Hacker News Reader

Comment by verdverm | Hacker News Reader

> legitmately helpful

Here are several things to ponder on

0. Why is popularity the goal? Can you be happy if it is useful to a small number of people

1. How do you know it is helpful? Open source tends to be organic and if users find it helpful, they will use it and tell others. If it is not getting traction, you might revisit this perception. It could be that it is hard to use and needs better documentation, or it could generally not be useful at solving the problem others have.

2. Do you have a place for a community? Word of mouth is better than trying to market it yourself

3. It takes time for most projects. Those rocket ship projects (and startups) are the exception

View on HN

0. Personal milestone really. Just want to see how much help I can be to others I guess. Hard to say.

1. One of my projects does almost the same stuff a proprietary software does so I am providing a free alternative. My other project is helpful in the sense that it does something no single-application does. Like there exists a site that can view ML models then there exists softwares that can edit and customize layers in a model. My tool does both in one. But yes you are right, maybe I am not selling my product in the right manner. Good idea to think about.

2. No :(. How can I go about creating one?

3. Fair enough

Thank you for these!

1. You are still defining usefulness from your perspective. What do the people in these niches say? If they say nothing, it likely means it is uninteresting or unaligned with how they work

2. Discord and/or Slack are good choices, I've personally written off Slack because they erase history after 3 months, which is really bad for small projects.

Why is customizing layers in a model through a UI helpful? To me, seeing their architecture is one thing, but if I change it I need to train it up and that's going to happen in Python. That is quite niche like robotics.

Free vs paid is often not about features, but about managed services, support, and certifications. If you run it yourself, that means you have more work.

Better does not always win out. There are several open source projects I am a big fan of, because they solve real problems, yet are rejected by industry because they are different and they are in a place of "everyone else is doing it this way" right now, despite the insane masochism this way of doing things represents (to me).

1. For my robotics tool, the response was really good. People enjoyed and even sent feedback and got a couple of issues opened on the repo too. I just think I need more reach for it.

2. Noted, thanks for the heads up!

You can quickly benchmark the model and compare it quickly by moving around/adding layers as you wish. But yeah your point makes sense too.

Makes sense. I too, know of many open-source projects in the same way. But when I compare with other open-source projects in the same niche which offer less features (as an example) getting more popularity. It just makes me think I am doing something wrong so that's why I asked.

How many people do you know who want to view ML models and edit and customise their layers?

Yeah not many. But the site that can only view ML model has over 30k GitHub stars.

Viewing seems useful to fulfill curiosity. But what can you do by editing? Any nontrivial edit to an ML model without retraining will just break it. Maybe you can quantize or reduce context length only.

Reply on news.ycombinator.com