Show HN: Peekier – A new way to search the web | Hacker News Reader

Show HN: Peekier – A new way to search the web

peekier.com

120 pointsfivesigma62 comments

Thank you for sharing this, it's a great first impression.

- Clean design - Fast search results with previews - Layout options - User privacy

On the second visit, I noticed the "tags" feature: it's possible to narrow down the search results by suggested key words. Very nice user experience.

One idea: perhaps there can be a description of how to set Peekier as the default search engine (for example: https://duckduckgo.com/install).

I wonder about the practical benefits of having previews, whether it helps me find what I need quicker. The movement of the eyes while scanning through the results feels a little less efficient (scan sideways; look for beginning of next row; repeat) - compared to scanning a list straight down.

Still, the site is intriguing. I'll continue using it to see how it feels.

Another idea, not so related to the search engine feature: I wish there was an "explore" page (like on GitHub) to see what's new/popular/etc.

Thanks. There could be a description on how to set the default search engine as it is a different process on each browser.

You should get some feedback on the actual quality of the search results too :) DDG is plagued with terrible results, at least when localized to "Sweden". At a first glance, it looks like Peekier gives me Google quality results!

Where do you get your data from? Do you use your own crawlers? I wish there was a traditional list based search results page too, then I could actually consider using this. I use Startpage now, but it has some strange bugs that's been plaguing the result pages for years.

I absolutely love the previews. I think those png images take up less bandwidth than the js heavy page load. Plus it's a static image not tracking my every movement.

Another question I have is... who's behind peekier? And how is it funded (servers are expensive, development costs, etc), how does it make money (Google makes money precisely because it doesn't offer those privacy features), and if it doesn't make money what's the agenda behind it?

That would be me. I thought it would be a cool thing to have - previews of a website before you visit it - so I built it. It does not make money but its monthly cost is minimal at the moment.

It is too early to discuss "the agenda" behind it because there is none. If it catches on and becomes popular enough, I have a few ideas on how to make it sustainable without sacrificing any of the privacy features.

Have you built your own crawler or do you rely on something like YaCy or Searx, or in other existing search engines?

A homegrown crawler is used but most search results are interspersed with existing search engine results.

May i ask what is the benefit of using your own as well as an outside solution?

I did a couple of searches for various terms and you had a huge index for the various terms. Surely such an index is only possible by hooking into search results from another engine, eventually this will lead to some sort of issue?

What's really made it difficult for competitors is being able to match Google's index. Bing crawls a lot of pages but is much more picky about what it index?

Also can you say where you are pulling the results from the Yandex API, Bing, Yahoo or Google itself?

Do you really do your own crawling for purposes other than image retrieval? Is it really any good?

I highly doubt that your queries are primarily coming from your own engine. The results are too good. It has good ranking, spelling correction, super large indices, and is fast. For example searching "GTGACCTTGGGCAAGTTACTTAACCTCTCTGTGCCTCAGTTTCCTCATCTGTAAAATGGGGATAATA" works even though it only occurs on a few pages and as the blog post that string came from explains, you need super fancy indexing techniques to handle things like that quickly. You also talk about this as if it is a single-person project, which makes it even less likely you made all this from scratch.

I like the concept and the parts that you undoubtedly make yourself like the UI, image retrieval and caching, are really good. This is a great site don't get me wrong. I just think you should be more forthcoming about where your results are coming from.

Where did I imply that queries primarily come from my own engine? Right now only 15-20% of them are. Your search was routed through Bing.

Oh sorry for not being clearer, I didn't mean to imply you misinformed people. I just thought you should be more transparent about where results are coming from.

For example it would have been nice to see a "results from Bing" message somewhere on the search page, or an item on the about page saying that you use another search engine's results.

It would actually have increased my confidence and impression of your project. I know that you can't have made a fantastic query engine like Bing/Google's without a ton of engineers so you must be using someone else's, and it would have given me a better impression if you said that up front instead of my having to infer that.

Totally unrelated:

oh man. My absolute favorite assignment in undergrad computer science classes involved searching ~50GB of compressed text files containing protein sequences, to see how many times "ATG" or something occurred.

It was a ton of fun. Minimum requirements were to get the correct counts. Then it turned in to a competition to see who could make the fastest solution. You had like 8 machines at your disposal, each with the full dataset, to distribute whatever you wanted.

I think we did it in 3 languages or so (java, erlang, something else...).

You got my interest, care to share a link with the write up?

Based on only spending a couple of minutes with it, I have to say your search engine is the first one that has made me "take notice" in a long time - well, since Google came on the scene.

I definitely plan to check it out more fully when I have some free time - but so far, I'm impressed with it. I like it!

First, I'm impressed.

But I'm having trouble believing that it doesn't cost much. Isn't that lot of bandwidth (getting pages and generating previews), especially if you are doing that on the fly? Or does your crawler generate and cache preview images? If the latter how large is your index?

Is the monthly cost minimal simply because you don't yet have much traffic?

Suggestions:

1. Do a list results view like Google - you're competing on so many levels that you don't want to challenge standards unless you have massive justifications that you're UX is better. I don't think the tiles are better and even if they were a little better it wouldn't be enough to challenge the paradigm.

2. Can you (or do you) restrict search results that have a high AdBlock/PrivacyBadger score? I've always wanted that so much more than I'm concerned about my privacy vis-a-vis my search engine provider?

> that you're UX is better

> you're

The list results was the first I noticed. I for some reason thought I was only getting video/image results. not a huge fan of the results style, though I could see it working for some.

Second was that I get a TON of results for foreign languages: ie, a search for "gw2 wiki" results in wiki-de.guildwars2.com wiki-fr.guildwars.com and ja.gw2.wikia.com .. 3 out of top 10, 5 out of top 15 were non english, and I dont see anywhere to filter that. That might be an edge case (where 33% of results are non-English) but it highlights the shortcoming.

On the note of thinking I was only getting video/image results... I dont see anywhere to filter by result type - which is pretty standard in other search engines now.

Your privacy is of the outmost importance...

You probably meant to say utmost. Outmost could imply a lesser importance or the complete opposite if read into too much.

It will be corrected. Thank you!

That is definitely a very interesting search engine, but I'm personally only interested if this will be open sourced. Currently I host my own instance of https://searx.me which is a search engine aggregator, which only sends the least amount of needed information, it even works with yacy a completely decentralized search engine.

That said there are some interesting ideas here, but I need to use a lot more to compare results to google, so far they've been pretty accurate Screenshots are pretty neat, but can also be distracting. I like what duckduckgo does with its instant answers, giving me a short answer that doesn't distract the overall flow of reading.

Edit: searx even somewhat supports ddg instant answers, though very wip.

It looks great! Where do the results come from? Another search engine (a-las duckduckgo) or is it crawled internally?

This is really cool! I'm honestly sad DDG doesn't use this, because i really like them, but i had no inherent "feature" (personally) that i felt was better than Google Search. This, seems to be better (assuming the search results are even halfway decent).

I could definitely stand for a "no animation mode", though. Ie, screw the css animations. They may be nice to make your site pretty, but when i am trying to ingest lots of results to find the correct one, i want things fast. The animations feel far too slow and cumbersome.

There is a low graphics mode in the settings. May I ask what browser/OS you are using? I've tested most combinations and even with the CSS blur it should be snappy.

OSX Sierra, Chrome, Macbook Pro Retina (.. 2013? can't remember haha).

It's fairly snappy, it's just that any animation just bothers me when i'm trying to click from link to link to link, you know?

Ie, if i'm using Google, i open tab after tab in the background for any sites i need to research. That has near instant feedback, no delay. I know it's different, because i'm not actively looking at the background tab, but effectively for my use case that is the UX you are trying to replace by viewing the content in real time, inside your site. So i need to be able to click, and then click away, with as little delay as possible (instant, preferably). An animation just gets in the way for that, to me at least.

Note that i am more than willing to wait for the page to load. I'm not trying to be unrealistic. I just don't want to wait, any time, to popup already loaded content. I suspect the animation also gives the content time to load the higher res version, so that i can't say much about. If it needs to be, it needs to be :)

I'll check out the low graphics mode, thanks!

Very impressive project! It's results seem pretty good - what's the size of the index?

The previews are nice, but they do tend to catch the eye and make it difficult to quickly filter out unwanted results. It's hard to skim the result titles. I do like the tiles, however.

The clickable suggestions at the top are an awesome idea. It's an improvement from having to start typing again to get suggestions.

Overall, this looks polished and well-designed. Competition in the search engine space is still sorely needed.

Suggestion: Start loading more results when I'm scrolling already. No need to display some text saying I should scroll. It should just work.

And on this note, I use the Vimium browser extension (vim-like keybindings for a lot of things) and scrolling all the way to the bottom doesn't cause more results to appear.

I do expect some breakage when using a niche extension like this, but it seems like it would be fixed for my use case if something like plainOldText's idea were implemented.

edit: Or if the scroll button thing were an a tag, I suppose I could "click" it like any other link from the keyboard. It does load more (somewhat reluctantly) if I scroll down with the arrow keys.

Every time you scroll 20+ CPU cores have to hum away and generate previews of websites (assuming they aren't cached already). So it's mostly a server load optimization.

I really like the look of this. One suggestion - keep the amount of characters in the title the same as google (50-60 chars). Most websites optimize for this length and without it you miss some important info in titles.

Also the description for your site needs to be recrawled :)

https://peekier.com/#!site%3Apeekier.com

Interesting concept. I am keeping a stack of a few million screenshots from the newscombinator, but I am not using it. Actually I thought about deleting them recently because I found the concept of having full-screen screenshots pretty bad. But now that I see it, I like it.

One question tough: How large is your index? I haven't seen any information on that, or did I miss anything?

From https://peekier.com/about

"Peekier (pronounced /'pi·ki·er/)"

If it's spelled exactly as it would be if you wanted it to be pronounced another way, it should be pronounced that way

That's almost a plus: People in the know can feel smug pronouncing it the correct way, while the majority of people pronounce it the obvious way. See: Wikipedia's pronunciation.

I don't get it - how else would it be pronounced? To me it looks like the comparative of a hypothetical adjective "peeky" (compare "peakier"), or perhaps a form of a verb "to peeky" (compare "carrier"). Either way, the pronunciation seems to line up with theirs...

Jesus Christ, though... I searched for "dog" and I'm getting "dog fucks teen porn video" with porn videos among the most relevant results...

(nope, I don't have weird cookies in my browser and it was an incognito session anyway)

Wow, that sucks. Never had anything like that happen during testing. Sorry about that.

I have changed the default safe search setting to Strict.

I don't know, maybe it's wrong localization... I'm in Italy, I see that there are some articles in Italian.

The link is gone, now—good job, I think it's a safer choice.

Is there a way of turning off the screenshot previews? I don't particularly need it all the time and I'd imagine it stresses the engine a bit more than spitting things out in text-based, list format.

I'm afraid not. It's the major differentiator, otherwise it'd be the same as everything else out there.

Are you testing the value or user demand for that feature? I agree with the GP. Screenshots reduce information density, increase bandwidth (load time), and for me take even longer to visually parse than a text result. That is, they interfere with what I consider to be the most important things I need from a search engine. Having screenshots populate over time as they load moves my eyes around the page and acts as a distraction--at least for me.

Google tested a similar "view a screenshot of the search result" feature a while back. You'd hover over a result and it'd show the screenshot in a column to the right. For whatever it's worth, they axed it, although I don't know the reasons why.

I think it works better than text in some types of searches (shopping for example). Distraction isn't really a problem once you've used it for a while. Again, different strokes.

Hovering over the link to get a screenshot is a completely different dynamic and doesn't seem to save much time over just clicking on the link and seeing for yourself. This gives you a high level overview. I just tried it with a search for some information about linux network drivers and I was able to read the text on the screenshots without opening any of them.

Also Google is notorious for putting out half-baked prototypes and quickly abandoning them, so it might very well be that the concept does better with a more dedicated team. I'm going to give this a shot for a few days and see how it is.

I think the 'random things popping all over' problem could be solved if the display of each screenshot was deferred until the previous one had appeared. That way you'd get a continuous flow of things popping up in order of importance, instead of stuff all over.

Hm, you may want to revisit how you preview the results. The 'cards layout' isn't well suited for scanning/comparing multiple options from an ordered list[1]. It removes a clear hierarchy which is the primary directive that a search engine provides (ranking queried content in a list).

I get that it's trying to be different, but it's going against what the users are trying to accomplish. Would be interesting running tests on this for certain query types getting certain layouts, etc. Anyway, good luck!

[1] https://www.nngroup.com/articles/cards-component/

That's surprisingly good at finding what I'm searching for. Great work. Now just get the images loading a lot faster and I might even try to switch from Google.

Cool approach!

If I type in "adult learning", I get no search results at all (no results found). A slightly different search, "Adult education", gives me the results I would expect. Ditto just "learning", just "adult"; so I'm not sure if its an issue in the index, or the strict search filtering..

How does it compare to DuckDuckGo?

Well, for one, it shows you a scrollable preview of every website.

I saw that, that's nice, usability seems better than ddg, but what about privacy? Is it better? similar? worse?

Edit: If you trust Peekier enough, privacy as whole is actually better because you don't even have to click on search results on some occasions, simply because their content is of no interest to you.

More information here: https://peekier.com/privacy

> It is actually better because you don't even have to click on search results on some occasions.

Fwiw, i like the site but that seems like a silly assertion. Yes, some sites could do terrible things with your page visit, but a search engine you frequently use could do far worse things than some random site i visit once.

Not knocking the product, i just dislike that statement, and feel it should be reworded. I understand your intent, the sentence just doesn't convey that accurately, imo.

Agreed, trust is a hard thing to build but you have to start somewhere. I've changed my wording.

This is so refreshing, loving the layout and simple customizations. Would love to see a new search engine in the space :) Good Luck!

Cool, what if adding a timestamp to each result? In Google search I always use timestamp for filtering results visually.

This is certainly something that will be added.

I love it! Keep going. Feels very nice on mobile too. Very good selection of settings, and right where I want them

Looks really nice. Id wish that you could change the layout of the search results though.

Is there a easy way to make this my default search engine on Chrome?

Go to Settings -> Manage search engines -> Other search engines and it should be there.

Reply on news.ycombinator.com