Show HN: I ranked news websites by speed | Hacker News Reader

Show HN: I ranked news websites by speed

legiblenews.com

320 pointsbradgessler198 comments

I've been working on building "the fastest news website" for a few reasons:

1. I got tired of waiting for news websites to load, so I made a text-only news website that only has major news headlines.

2. I wanted to demonstrate to the world that if you want to build something really fast on the web, you can do it without loads of JavaScript.

3. I wanted to show that you can design something that looks good without having tons of images, etc.

I put together the speed page at https://legiblenews.com/speed to hold my website to be more accountable for speed, but it's also interesting to see how fast other news websites are (or in most cases, are not).

Some feedback I'm interested in receiving:

1. What's your take both on the speed ranking methodology for Legible News?

2. Are my descriptions of the metrics for a non-web developer reasonable? Example of that at https://legiblenews.com/speed/websites/associated-press, and if you click through the links on that table, you see a description like https://legiblenews.com/speed/audits/cumulative-layout-shift

Sorry ahead of time, but I can't fit all news websites on the speed report. I had to target general news websites, not ones for specific niches like HN for Tech. If there's something you think that's missing please post it, but I can't promise that I'll add it.

If you like it, please consider subscribing! Thanks!

Open articleView on HN

Nice idea on legiblenews.com, however it looks to me like it's just using https://en.wikipedia.org/wiki/Portal:Current_events and changing the format? All the articles are from there, verbatim, and you are exactly copying the list of articles - even the categorisation of them.

I'm not sure that a 10% donation to Wikipedia from subscription revenue is quite enough when you are doing quite a simple reformat of all the volunteer work that goes into the current events section of Wikipedia.

You are effectively changing $9.99 for a weekly email digest of the https://en.wikipedia.org/wiki/Portal:Current_events page and giving them $1. I think if you went 50/50 on it I would probably feel it was more fair, you are adding value but I assume thats all automated.

I will give you credit for your 100 score, that page on wikipedia gets an 86 on page speed insights.

If true, and appears it is, if the source for the feed really is Wikipedia, shouldn’t it be credited on the site?

Related Wikipedia page on reusing it’s content:

https://en.m.wikipedia.org/wiki/Wikipedia:Copyrights

It is, see the "License" link at the bottom of pages that display Wikipedia content. Front page currently points at https://legiblenews.com/june-27-2022/license.

Note: there was a bug that I fixed at around noon PST that wasn't display this link for daily news articles, but it was for the contextual articles.

I intend on fully complying with Wikipedia's copyright notices. If you see otherwise please LMK because its a bug.

PST or PDT? I didn't think anyone was using PST anymore :-/.

How about PT :-)

Um, you mean they're not using it until November?

https://www.timeanddate.com/time/zones/pst

The US hasn't used PST to describe current time since March 13th.

Yeah, and then it will be used again starting this November. It hasn't gone away permanently IOW, unless you know otherwise.

… I mean, I use it right now, if I have a timestamp in America/Los_Angeles that's between about November and March. Not all timestamps are right now.

Heck, my coworkers use PST all the time (ah, timezone pun?). One of these days I should honor those timestamps to the letter.

Might just be me, but hiding that information behind a "license" link feels like you want to hide that you are just copying from Wikipedia. In fact, the Wikipedia current events section also has a pagespeed of 100, so what do we need your website for?

Yes! The news content and links from them are from Wikipedia, which is why I donate a portion of revenue to them. The value I'm delivering is speed and focus, going further than Wikipedia, hence your observations from the page speed score. The design of their pages is quite busy and some articles actually take a really long time to load when they're re-built from scratch and not served up from a cache.

Regarding, "not enough", I'm currently losing money on Legible News, so if I increased the amount, I'd lose even more money. For now I consider a 10% donation generous given the current unit economics. I may raise prices to offset these loses, but I'll grandfather in people who subscribe today.

If Legible News reaches a large enough scale I'll consider switching from the current donation model to a Wikimedia enterprise subscription (https://enterprise.wikimedia.com/pricing/) or hiring my own editorial staff to put together the days headlines.

---

Edit: Oops, there's a bug on the news articles that are not displaying the license. If you want to see what this should look like, check out https://legiblenews.com/articles/Mekelle and scroll to the bottom and click on "License".

Fixing this now for news articles.

---

Update: Licenses are now working for news pages. You can see today's at https://legiblenews.com/june-27-2022/license. This broke navigation at the bottom, which I'm now fixing.

So I actually love the idea, even using Wikipedia! Their news/current affairs section is amazing, but not accessible or readable in a "new site" like way. If I were you I would be championing that aspect, making it a key feature, talk about how it is making this amazing resource accessible. But part of that would be to donate a larger portion to them. Just charge $20/year and give Wikipedia half...

The mobile version https://en.m.wikipedia.org/wiki/Portal:Current_events is less busy and arguably better readable/accessible than the desktop version.

I'm not committing to that specific price change and donation reallocation, but I will raise prices in the future (if you subscribe now you'll be grandfathered in) and changing the donation allocation with Wikipedia.

The problem now is that fixed costs are barely covered. Once that hump is achieved I can start playing around with donation allocations more. It will take some time to get there as the cost structures aren't that well established yet.

> I'm currently losing money on Legible News

> The news content and links from them are from Wikipedia

How are you losing money if all of your content is from Wikipedia, and presumably free?

Close to zero income plus hosting fees most likely. If I built something like this I’d expect to lose money on it for awhile as well. I’d also expect to have to deliver a bit more “product” but I feel this guy is having a ton of fun with this project. I’d love to see it grow and develop into something amazing.

In fairness, it doesn’t seem to be offering much at the moment, but I’m genuinely curious about the directions the owner imagines taking the project.

This is accurate.

Current costs are:

1. Hosting - $30-$60/mo depending on traffic levels. The more subscribers, the more this will cost.

2. Workstation - Its fractional, but the workstation I use to build this isn't free.

3. Dev time - This is the big one. If this is to become sustainable, I'd need to build up subscription fees to pay for a team. $200k/engineer, administrative staff, and whatever an editorial staff costs if I go down that route.

Saying "it costs nothing" would be a recipe for eventually shutting down this website since I'd be working towards none of the infrastructure above that's needed for this thing to continue on without me.

I also want to build out a mobile application, which also costs money in the form of App Store fees and dev time.

Keep in mind: I'm not targeting the HN crowd with this, who know how to find CurrentEvents, throw it in their RSS feed, find the optimized mobile version, etc. This was built for people who don't have the time or knowledge to do the digging needed to find and format this content.

What HN found interesting was the relative rankings of news website speed, which I hope brings awareness and results in faster websites for us all.

Hosting - $30-$60/mo depending on traffic levels

Stick it on Netlify and that cost will drop to $0.

> 1. Hosting - $30-$60/mo depending on traffic levels. The more subscribers, the more this will cost.

Does the cost of subscription pay for the hosting costs? It sounds like this is the only current hard cost, so hopefully it's growing slower than the associated subscription revenue!

> Hosting - $30-$60/mo depending on traffic levels. The more subscribers, the more this will cost.

> Workstation - Its fractional, but the workstation I use to build this isn't free.

> Dev time

I mean, no offense, but it sounds like you don't actually have any costs besides hosting, and you'll break even somewhere between 40 and 80 subscribers.

I think what everyone's objecting to is your misrepresentation of the content on the home page as your work. Nothing on the "today's news" or "about" pages credits Wikipedia or its contributors.

Photopea.com was able to keep their hosting costs down to about $50/yr while still serving 7 million people[1]. Their website was hosted statically at a size of 1.8 megabytes. Yours is different of course because you scrape Wikipedia regularly, but there may be a cheaper solution than your current setup out there for you fyi

[1] https://old.reddit.com/r/IAmA/comments/i8j5te/i_made_a_free_...

If OP had 7 million people, they wouldn't be losing money, even if it were serving ads.

I don't think GP was talking about profitability, but about the absolute level of the cost.

You are losing money based on the fact you might have to hire staff one day??

What??

When you say losing money, do you mean opportunity costs from your time developing it?

Because running a website like that should cost basically nothing.

> Regarding, "not enough", I'm currently losing money on Legible News, so if I increased the amount, I'd lose even more money. For now I consider a 10% donation generous given the current unit economics. I may raise prices to offset these loses, but I'll grandfather in people who subscribe today.

This feels in very bad taste to me. As another user pointed out, these articles are being authored and maintained by volunteers, who are donating their time and effort to make information accessible to people. Instead of joining them in that pro-bono effort, you're trying to leverage their work to your own profits. Yes, it requires some up-front work from you to format the articles, but it's a one time cost while the authors are continuously working. And sure, you have to pay for hosting costs, but then you could still run it as a non-profit that only recoups costs and donates the rest to Wikipedia.

If you want to make a business out of this, I'd suggest investing in having people author some original articles. But taking someone else's volunteer work for your own gain is not something I will support. Not wanting to invest your own time and money to make your project work long-term doesn't justify using other people's time and efforts to make your project work long term.

If there's any part of your data pipeline that I misunderstand and rectifies the situation, I'd be happy to have my understanding fixed.

I occasionally make edits to the structure of the original content, but given my limited resources, which is a fraction of my personal time, that's all I can do. In the future if this brings in more revenue I will consider donating more to Wikimedia, hiring editorial staff to contribute back to CurrentEvents, or purchasing a Wikimedia Enterprise subscription.

I do want to acknowledge and be upfront that I won't be able to satisfy everybody's level of fairness, taste, etc, but this is currently what is sustainable for me given the amount of time and resources I have available to put into this project. Some will think this is done in very bad taste, like yourself, and others won't. That's fine! I accept constructive or thoughtful critiques and will reflect on them in the future as things change.

The ease at which you gloss over the fact that your site is simply accepting money for the time and effort of volunter Wikipedia editors is truly innovative, good job.

I don't understand why you would feel that way. Under Wikipedia's license:

> You are free to:

> Share — copy and redistribute the material in any medium or format

> Adapt — remix, transform, and build upon the material for any purpose, even commercially.

This seems like exactly the kind of behaviour they want to encourage. It's not taking anything away from the project, just adding another way to consume the information for those who choose to.

You forgot this part:

> Under the following terms:

> Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.

> ShareAlike — If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original.

> No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Proper attribution is not optional. While OP's license link might put him in the clear legally (although IANAL, it might not) it is can still be questionable ethically and I agree with the other commenters here thinking it is.

I was only addressing the comment that OP is "trying to leverage their work to your own profits". I agree that the attribution could be more explicit on OP's webpage, but I don't see that for/not-for profit really has anything to do with it (legally or morally).

For those that believe Brad's take on Wikipedia's content isn't adding a fair amount of value...

Presumably, if Legiblenews takes off and doesn't add value, Wikimedia is free to compete with it. Wikimedia could form a for-profit company that donates more of user revenue to the foundation and feature that company on Wikipedia. Therefore, for the foreseeable future, Wikimedia effectively retains the option to copy & crush Brad's side-project. And if Legiblenews takes off Wikimedia does nothing at all...they will still almost certainly receive more $$$ than the current state of the world.

In that way, Wikimedia is outsourcing innovation to Brad, who Wikimedia is well-positioned to copy/compete with in the future, and being paid to do so. While Brad loses money.

I'm not 100% sure who is on which side of "fair," but ultimately if it complies with Wikimedia's license I'm good with it. And if you don't like it, why not start something better?

OP sounds like they're honoring the terms of Wikipedia's content license and donating money on top. Why is that a problem?

They're presumably also benefitting from the work volunteers put into Linux, nginx, and whatever else is in their stack, but nobody suggests that websites using open source technologies should be forfeiting all profits.

I generally agree with this and would add another take:

Delivering good news is expensive! The big players are not doing that for the most part, instead doing "newstainment" and I almost never pay attention.

Indie journalists are challenged right now, but are doing News. Most can barely make it.

Your platform could help! Having them author and get some revenue would be something I would pay for. And I do that now via Patreon, Substack, others. It is all too coarse though.

All that said, maybe plenty of people will pay. You are likely to find out. And like I said in my other comment:

Speed is nice, but compelling rules!

Maybe invite those getting views to author original material. Helps you, them, the reader / viewer.

Wikipedia: “please reuse our content”

Random internet user: “such bad taste to reuse Wikipedia content, I should complain on their behalf“

Why are your costs so high?

Multiple edge servers for extra speed?

How the hell are you losing money on this?

Thanks for mentioning and noticing it. I like the original idea (one of those that I thought myself doing back during the pandemic), but seeing it is just a wrapper around another source made it lost its magic, at least for me.

If it introduces more people to Wikipedia's excellent Current Events portal, then I'm glad it exists.

TIL - actually I like the wikipedia's tree like formatting much more than the mirror options mentioned in this thread.

Also it's been done better by https://newsasfacts.com/, which actually does give 50/50 like you suggested.

News as Facts also gets a 95 on page load speed:

https://pagespeed.web.dev/report?url=https%3A%2F%2Fnewsasfac...

See also https://currentevents.email

That site seems supremely... Not optimized for utility by humans.

Wikipedia needs money?

Please don’t complain on behalf of others. Wikipedia has a licence that explicitly allows (even encourages) what OP is doing.

This is completely BS. What is he showing here? That a static copy of wikipedia "news"[0] is faster than a full featured news website. Is this just a contrived ad? My news aggregator also achieves 100% performance[1] but I don't dare to use that info to make such a irresponsible claim. Is also some orders of magnitude easier to be a news curator than a news producer.

[0] Is hard to consider local news like "Two policemen and a polio vaccinator are shot dead by gunmen on motorcycles in North Waziristan, Khyber Pakhtunkhwa, Pakistan" relevant news.

[1] https://gtmetrix.com/reports/www.slowernews.com/XOAMfY1i/

Back in my Newsweek days, we rebuilt the entire site with a blisteringly fast homepage. Proper caching, lazy loading where needed, and fully optimized CSS, images, and javascript. Then the sales team sold a 600px wide takeover ad for scabies medication showing an old dudes infected back.

So it doesn’t matter how fast you build the site. You don’t have control over the things that make it bad.

I was hoping some folks would chime in who built these websites!

Yes, I've run big websites and they're kind of like sea faring vessels -- given enough time with a large enough team, they tend to collect barnacles and become slower over time.

I made the speed rankings so that Legible News wouldn't lose site that speed matters.

Newsweek was more like an active scuttling.

Advertising is related to technology in much the same way a parasite is related to the host

So with ublock origin it would be great?

Do you see why we all use it?

Yeah, same when I was working for Nature. Website was super fast, but then we had to load the (many) ad scripts that the sales team wanted and you’d add 2+ seconds to your page load time.

No way to win. At least I could wrap all the ad locations so they didn’t shift the page when they finally popped in.

I've been in the same situation, and it's one of the few times I've appreciated the bureaucracy and red tape that come with being at a Fortune 500 company. Whenever the sales team comes to me asking for another tracking pixel, I just say "We'll be happy to add this once you've submitted the necessary paperwork and it undergoes the the company-mandated security, privacy, and legal reviews". 90% of the time, I never hear back from them again.

Yeaaaaap. I ended up implementing a library for Newsweek that converted document.write ad units into async writes. I'm still amazed that worked, and that ads used to use document.write.

Also nowadays the marketing guys will want all manner of tracking pixels to measure their marketing efforts

I'm super pleased to see the FT towards the top of the charts. It's a couple of years since I worked on the project[1] to make it faster and it's really great the team have been able to keep it that way.

We bet on measuring site speed with user centered metrics early on which was going against the grain at the time - removing above the fold CSS, are you crazy!? It took a lot of demos to convince people that what we were doing was faster, and that they really needed to trust their own eyes even when the tools disagreed!

Keep it up, folks.

[1]: https://medium.com/ft-product-technology/designing-a-sustain...

Somewhat ironic to host a performance-oriented article on Medium. It took about 1-2 seconds before the text appears (in a mostly text article), even though I have a really good internet connection right now (ping to medium.com average ~20ms, upload/download speed is ~930Mbps). Doesn't seems it's a one time thing either, hard-refresh of the page leads to same behavior.

Tell me about it!

"Chickenshit Minimalism - The illusion of simplicity backed by megabytes of cruft."

https://medium.com/@mceglowski/chickenshit-minimalism-846fc1...

(The article is also available on my personal site - about 18kb all in - for anybody who would like to read it not on a slow and increasingly walled garden website.)

My least favorite part of Medium (and why I actively avoid it) is the incredible amount of recommended threads/articles focused on bad-faith arguments from neonazis or religious extremist views.

"Bad faith" isn't the right characterization. Neonazis and religious extremists believe what they're saying, and they're not trying to con you or humiliate you, they're trying to convince you.

One of the many reasons to use a redirect extension or something to replace medium with scribe, e.g. https://scribe.rip/ft-product-technology/designing-a-sustain... - compare that experience

wow, what a different user experience!

Which redirect extension do you use?

I really didn't realize how much I dislike Medium as a reader until now. I guess I got used to waiting 3-5 seconds to get the content. What's even worse is the constant noise, some of which stays even after the initial loading phase: cookie banner, login with Google, "applaud the article" (or whatever Medium's like button is), stats about the creator and the post, 5 different containers to ask me to read something else immediately...

No wonder most of the time I don't get to read a full article on Medium: they constantly bother me to do anything else but reading the article.

This is awesome! I hope more news websites take notice of this report and clean up their act to make things faster.

Good decision to focus on user perceived performance rather then "Google/SEO performance". Also returning user experience, and actual user experience (using the site app/as a user) is very important meanwhile most only measure first load performance.

great work to you and the team. I absolutely adore the FT website. the subscription is worth every penny and a large part is the user experience for me

I hate to rain on the parade but there's no way FT should be up there with USA today. I clicked these (being the top two). USA today was almost instantaneous. I was excited to get some more serious focus news at what this index said would be a comparable speeed. 8 seconds. 8 seconds to settle down. 3 of which basically blank screen.

Congrats on the job you people have done, it's really super, duper fast compared to most of the other news resources out-there. The first page layout is also quite helpful, you can skim it really fast and get an idea of "what's happening" (I'm doing that two, three times per day with with the ft.com first page).

Odd to see CNN last when https://lite.cnn.com/en exists. Maybe that should be another entry?

I thought about including the "lite" websites, but decided against it since those are not the primary news reading experiences that they're pushing. What I might do is add a link to any news website that has a lite/text-only website.

I'll go ahead and say that yes, this is somewhat self-serving, but I'm hoping this puts pressure on some news websites to get their act together and make their primary news websites faster for everybody.

---

Edit: I've pushed out the updates

1. Individual speed reports now include a link to the text-only sites.

2. Methodology makes a mention of it at https://legiblenews.com/speed/methodology (I still need to refine it; wrote and published in haste)

Why don’t the lite versions count? They are what I use on mobile for primary reading. If the goal is pressuring sites that are too slow you should reward the ones that are quick.

The goal is clearly self-promotion, and not neutral observation.

This is crazy. You made a lightweight, text-only news site, but then refuse to compare it to the other lightweight, text-only news sites.

As well as CNN, NPR has one too: https://text.npr.org

There's also http://68k.news which is particularly nice because it hyperlinks to images in the stories rather than embedding them.

Tangentially https://lite.pbs.org/ also launched this year.

I came here to point out the existence of text.npr.org but it seems I'm way too late!

For some reason these just look unreadable to me, a huge wall of text and just too much information thrown at me at once from all sorts of random topics that I may or may not care about. I actually really enjoy the newspaper style visual layout, even if they use stock photos. It makes it easier to navigate with my mind. The ads and bloat can all be solved by installing an ad blocker.

You're not wrong. This is how the internet used to look like. Modern web layouts and designs evolved to be this way because they result in higher engagement metrics.

There's also the fact that screens have gotten so much bigger since the early web. A wall of text didn't look as much like a wall when it was in a 640x480 window. (Minus a couple of banner ads, of course!)

Well, higher engagement metrics seems like a cynical way to say "more readable" especially given the parent comment's complaint.

You don't need commercial interests to think about UX. The early web looked the way it did mostly due to hobbyism and undeveloped UX standards that we take for granted today.

I didn’t intend to use engagement metrics in a negative way. We also have to note that the audience has changed. The audience of the early web was very different to the average audience today (almost everyone).

Didn't know https://lite.cnn.com/en exists

Looks very clean, thanks for this

Got a 96 for me: https://pagespeed.web.dev/report?url=https%3A%2F%2Flite.cnn....

Oddly enough legible news only got an 86: https://pagespeed.web.dev/report?url=https%3A%2F%2Flegiblene...

Oops that was on the Mobile tab. Both score 100 on Desktop.

Yeah speed varies depending on time of day and whether or not your Heroku instances are being subjected to an influx of HN traffic.

:-)

Cranking up the Dynos to keep up.

Which BTW, that's the funny thing, Heroku is super duper slow, and I still manage to run a reasonable fast news website.

Here's response times under HN load: https://s3.amazonaws.com/bradgessler/Eu1cKE24dpKjTkqNcK3qPqH...

99th percentile is at 175ms.

If your page is only updated once a day, why not render a static html site and put it on a host like netlify?

This. There is no need for a cloud server. Even better get a 10GB port private server from Interserver and stop using Heroku.

Hadn’t heard of Interserver before. How are they?

Really good for the price. I think you can get 150TB bandwidth on 10G port with a dedicated server for around $55/month.

It's annoying the the lite.cnn.com URLs look nothing like the normal ones, e.g.:

https://edition.cnn.com/2022/06/28/entertainment/tom-hanks-f...

https://lite.cnn.com/en/article/h_fdb58e07457b1cd96266160502...

Does anyone know how to automatically map one to the other?

you can't connect that two at the moment. Though there will be an update soon to use the same slug as www.cnn.com

Also, I will say that the normal CNN site is slow to completely load, but decently fast to become interactive. Like on my iPhone, everything above the fold is loaded and interactive reasonably quickly, even if the little blue loader bar does take quite awhile to finish. But if you’re not focusing on that, the site doesn’t FEEL slow, as the key content you want to start reading right away does load decently fast.

Me too :)

https://webperf.xyz/

It's a 3 day running average that is focused on Article pages (not the homepage), against the mobile version of article, tested using Lighthouse.

Each site gets tested 1-3 times a day.

All historical data is available in Google Sheets (linked from main page)

https://docs.google.com/spreadsheets/d/1sGKmbnW74u9r1GOzAQcI...

Hey this is the site I use! It's been excruciatingly difficult to get a higher ranking when you have any ads at all.

Very difficult. It is not just technical but there are business reasons too.

That said, it is possible to make drastic adjustments that help monetization, UX, and still have ads. I did that as a short-lived experiment for a very ad heavy mobile site (SFGate):

Main points are here (not a ton of details) in this sanitized post-redaction deck:

https://docs.google.com/presentation/d/12ds0b4nTxzcDy23te0Zm...

I had an extra labeling for non-ad sites so there was some distinction and would like to restore that.

However look at El Diario - ads and number 1 spot!

DotDash has numerous high performant sites (ThoughtCo etc) that have ads too.

El Diario has to have some kind of crazy loophole. That or some rogue genius has solved a pernicious problem and hasn't shared their wisdom with the world. I see normal google ads and just everything loads so incredibly fast. I'm not a perf expert, but our page speed insights seem to pin almost everything on ads.

I would encourage you to see if you can find out! Someone there should do a talk/post about it

> https://webperf.xyz/

This is the superior tracking-news-website-speed tool.

Nice work on the page and on shipping! What surprises me the most from this is how fast USA Today is given the amount of stuff going on on their page.

I like how you've put the ranking by speed and I think that's a really good way to advertise your service. That being said, $9.99/mo seems pretty steep for what is basically a really fast aggregator of other sources. The $1/mo to Wikimedia is good but really just a small amount going to the creators of the actual content. $25/mo for the NYT does give you a slow website, but you know they're funding investigative journalism with it.

Edit: My mistake, it's $9.99/y, not /mo. That is way more reasonable

Website says $9.99 a year, which would be 83-cents a month, not $9.99 month as you mention.

Source:

https://legiblenews.com/plus

Oh my bad. Yeah that's much more reasonable!

I was also surprised to see USA Today so high on the list. About 10 years ago they did a complete redesign that felt like a SPA and although it was pretty cool at the time, I remember it being very slow and clunky. They've really turned it around!

They got a new team together. The main architect is very detail obsessed about page performance.

Thought the same thing, then clicked over. There’s quite a bit going on for how snappy it feels.

Interesting.

Honestly, I never notice load times from news sites because they're all an advertising cesspool and I run Brave + layered ad blockers on my computer and mobile devices. As a point of curiosity I wonder how these sites would compare with ad blocking enabled.

But at the end of the day I care a lot more about their serious, reliable content (or lack thereof) than a few milliseconds. I'll wait 3 seconds for good news rather than 200ms for trash.

(Not at all saying legiblenews.com is trash. I've honestly never heard of it. I'm just speaking in general terms here.)

Ad blockers are interesting because they do make a pretty big difference.

I wrote a little about that under "Less is more" at https://legiblenews.com/speed/why.

   As of June 2022, CNN.com downloads over 8.9 megabytes of images, videos, and advertisements. With an ad blocker, CNN.com still downloads an incredible 2.4 megabytes of data. All that heft means you have to wait longer for everything to download over the Internet, then wait for your browser to process it all and display it on screen.

I don't really expect slow/fast news websites to be a big deal for the HN crowd since we all know how to run ad blockers in our browsers or PiHole.

My goal here is to make less technical people aware of the speed they're sacrificing when they read websites like CNN, Fox News, etc. There's a lot of privacy implications too I'd like to make people aware of, but I haven't found an easy service/scanner like Google PageSpeed Insights.

"Trash news" is a real problem, and no offense taken to the fact you've never heard of Legible News. I source headlines from Wikipedia, which is why I donate to them for people who buy a subscription. Legible News is more of a news aggregator than anything, which links of to news institutions that people have heard of.

Sourcing your news from Wikipedia is a really smart idea. Looking at your site and the Current Events portal on Wikipedia I see some differences:

https://en.wikipedia.org/wiki/Portal:Current_events

How are you pulling your data? (there are 3 different Wikipedia APIs, with various levels of data).

I suggest you look at the the Features Content API (https://en.wikipedia.org/api/rest_v1/feed/featured/2022/06/2...). This could be fun for you to use, since it also gives you a list of the most visited pages on Wikipedia in the previous 24 hours, which often surfaces interesting topics of current event.

Nice job on a cool project!

News websites with JavaScript disabled can be a pleasure:

- https://www.theguardian.com/uk is fast and snappy

- https://www.bbc.co.uk/news feels like a hyper fast mobile website

Both degrade wonderfully well (minor glitch on the Guardian "most read articles" HTML on mobile but otherwise all good)

Yes! There's so many things the HN crowd does to make websites bearable. It's sad how hostile the web has become to users, especially non-technical folks.

I built Legible News for people who are non-technical and don't know how to disable JS, install ad-blockers, or run a PiHole.

I really appreciate this effort. I think the web needs more of it and it’s why I built something similar over at https://brutalist.report.

Can you share any stats around the conversion rate for how many visitors end up converting to your Plus service?

Have you heard of News as Facts? It does a better job at telling you that the news came Wikipedia and half their profit from subscriptions goes to support Wikipedia. It's also more up to date.

This seems like a https://newsasfacts.com/ clone with less features.

Shouldn't this be considered a news aggregator (e.g., Google News), instead of a news website?

Regardless, I do believe an even simpler aggregator fetching only AP/Reuters could be maintained with something like Brave rewards in the future.

If you're focusing on speed, why any javascript at all? I don't see any images, and the site is only updated once a day? I think you could easily go from the current 272k of resources down to under 10k.

It was like that before when this was a static website, but JS actually speeds up subsequent page loads via Turbo links. The other great thing about Turbo's JS architecture is that it doesn't really block things on the initial load.

Minor nitpick: India Times operates its own portal, which includes news, astrology, classifieds etc. It has much lesser footfall & lighter than Times Of India, which runs as a separate internet entity. A lot of the assets (news, photos, articles) are shared & they have common SSO as well Both are owned by Bennet & Coleman. Funnily enough, TOI runs like a subsidiary to indiatimes going by site organization structure (timesofindia.indiatimes.com)

Indiatimes & Times of India aren't the same, from the internet aspect.

We made a reasonable amount progress in building our own minimalist website aggregator [0] until we gave up.

The plan was too source news from Reddit, HN, Several Newsletters and of course the Wikipedia News Portal. I was proud of the stack (Backend is google sheets) but I underestimated the dedication. To build a news aggregator you have to pick the news links on a daily basis and not skip a day. It takes an hour to just to sift through all the days mess as these days significant world news, in my opinion, often doesn't turn up on the front page anywhere. Also some significant world news have so much frequency you have to actually think what to post and if you should remove redundant past articles.

Just picking the article took me an 1 hour a day and I wanted to summarise the content which would have easily taken my 1-2 hours more.

I respect that OP is charging atleast something for their service, because if they didn't I personally don't see how they can continue running the site.

[0] https://worthreading.netlify.app/

The metrics focus on the website but ignore the browser/client.

I use a text-only browser. It does not support Javascript. I also use TCP clients to make HTTP requests and fetch the HTML, then I read the HTML with the text-only browser. I use stored DNS data so there are no DNS lookups except one authoritative query over the loopback. With the exception of Washington Post, all news sites "load", i.e., download, at about the same speed.

If one is trying to demonstrate how fast the www can be without Javascript, then it seems counterintuitive to exclusively select only big, fat, graphical, Javascript-enabled browsers for the demonstration. Those enormous, complex programs^1 are part of the problem. Keep in mind that popular websites are generally^2 designed to take advantange of "browser features" such as images, CSS, Javascript, and automatic fetching of resources from external sources. "Tech" companies desperately try to make these "complex" websites faster through "engineering", but not by changing the design to obviate the need for features and external resources. The later method works. It has been thoroughly tested by yours truly for two decades.

If one uses a client that avoids images, CSS, JS and autofetching, then websites are generally faster. This method is not suitable for all websites, but it is ideally suited for websites that are textual and meant for reading, e.g., news websites. I cannot read for very long if using a graphical browser, it causes too much fatigue on the eyes. Whereas I can read for extended periods using a text-only browser on the console with zero eye fatigue. YMMV.

1. The Chrome binary, without all the libraries, is something like 150MB. The text-only browser I am using to type this comment is only 1.3MB, compiled statically.

2. HN is one exception.

Images and videos add a lot to the news and in many cases not seeing a visual leads to an incomplete story. Of course text only will be faster, it’s not something that needs to be explicitly pointed out.

Interesting setup. Do you use a command line only in general or only for browsing

In general. Most time is spent on command line in so-called virtual terminals (VTs) with no graphics layer and no GUI. All programming is done without a GUI or IDE.

I do not use a terminal emulator^1 inside X11, or something like X11. I no longer use X11. Nor do I use VMs. I write scripts and compile programs on small, underpowered computers.^1 IME, switching between graphics layer and console text VTs (e.g., Ctrl+Alt+F2) is slow, so I have learned to stay in one context. I prefer VESA textmode on older computers and something like VESA textmode on newer ones.

It would be nice to have a framebuffer instead of GUI where I could view images, watch video, etc. without having to use a different computer or switch contexts.

1. I do use a terminal multiplexer (tmux).

2. If I wrote scripts and compiled programs on a larger, more powerful computer I would be less sure that they would also compile quickly and run fast on smaller, underpowered ones.

%s/VESA/VGA/g

I would love to see these stats for sfgate.com and sfchronicle.com. Both of them are infested with so much added fluff. It's frustrating because I pay for a subscription to SFChronicle (and NYTimes), but unless I want to install adblocking, the experience of trying to read the news is often so user hostile, even on a fast connection and fast computer.

You can run those pages through speed reports at https://pagespeed.web.dev

I don't include in the rankings because their distribution isn't wide enough, but I'd guess they're relatively slow for the reasons you mention.

If you want a text-based format, these exist.

https://text.npr.org

Personally, I do not find speed to be a significant exception or attraction.

Actually getting news is the clear winner!

Thought experiment, two sites:

One is quick, easy to read, but boring, low clarity.

The other is slow, harder to read, but is compelling, high clarity.

Now another thought experiment, and please here me out! This one is from broadcasting:

Two streams or stations. Can be radio, TV, whatever:

One, is boring, but pristine! Quality is off the chart good.

The other, for whatever reason, is compelling! Quality is low. Crappy.

-----

Answer these for both experiments:

Which one do you use?

Why?

Which one do [you think] other people use? [And you might benefit by asking them]

Why?

-----

Results!

In broadcast, the number one criteria was whether it was compelling. Full stop, that is what mattered to most people by a very solid majority. Almost nothing else did. And that played out with almost everyone doing broadcast of any kind moving to offer more lower quality choices to people, while at the same time delivering compelling programming to their audiences.

You have probably heard the phrase, "Content is king!"

For news, this will be less clear, but I bet being fast is analogous to quality in the broadcast realm.

And if I am at all right about that, fast just won't get you there. Don't get me wrong. There will be a set people who go for it, but the bigger numbers will be on compelling.

And that is both a different, and IMHO, a very interesting topic. Public trust in mainstream media "news" is super low.

If you were to deliver compelling news from a variety of sources that have more public trust, you are very likely to attract and keep users. Again IMHO.

Why just accept mediocrity? The slowness of the website has no connection to quality of work that is displayed. So they can be improved separately. It doesn't have to be always choice of bad vs less bad. We should strive for something that is actually good.

Everything costs something.

Optimal profit will come from compelling delivered on good enough.

It is a personal judgement from there.

Secondly, one needs an audience to profit. That means speed is not top priority. Content is.

Got a bit more time.

I agree with you! And there are some of us out there too.

In broadcast, where I live, only a single TV station is going for best quality. All others compromise quality to offer multiple program streams or substations. There are a few in radio doing quality first, with most offering multiple streams.

Online, I often see quality emphasized when it is not compelling in any other way. In this discussion, I feel speed is not a meaningful differentiator. Great news would be!

The rest of my comments are supporting why that is.

Nice, and compelling websites, programs, and the like are super nice and appreciated by me for sure. Sadly, that combo is rare.

It sometimes comes at a small premium too. I will often pay that, given it is modest and covering the added cost and not over the top of the value added.

I would have ranked the speed by the article's page and not homepages. As most of the users they only see the article's page, that page needs most of the attention. Homepage is usually for the hard users and they wouldn't mind a bit slower page than the article's.

Oh I misunderstood the title implication. Nothing wrong with the title, but I thought (or was hoping) you were ranking news websites based on how quickly they publish the news in relation to the trigger event.

Maybe there are solutions for this out there already, ill have to look

Its easy to run a fast website when you don't have to host any ads. You know, the thing that gives media companies their revenue.

And yet plenty of websites that don't have ads are painfully slow. The snark is a little unwarranted in my opinion.

I opened the last entry (cnn) and it actually loads very quickly. The indicator keeps spinning but the content was ready in a second

While LegibleNews does well in speed, to me, a newspaper or a news source needs to be extremely dense. I don't want to scroll forever to painfully be blasted with negative space. A single glance at the page should reveal all headlines. NYTimes and WSJ have a really good density of information. But, my favorite is this tech news site that loads fast and is super dense: https://skimfeed.com/

Nice analysis.

For BBC though it's somewhat unfair - outside of the UK you get redirected to BBC.com rather than BBC.co.uk as the BBC in the UK is a public service and funded via the national TV licence so the UK version doesn't have ads (just like the TV channels & radio stations). Would be interesting seeing a comparison using a VPN/proxy of the UK version vs .com/commercial version.

Also they are testing the bbc.com home page, not bbc.com/news, which probably contains a lot of non-news. Certainly bbc.co.uk does.

Quite right. BBC.co.uk is mostly edge cached. Lighthouse can't test that version because of the redirect.

No love for Brutalist Report? https://brutalist.report/

What's a real shame is that I know that multiple news sites performing poorly on this list specifically hired outside consultants to increase their page speed. I worked at one in the past.

As soon as the outside consultant leaves, everyone pushes their internal initiatives again and the sites get slow and bloated.

I'm getting around 250ms response times for https://legiblenews.com from the UK after multiple refreshes. 500ms occasionally.

I feel like it could be much faster at around 50ms or so with a cloudflare caching setup on the page, may be wrong.

You're not imaging things: I'm running a bit hot right now from HN load (https://s3.amazonaws.com/bradgessler/Eu1cKE24dpKjTkqNcK3qPqH...) w/ ~170ms for the 99th percentile. It doesn't help that I'm running this on Heroku, which isn't known for being super fast.

I'll probably deploy this to https://fly.io/ in the future to a few different geographies around the world so that I can bring latency down for the entire planet while saving myself a boatload in hosting costs.

That is interesting to see.

I wonder if fly.io/heroku matters for the servers however. Caching the webpage with a CDN sounds like most bang for buck right? One request is slow at the start of the day for the news (maybe its a user requesting far away the main server worst case) but every subsequent request is served from close to user CDN. (Not 100% if thats how it works in practice)

Speaking out loud here just to see if server location matters if you have a good CDN setup or something I'm missing which would make requests need to hit the main server.

Why not generate it as a static site?

Very curious that NYT's is so slow given their fairly advanced tech team. They do a lot of neat stuff with that team. Though maybe it's another part of the company that manages the rest of the site. Or maybe it's all their neat stuff that makes it slow.

I am a NYT subscriber and I love what they do online. Some of the one-off special effects are a bit excessively heavy, but I appreciate the creativity and willingness to push the boundaries or explore the design space of the web.

I also get the actual paper delivered in the morning, so if I need something "performant" or have a poor connection I can simply read it in print. So I appreciate that the website offers something more than just an online duplicate of the print experience.

I wonder if the NYT would see an increase or decrease in conversions if they switched to a more performant but less progressive design system. Would more people convert on the fast load time? Or are the special effects diving more conversions (as they did for me personally)? I think they must have done the math and concluded that their business aligns more with the latter.

Is this a Show HN, or a blog post? It seems like OP is trying to get attention for his news website, and perhaps that would be best done by doing a Show HN of that site (not this comparison page, which conveniently lists his website as the top-performer).

Higher pagespeed score does not mean faster. Even a fast website can get a poor pagespeed score. In News, a lot of the pagespeed score boils down to ad networks, TTLs on 3rd party content and, importantly, content shifts after initial paint.

I also made a fast-loading news website. It should be roughly as fast as yours. https://www.thenewseachday.com/

Some newspaper provide subscribers with the PDF version of the printed edition. I subscribe to the WSJ and find reading the PDF instead of the website a more focused, tranquil way of consuming news. Awkward as PDF is on mobile devices, it allows having a "bird view" of today's headlines, by which you can pilot or dive in freely. If the newspaper of your choice doesn't provide such service, you may try access PressReader with your library card; lots of newspapers have digital replica hosted there.

Low latency fiber internet + network wide pi-hole ad blocking + new Apple silicon based laptop makes the entire web feel instantaneous! It’s really quite remarkable.

As someone who hates reading the news, this idea really appeals to me! I'm tempted to subscribe, but I think I wouldn't click on the weekly email -- even one page weekly in my email seems like too much. For me, perhaps a weekly text message with just a few headlines would be better. That seems extremely low effort to check, and if I'm interesting in something I can still follow up on specific items.

I found that whenever I am stuck on browsers without ad blocking I tend to use my home country's public service website. It is amazing compared to the main page of just about all other news outlet, despite containing more rich content than I'd like: https://www.svt.se/

https://upstract.com has done the Wikipedia thing for ages.

Didn't google used to incorperate this into their search rankings? There's no way they still do that today right? Imagine how amazing it'd be if they suddenly turned on the button to downrank pages as a function of how many cookie banners and autoplaying videos etc you have to click away to get to the content.

Do you consider lite editions of sites, such as https://text.npr.org or

https://lite.cnn.com?

I use these often and they come in handy during extreme weather conditions (tornadoes) when cell service is low or impaired.

Not for the rankings, but I do mention them in the footnotes.

Rationale at https://legiblenews.com/speed/methodology toward the bottom.

Ah, awesome. Totally reasonable explanation.

If anyone is looking for a text only free version of this, check out 68k.news, found it here a few months ago.

It would be interesting to see the results when the requester is put behind an ad-blocker like pihole.

in addition to lite.cnn.com please add text.npr.org

I linked to it at https://legiblenews.com/speed/websites/npr

CNN too at https://legiblenews.com/speed/websites/cnn

Surprised that Indiatimes (TOI) made it so far up the list, considering how much stuff is there on their homepage.

Not surprised CNN is the worst offender - it's an absolute dumpter fire of a website, filled with auto-playing ads, videos and low cunning.

BBC should be tested against bbc.com/news, not bbc.com.

They're very different pages/sites.

Thanks! Updated that to point at the URL you suggested. Didn’t do much for their rankings, but I’m pleased it’s pointing at their news website.

Looks cool. Just a note might be worth mentioning you're benchmarking the desktop scores as the mobile ones can be significantly different on pagespeed insights. Maybe show/track both? Seeing as mobile usage is significant

I've helped work on https://www.cbc.ca/lite over the past year or so for similar reasons, very good work on legiblenews! :clap: :clap:

What about CNET? They’ve been super fast for years. Granted it’s not general purpose news, but if you want to feature news websites that manage to load quickly, you should include well-trafficked sites like CNET, IMO.

I had hoped for a comparison of how quickly news events appear on the news sites. But the comparison would only show who copy-pasted the fastest and not who researched the best.

I find the speed of page views even more meaningless. ;)

We built something similar, but focused on cybersecurity news: https://cyber.report/

And we personally curate the articles, no automated scraping :-)

You should add https://dallasnews.com to that list just for laughs. Last time I checked on Page Speed Insights, an article scored a 0.

I think the time to a visible site is pretty important; i dont understand why a site that takes longer to see is faster than a site that i can start looking at sooner. can you explain that?

The lite CNN website is missing from that list: https://lite.cnn.com/en

Man what are you doing wrong Drudge to end up near the bottom, if there ever was a site that should be at the top of this list it's Drudge Report.

Yeah I was surprised by this too because their total page payload is pretty light, but I ran the test a few times and it hovered around in the 40's. They're doing something wrong with their JS or their ad-auctioning script takes way too long.

Details at https://legiblenews.com/speed/websites/drudge-report and of course you can run it yourself and dig in at https://pagespeed.web.dev/report?url=https%3A%2F%2Fdrudgerep...

If you look at it in your own browser, be sure to turn off your ad-blockers. Pretty zippy when you do that.

I think it's a bug when I visit drudge in the browser it extremely does not take 7 seconds until it's interactive

How do you feel about the need to customize or filter content to personal interests? I like the way reuters.com has done it with "MyView".

I’ve thought about it, but I’m not a fan of being able to customize what people see (or don’t see) for major news headlines.

One vision I have is that millions of people from different ideologies read the exact same “front page” of news. I figure if the headlines are not sensationalized and are more factual, legible news has a shot at that.

I have considered breaking apart the pages if there’s ever too much content, like having a Sports and Finance section.

I'd be interested in seeing CBC.ca/news on the rankings, as I imagine they'd be near the bottom and they need a kick in the pants.

Good suggestion! “Largest general news website in $COUNTRY” is worth shooting for.

They did OK scoring a 62.

https://legiblenews.com/speed/websites/cbc-news

hello! Have you tried https://www.cbc.ca/lite? Would be interested in thoughts

I remember the days when the BBC cared about speed and usability. They had a great designer at the time who posted the breakdown of his work.

Oh, we are talking about page load speed, not like "which news source gets the story printed first".

I was really excited about the latter

FYI you have "Al Jazeera" misspelled as "Al Jazerra" throughout the site, though correct in URLs

Thanks! Fixed.

So, when rating CNN, what about their text-only site?

Are you rating any other sites that have both full-media and text-only sites?

Read https://legiblenews.com/speed/methodology

NPR.org is also pretty fast. They’ll also deliver a legible page if you disable JavaScript. No ads to boot.

You might be able to squeeze out some more speed with Brotli and enabling http/2 on Cowboy.

Reuters used to be one of the fastest. They took care of that with their 2021 redesign.

do text.npr.org please :-)

funny, I was very surprised that USA Today was so high in the list. So I typed it into my browser and was greeted *very quickly* by a full page ad modal.

That modal must be baked into the site. It bypasses my ad blocking which blocks the rest of their ads.

Would like to see a comparison with wikinews

This site takes all the content from wikinews. There, that's the comparison.

Reader Mode in Safari doesnt work.

Where are the pictures?

god job

Reply on news.ycombinator.com