Building a search engine for programmers

I agree that you could do better. But..

• The problem is not that intense (Google works kinda okey-ish)
• The market is very limited (~30 million software engineers in the world)
• The solution is very hard (crawling and indexing the entire web)

So, nice idea, but the above 3 criteria for commercially viable product point to suicide.

This reminds me of Dash; https://kapeli.com/dash. I've been using this for searching documentation and code snippets. It has been super useful, and I'm a paying customer. What's nice about it is that it works offline and is super fast since docs are downloaded (and auto updated). And if what you search isn't in the docs, it uses Google right in the app. I believe it can scratch much of the same itch you describe.

I definitely want to see something like this! I would not focus on official documentation though, because most of that already allows navigating through versions etc.

Instead, I would focus on other locations where less context is available. For example, StackOverflow or blog postings might be years old but still seem to be good enough for Google. If an engine could narrow this down, for example according to available versions at the posting time, and rank postings accordingly, this would help tremendously.

I'll give you a concrete example where I keep experiencing such issues regularly. I've been an early adopter of Ember.js, a JavaScript framework. It has seen massive improvements over the years and lots of the information you find online is outdated and not valid anymore. I think this is particularly because the community is smaller and the influx of new content isn't just as big as e.g. for React, so old content isn't superseded so easily.

Of course, there are tricks to narrow your Google results down to what you need. But a search engine that actually makes sense of the context, like version used or post date, and ranks results accordingly, would be super helpful.

Admittedly, I am not sure if this is easy to monetize. Software engineers are good in blocking ads and I am not sure if the pain is big enough to pay for the service.

Great idea!
I know he R language has a pretty good search engine: https://rseek.org/

We saw on the last months many Developer-focused companies go IPO or grow organically very well. I think that it's an excelent market.

Crawling, indexing and understanding what the developer meant it's quite hard. Basically (I'm from Deep Learning field), you need to create a cross-reference corpora with "What the developer searched and found" x "What the code was" (basically 2 language models, and a good optimizer to align them).

Also, did you checked out "Dash"?

Sound awesome, specially code snippets directly on search results.

However for me personally is a compromise between something specific or ease of access, sometimes I find it faster to search on google rather than going to the site directly even if I know is there because in a sense github search is a search engine but google does a better job.

Google might not have best results for everything but they do a good damn job with focus keywords, site url and other search methods which you can narrow down.

Also it is there directly on search bar which I don't see anywhere soon replacing with either duckduckgo or a specific dev search engine because what I will search depends very much on what I am doing.

Where I would use such engine is if it was e.g directly on VS Code or my local environment.

Ranking is a very complex story, for the docs sometimes people do a better job explaining shortly on a tutorial which is majority of how I learn things rather than official documentation.

This is a rough idea on what I think , you might give it a try for me is a first time hearing something like this and would love to test.

I get what your saying, I get how it's a bit of a better mouse trap, not convinced if I'd bother remembering it for a while, I don't see the pain that strong due there might be more at specific times... Possibly if you market it which teachings of development you might get some traction.

Depending on you implementation details it could be a high project, but if you just sublet the searches and reorder them according to your rules, it could be half viable possibly.

Cool idea! It would be cool to have a "docs" tab that only contains documentation, similar to the Google "images" tab.

Whilst I'm not a developer Invision have developed a similar niche targeted search engine for design inspiration which might inspire you in terms of UX rather than content: https://search.muz.li/?utm_source=Muzli_medium&utm_medium=muzli_medium_banner&utm_campaign=search_banner_main