploy a toy system (all the components but toy code) early on in the project. We did a lot of work and couldn't deploy Qdrant, even though it was working on localhost.

raydenvm · 1d ago

How much time/money would you save then?

apineda · 15h ago

even small projects cost a lot of time ( marketing !! ), better to focus on something you really care about rather than opportunistic items

account-5 · 18h ago

I learned I hate docker and elastic stack.

eyesofgod · 1d ago

To just not bother with side projects. Got too costly too quickly.

raydenvm · 1d ago

Have you ever felt tempted to drop your main job because your side project was growing?

aristofun · 1d ago

1. There is no link between quality and popularity of the product. As long as the product is just good enough - it is only a function of marketing.

2. There is no universal lessons to learn, or wisdom to apply in every situation. Including the point above. Every complex enough situation is unique. “bad generals prepare for the past war” as someone said.

raydenvm · 1d ago

1. What about the concept of minimum loveable products (MLP) versus minimum viable products (MVP)? They say go for MLP and that with MVP, your chances are much lower due to the excessive product and information noise.

ben_w · 17h ago

I say that everyone doing a startup needs to know what the actual minimum in the "minimum viable product" really is for them.

The idea of the MVP is there to say "publish this as soon as you can, don't wait because the market will have moved on and you'll run out of money if you wait for perfection".

MLP is a reaction against this, but I don't know if it's a reaction by perfectionists (who are wrong because then you go bankrupt), or a reaction by people who've tried one too many buggy and unready alpha products that was shipped to test the market before it was ready (who are correct, I'll let you think of your own example here).

aristofun · 22h ago

It’s just bunch of word fluff for people to show/think they are smart. In reality money or/and luck usually wins.