Vercel also supports Gatsby deployments. Well, they have an option when setting up the repo integration at least.
My team has just started a Next.js app. I chose Next.js over Gatsby because on the face of it, Gatsby looked like a lot more configuration. I fully expect we'll "disagree" with Next.js at some point but for now I am really liking that it is opinionated, as it it letting us build really quickly.
Can relate, next.js is more progressive while Gatsby get things done so quick at first, but you find yourself in a dark room of settings, let along GraphQL.
I setup my personal blog recently. Tried Gatsby and decided I didn't want to spend the time tweaking it so just used Hugo. Ended up deploying it on Vercel and its been so easy. I have less excuses for not writing blog posts now :)
Could you describe what kind of pain points have you had? I've been building a photo-intensive static personal website lately. I started off with Next.js, but later switched to Gatsby, because so many things just worked out-of-the-box with it. This was especially the case with optimizing and resizing images.
For Gatsby there is gatsby-image, which works super well. For Next.js there is next-optimized-images which is not very mature. The latest version of Next.js also provides next/image, but as far as I have understood, it requires you to run the Next.js backend server, so it is not an option for 100% static websites.
> Instead of optimizing images at build time, Next.js optimizes images on-demand, as users request them. Unlike static site generators and static-only solutions, your build times aren't increased, whether shipping 10 images or 10 million images.
Sometimes I also bumped into issues with Next.js where the framework was trying to send some code to run on the client browser instead of running it during page generation. The division between server and client seemed a bit too hazy for my taste and felt magical in the wrong ways.
Yeah, Gatsby-image was what drew us into the platform to begin with. That's what made it fast to get started.
I found the plugin system confusing, and poorly documented so it was hard to know how to get the most out of a plugin and adjust it to whatever our own funny use case was.
The graphql thing was annoying, we don't use graphql and don't know it and didn't find any advantage in having it around.
It's been a while since I touched it, but every time I need to a put it off because it's not a nice env to work in compared to our "from scratch" react apps
NextJS supports export for 100% static site generation (like Gatsby). Unlike Gatsby, it _also_ supports a server runtime and SSR, and everything in between. Its magic is the good kind that decomposes into sane primitives. Respectfully, it sounds like you might have needed to spend a little more time with the docs.
It does support static site generation, which I was using it for, but for example next/image requires the backend API server. The case I was referring to was such that defining the exact same function in a component file behaved differently than in a page file, I think. In one case the build failed if I remember correctly, because it was trying to build client-side code but the code referred Node.js stuff.
No need to be aggressive. I don't have anything against Next, I just felt much more comfortable with Gatsby.
The common complaint about Gatsby is that it's slow. Once you reach a certain scale of records you're ingesting from Contentful or wherever, Gatsby really starts to chug down trying to infer all their types for GraphQL (I believe this is what causes the slow down).
Gatsby is a weird platform. It's gotten absolutely humongous for a product that is designed to only solve the simplest problems. Netlify has taken it to it's limit by adding redirects and lambdas. But it's still a platform for fancy brochures and blogs.
I agree with this weird mismatch of overcomplexity for the problem it's trying to solve. I did really like though the Gatsby model of ingesting all your source data into GraphQL and then having a standardised way to query and build pages from this. I wish Next.js would learn something from this.
I was using Next.js to build a staticly-generated website (so within the remit of Gatsby) but eventually i outgrew the limitations of static site generation (23k pages and counting...) it was so easy to switch to a traditional server-generated model with Next.js. I'm incredibly greatful for picking up and sticking with Next.js for this.
I think for any commercial site, it's likely cheaper to use a platform like next or nuxt and a CDN rather than choking your build pipeline with static generation. You still get static performance and just offload it to an edge network and CDN costs are racing downwards.
"incremental static rendering" is... server-side rendering, at least that's what I meant. Rendering pages on demand for users as they request them, with (optional) invalidation to render it again later.
The whole JavaScript SSR/SSG/hybrid app scene is changing too fast to be driven by anything more substantive than trendiness. Reminds me of dot com hysteria.
Vercel communications team: I watched your quasi-live streaming presentation/announcement several weeks ago, and could barely fathom what anyone was talking about.
Was it your intention to completely talk over everyone's head by using arcane or invented abstractions? Trying to make investors feel they are onto something really advanced?
Or just caught up in VC/SV speak? Because products that nobody understands have a harder time taking root in the commons.
Edit: Literally there is a question above (top at this time) 'What is Vercel?' - this is my point, a communications failure. If people Googling 'Vercel' and not getting the answer very quickly, then it's a communications problem.
Speak in plain English, describe 'what it is' and the direct material value.
The front-page is pretty unclear, but if you open the documentation page, it is fairly clear: https://vercel.com/docs
> Vercel is a cloud platform for static sites and Serverless Functions that fits perfectly with your workflow. It enables developers to host Jamstack websites and web services that deploy instantly, scale automatically, and requires no supervision, all with no configuration.
Marketing is responsible for communicating value to their users and buyers, not the entire population. Consider the possibility that if you can't wrap your mind around a product even after several explanations, maybe you're not the target audience.
I don't mean this in a negative way, and concede that Marketing often fails to even get their users and buyers to understand.
This is the wrong answer (and maybe a little condescending) because the product is targeted at the widest of tech audiences.
Vercel is failing on communications a little bit.
They don't have some highly specialized tech that 'nobody should understand' - they're literally making tech for the front-end masses, basically 'tech commoners'.
There are literally a handful of people on this page saying 'what does this do?' - how often do you see that on an HN page?
Anyone in tech, should be able to plainly understand what every other tech companies does, within a few sentences.
Some of it might need to be abstracted, and some of it may not fit very well into a sentence - but at least the Wikipedia version should be clear.
Explaining Next - the frame work in slightly more mechanical terms, and then, separating the notion of 'hosted service' would be great. (If that's what it is, because frankly I'm not even sure).
If you go to vercel.com - nowhere does it really say 'what it is'.
The introduction is 'Vercel combines the best developer experience with an obsessive focus on end-user performance.
Our platform enables frontend teams to do their best work.'
That is 'bad communication' because it's completely not informative.
Sometimes marketing speak lingers on 'why it's great' - fair enough, but that statement above isn't even that, it's secondary jargon.
'We enable teams to do their best work' <- no place in communications unless you really need to fill a panel with something.
I did consider that possibility but as someone whose entire career has been involved in various parts of web development I figured I should be able to at least get a general understanding of where they fit in the stack. If someone like me isn’t a potential customer then who is?
Congrats! I've been using vercel for a month now and it's just so nice and easy to use; without sacrificing performance/flexibility. I hope they don't lose the ease of use as they get bigger, like so many other similar products
Hi, serious question what is Vercel? As a developer, what do I use it for? Back-end as a service? Hosting? Database? from reading their homepage several times and looking at their products I still cant tell... can some explain?
With Vercel you can easily deploy a static/SSR app in minutes, either using their NextJS framework or any other JS/TS framework.
It also provides serverless API endpoints capabilities with other languages (Python, Node, Go, Ruby). Most of this comes with a great free tier (including bandwidth which is unlimited).
Team accounts allow your team to collaborate, deploy, run analytics etc seamlessly.
It's great, they've done a great job. I might be forgetting a couple of things, but this is a big part of why they are so successful.
Vercel is a serverless platform like AWS Lambda or Netlify. Run your API endpoints and your client application on it but your database elsewhere. They'll make deploying your website one command but take care of an enormous amount of developer operations, including dev/staging/production environments, secrets management, server scaling, and https certificates.
Vercel is like Netlify, but I would not say it is "like AWS Lambda". Vercel a deployment + hosting platforms for static websites and "serverless" functions, but Lambda extends past what Vercel/Netlify is for.
In terms of actual hosting (removing the build/deployment pipeline), Vercel is a subset of AWS Lambda.
The term "serverless", to me, makes as much sense as "No Software" (Salesforce) did back then: none. I do get the point of "you're not managing any actual servers", but still.
Having had it explained to me later, yes, but when I first heard the term I thought it was a rebranding of standalone desktop applications rather than something that involves servers.
To be more precise:
Vercel is the company (fka "Zeit").
They have a suite of complementary software products and services, including NextJS (a "batteries included" React framework with robust support for SSR and SSG), Now (hosting / CDN / runtime in "serverless" paradigm), Micro, etc.
World-class DX and remarkably powerful tools that somehow combine simplicity and flexibility.
Vercel is the company and the product. Now is the old name of the product. It's been thoroughly re-branded. If you go to now.sh it redirects to a page on vercel.com with no mention of "now". The command-line tool is "vercel", aliased to "vc", and it is configured in vercel.json (no longer now.json). If you go to GitHub and look at the repos containing with "now" in their name, they're archived or deprecated.
Thanks, quite right. (I realized same too late to edit my comment.)
FWIW I liked "Zeit" better than "Vercel" (shrug). But preferences aside, I was annoyed that the new redirects for "now.sh" URLs were -- uncharacteristically for a team famous for attention to detail -- mishandled, e.g. breaking a fantastic preso I'd bookmarked (which grauch used in his React Conf 2017 talks) and which I had cause to reference today in promoting NextJS with a client. It's the only real "footgun" move I've seen them make. Here's hoping it's an anomaly!
Hey chrisweekly. "deck.now.sh" we chose to redirect because it had some very so slightly out-of-date information. Thanks for calling this out. The right thing here would be to add a banner to the presentation instead, giving an option to go visit the documentation or newer presentations.
I had the same question. I’m an experienced developer and I could not figure out the substance of what this product offers from a glance at its front page.
It vaguely seemed like a replacement for react but it’s hard to believe you could raise $40M for that so something else must be going on. I generally was under the impression that the industry was moving away from client side rendering since it’s so slow. At least I browse the web with JavaScript off by default.
I’m using Preact X on a Next project and honestly it was more pain than I expected. Their Preact support is pinned to older versions with a different module structure.
I’ll be releasing my solution as a config wrapper (or offer as a PR for their Preact example) before I launch my site so other folks don’t need to go through the same process.
One of the major selling points of NextJS is precisely to address the issues with CSR (client-side rendering). It offers SSG (static site generation) for build-time rendering -- no client JS required. But also SSR (server-side rendering, ie at runtime), and automatic hydration for CSR too.
Is a tool to deploy frontends super easy. It's called JAMStack, but the main idea is make your life easier. I like it.
If you want to deploy a JS frontend (not just static files, though you can), this automatically handles a lot of things like SSR, SSL, etc https://vercel.com/pricing
Vercel basically makes frontend dev super easy. You can literally type in "vc" in your terminal (after installing their cli) and it'll build and deploy everything you have locally in seconds (think React or Nextjs, a React framework).
You can also use their serverless functions; which add "backend" functionality to your existing frontends. You can deploy these serverless functions in a bunch of languages (see docs) but personally I've used them in Nextjs's baked-in functionality of api routes to basically add an endpoint for mutating/querying to/from my DB for example.
I don't think there's an official recommendation. Any database host will work just fine. However if you want to optimize based on location, here are the regions where Vercel hosts: https://vercel.com/docs/edge-network/regions#routing. So hosting a database in AWS in any of those regions probably makes the most sense.
I personally have Caprover running on a $5/month DigitalOcean droplet where I can create DBs on the fly for small projects. I normally switch to a better hosted platform for bigger projects. Heroku is also good but the free plan is 10,000 rows and the next plan is $9/m.
Front end deployment, when done correctly, is already easy. S3 + Cloudfront and done: all your static assets are globally available everywhere with good response times in all major geographic areas.
For being "easy" it actually is quite hard to find documentation on how to run an S3 website with CloudFront "properly". Amazon makes it very confusing to figure out that, e.g., web server mode is different than regular S3 mode. CF itself tries to get you to do the wrong thing. Plus you have to jump through hoops to make your bucket private but still give CF access to it. I was doing that wrong for years. How many multimillion dollar companies wouldn't be relevant if Amazon just fixed their freaking docs?
It’s several things that package well together, plus some payable value add for certain use cases.
- they develop Next.js which is one of the two primary React-based solutions for static/server/client rendered websites (the other being Gatsby)
- they provide low friction deployment solutions for same (like heroku)
- they provide tools for specific markets attracted to one or both of the above, e.g. ecommerce
- probably a fair bit more, but those are their standout offerings
Worth noting that Next.js is a mix of custom offerings (they have their own solutions for things like static site generation and image optimization) and configurations for existing tools (a lot of what works well is a predefined set of Babel and Webpack configs, but they can be overridden unlike create-react-app, and there’s broad support for that).
It's a PaaS service like Heroku and others. However it's the new generation of "serverless" billing models where you don't worry about instances/nodes/servers at all and pay in much finer-grained increments.
Basically node-js servers running your code with an integrated CDN, but they also support Go, Python and Ruby for running server-side logic.
I'm a huge fan of vercel, it's saved me a lot of money[1] this year. Their developer experience is incredibly friendly and they've contributed extremely useful libraries to open-source, including micro and next.js. I've had great support experiences for some serious issues like platform bugs. They're in a tough position because the serverless paradigm is still foreign to many people, and many typical project dependencies won't operate in a serverless environment.
I have a set of critiques just like any infrastructure user, but by and large Vercel is moving in the right direction and I'm looking forward to seeing them succeed!
[1] My bill running CollegeAI.com on Kubernetes went from $600/mo to $20/mo with Vercel. It required a large refactor to serverless and the initial week of deployment the performance was so much better I got a massive traffic spike that increased my traffic from tens to thousands of daily signups. The bill that month was over $1000 because I didn't have the serverless parameters set up totally correctly.
Based on his HN linked Twitter and his LinkedIn from googling his name, that does not appear to be the case at all. Not sure why someone would lob an accusation like that.
It's from a bygone era when Vercel would charge for serverless invocations. I think it still might be possible to run into the problem in overage so I can describe it a bit.
Basically I was polling an API endpoint on our SPA client. Maybe one hit every second. We were getting over 10,000 (non-signup) users a day, so it ended up being a lot of requests. Our functions were configured with the default amount of RAM and CPU allocation. When I saw the bill shoot up, I reduced the rate of polling and adjusted the serverless functions to have a lower amount of RAM. This got the price down to $50/mo. Later, Vercel would make pricing per developer seat so the price dropped even more.
The traffic dropped to more reasonable levels after that. But it was a very good week and the traffic remained nominally 2x-5x. I'm sure google's algorithm played some part in the traffic spike.
(also, I am not an employee of Vercel, just a customer :)
I explained it a bit above :) generally Vercel works really well for high-traffic sites without worrying about auto-scaling. But you'll run into unique issues whenever you're dealing with two many digits haha
I think my problems were always more pronounced with Kubernetes fwiw :)
From first hand experience - Vercel is pretty magical. I’ve helped my wife develop an e-commerce site for her planetary-themed kids merch [1], we chose NextJS with a Shopify Storefront backend and deployed the entire thing to Vercel. Didn’t even need SSL certs, no fiddling with CDNs, it’s just magic.
We are super happy with how it turned out vs dealing with full Shopify front end/backend stores.
We developed from scratch, I frankly wasn’t aware of the template when we started but also wanted a very specific look and feel (and the whole planet section on the bottom of the welcome page obviously had lots of work).
Not aware of any issues. We did quite a tight integration with Shopify such as the entire website shop is 99% generated from the Shopify items (which in turn are generated from the Print On Demand catalogues. It’s turtles all the way down).
I have a site (https://familytreechart.com) with a similar integration with Shopify but one thing I've never figured out is how to send someone to Shopify's checkout page with a preconfigured cart. It looks like you are managing the cart on the nextjs side. How are you generating the checkout page with the right items? I've resorted to creating a DraftOrder, but those don't allow for discount codes so I also had to handle that on my end when creating the draft order. Email in profile if easier.
There are cart permalinks [1] that you could use but depending on how much you want to pre-populate I recommend using the JS Buy SDK. I found it the other day and set up a custom free gift with discount code for a Shopify Plus client. The discount code uses the BOGO but the custom functionality checks the conditions and adds the product(s) to the checkout if it is not already there then applies the discount (the client wanted to use the discount code to add the free gift instead of a permalink).
You can add to cart, then add the discount, then redirect to checkout all within one link (working on an app to help merchants generate them).. I believe that is the proper flow but I cannot remember off the top of my head I just know that is all currently possible with a single link.
That said, I have been made aware that the checkout experience will be changing next year. I'm not sure what that entails yet just that it is coming.
Also, having worked with Shopify rather heavily since August, I love it.
The website looks excellent but the perf could really do with some work.
After a quick look in the profiler I found that you're eager loading full-res images from shopify, and it's adding quite a bit of time before the page fully paints, with a lot of reflow. I'd recommend lazyloading them after the fact, with a super low res version as a background-image and adding a min-height to the images so that the whole page doesn't reflow as much.
Otherwise a great site and good looking products, If I had kids I'd buy a shirt.
Many congrats Vercel! I've been a big fan since Zeit platform version 1 - I remember the transition to platform version 2 was difficult but over time I can see how incredible of a platform you've become.
Just a note from a fan - I hope you all can determine a way to approach the future with longevity and corporate independence. The web really needs your entire teams dedication towards performance, end user experience, and developer experience in the coming years. I would love to see Vercel as an independent entity a decade from now and witness how your philosophy changes the web in that time!
Don't think that's possible after raising such amounts of money. Their way out is now either getting aquired for loads of cash or going public. Going public also means loosing parts or all of your control over the company. Those investors will want to see their money back some time in the future.
How long will that persist? Any such company has no chance of being included in e.g. S&P 500 so there's clearly strong pressure in the opposite direction...
If trillions of tracking dollars with a measurable boost on valuation and trading [1] and thus a measurable reduction in the cost of capital [2] is “nothing else,” then yes.
It wasn't a "transition". They straight up killed features that would never work on 2, that worked on 1. For example, websockets. Turns out you can't do those in their ephemeral lambda model.
It is always the same, you want database use a third party, you want websockets use a third party, this approach turns out to be very expensive if you are starting
> Airbnb’s engineering team, for example, isn’t a heavy user of Next.js yet, but they are a world-class team of designers, engineers, product managers, and marketers.
This is a semantically null sentence for the purposes of your article. Skip it and you'd have the same meaning without sounding like you're sucking up to a company.
> Hashicorp pushed back on the dogmatism of pure static. Their team is laser-focused on the user experience. They don’t care if an application is CSR, SSR, SSG, etc. as long as their end-user is delighted.
This has no meaning to me, personally. If Hashicorp doesn't care about pure static as long as their "end-user is delighted," why did they push back? Was there some kind of drama you're revealing? How is "end-user delight" related to any of the acronyms? Are you suggesting pure static pages "delight" fewer users?
The writing is pretty bad. Funnily enough, on the webpage they link from the "CSR, SSR, SSG" string doesn't contain the words "CSR" or "SSG", so it doesn't explain much unless you have the time to read the whole page.
it makes sense to people like me who have been following the vercel story closely, but youre right in that this press release probably should have been written with a much more contextless audience in mind.
Shout out to Vercel (formerly Zeit) for their awesome Hyper terminal https://hyper.is/
I love Vercel - it runs several of my Gatsby sites and an Angular website for free. It also builds out my PRs so I can see how the website is changed. If ever I need to use their paid tier, I'll do it.
No offense, I think people are crazy to rely on a stack they can easily make themselves vs. a series B startup. It's not that difficult to build and deploy full stack apps on your own server.
I Don't care. There are tons of examples of building shit on stacks that won't be around. How about investing a shit ton of time into Azure Kubernetes using Docker containers to find out they dropped support, just like this startup can drop out and never be heard of again. Good luck with that.
that seems like the nature of tech stacks, going all the way back to Flash and Dreamweaver... at least with Vercel it's pretty agnostic — I can easily switch a node app from Vercel to Digital Ocean or whatever. Never used Docker before ha, tried to get into it but the install tutorial flow was completely broken :/
This is a red herring. They don't have unique technologies. Just a very, very well done and accessible combination of best practices.
I have a bunch of apps on Vercel right now and zero worry that I will be able to deploy them somewhere else should they drop out for whatever reason. They are just Node/React apps.
I'm a designer w/ some developer experience that uses Vercel for our startup... I'd have no clue how to deploy my own server. And Digital Ocean costs money... Vercel is free for me right now
If this view had any basis in reality, and if you really believed it, you would be crazy not to build one yourself (easily) and make boat-loads of cash from all the crazy people paying you for it.
"not that difficult" is plainly untrue depending on skill level, even in the absolute best case it's miles away from the ease that is offered here.
I pushed a new proof of concept to Github yesterday and deployed it in literally 3 clicks on Vercel.
Please show how to do this with AWS.
Then take a look at the features like preview deployments, CDN, environment variables etc. Does your solution cover all of them with the same ease?
I think people are crazy to accept abstraction on the machine layer but keep reinventing the wheel for the infra in the middle. These services let's you focus on the actual product.
For the last two projects I really wanted to use Vercel but there is something about its interface that makes me unable to understand it. I ended up both times on Netlify instead.
Is it me not being a dev by trade or it happens also to others?
Glad to see this because we just started using Vercel for StatusGator.
Their API is great: we were able to automate deployments of branded status dashboards very quickly. There are some rough edges around domain name management but overall the experience is very developer-friendly, analogous to Heroku but for a different architecture.
> but overall the experience is very developer-friendly, analogous to Heroku but for a different architecture.
Just to elaborate on that, Vercel is awesome for stateless apps - specially front-end apps. Heroku excels on (small scale) statefull apps - like the back-end app connected to a database (add-on).
In my previous company we used both Vercel and Heroku, which was a great match. Both of them have Review Apps, for example.
It's ironic that the blog of next.js creators is not opening when cookies/localStorage are disabled. For two years I have whitelisted sites in the main browser and have a separate browser where everything is enabled but cleaned on close, so that I could read broken websites. I hope "the next web" they are building just works in a single browser without cookies/tracking at least for blogs/news.
Had to use Zeit/Vercel a few months for a “serious” project that requires SSR, caching and multiple staging environments. As the project progresses I noticed an increasing pain from constantly fighting with outdated documentations and limited configurability. I hope that’s changed now though, because there were some parts of Zeit that we loved. For simpler static sites, it’s hard to beat Netlify.
the docs could definitely use some work. there's a ton of nuance in advanced nextjs apps and the docs usually have the answer i am looking for, but it's tough to find it
the timing of the A and B are certainly unusual. curious what changed and whether it was "a formula starting to work" or "need cash" or "pre emptive B". i guess it's not really any of my business but it'd be nice to know.
Mesosphere was started by Berkeley PhD interned at Google Borg. They worked on some exploratory project to expand the capabilities of Borg platform and wrote some paper.
Unfortunately, that project was on a slightly skewed direction of the technical trend. The philosophy and direction embodied in that project was neither useful inside Google, nor it matches the industry trend. But nonetheless an interesting improvement.
With the reputation of the group and relationship they decided to start a company.
And once docker is out, and kubernetrd next, in 2014 many have seen that mesos the technology is bound to irrelevance.
That prediction verified itself rather quickly, when mesosphere rebranded itself d2iq.
Oh god. Even companies raising $40,000,000 can’t explain what they do:
“Vercel combines the best developer experience with an obsessive focus on end-user performance. Our platform enables frontend teams to do their best work.”
Nice, inspiring evocative words, but I care to find what the fuck Vercel IS, and this is not an answer.
Maybe if I reword the above it will become clear:
Vercel lets developers have a really great time concentrating super hard looking super close at how fast things are. The thing we do lets you do the best thing you do.
> Vercel is a cloud platform for static sites and Serverless Functions that fits perfectly with your workflow. It enables developers to host Jamstack websites and web services that deploy instantly, scale automatically, and requires no supervision, all with no configuration.
Do you think it's reasonable for someone arriving at the website with the very common question of "what exactly do they do?" to click the "Documentation" link?
It's square aimed at companies who hire developers. They want company money, not developer money. Sure, some developers pay for Vercel's products, but they want big chunks of money that managers approve to expense.
This assumption that any developer might automatically navigate to documentation page sounds silly. This is not true. Developers glance the page like normal people do and the OP's concerns are not wrong.
I would! I also like to look at the job openings page when evaluating a technical product. The experience they ask for can give a lot of insight into internal engineering health.
OP posted the non-tech audience answer and is hot under the collar about it. Do you need them to spell out every part of what they do? Part of investing is due diligence
YOU aren't the audience for this press release because the people they are trying to attract are more interested in abstract concepts than concrete ones. E.g. "I'm a great visionary CTO of a 100 person dev team and I use Vercel". A developer doesn't care that Vercel raised $40M, all they want to know is "if I push code to your stack, will it host my JAMstack".
NOW, I would argue that a good press release for a devchain based company can talk to BOTH audiences, and I think the criticism there is fair.
The quote "Vercel combines the best developer experience with an obsessive focus on end-user performance.
Our platform enables frontend teams to do their best work." isn't actually from the linked press release, it's from the home page https://vercel.com/
TBH I think the linked blog is targeted at a dev audience, and if it isn't then "They don’t care if an application is CSR, SSR, SSG, etc. as long as their end-user is delighted." (with those acryonyms linking to a Next docs page) is probably not supposed to be there.
I like this idea, thanks for the link. Just noting that it's funny for an article about effective communication to have a prominent typo in its key graphic ('concreet')
If you scroll a bit down on their starting page there is
> Vercel is the best place to deploy any frontend app. Start by deploying with zero configuration to our global edge network. Scale dynamically to millions of pages without breaking a sweat.
They want enterprise customers, not single developers. So they use the usual bullshit marketing wordings.
Their only goal is to get someone high level in a big company feeling the sentence is cool so that he lets his email in the "get demo" page.
This is just normal enterprise sales, and this is probably why they managed to raise so much compared to most similar services. They really have a strong business focus and a sales team
That’s a jaded view. The truth is that most enterprise customers are not making a decision based on a front page blurb and are going to Vercel because they have a problem they already need solved.
The problem here is “I want to avoid increasing the number of backend and SRE engineers for the customer-facing portions of the business and I have a team bought into next.js already or are platform-agnostic.”
People are not Stumbling upon vercel. The developers are coming through next.js. That said, I think this is evidence they could have better copy on the front page.
For an Enterprise customer there is not a lot information that helps to make a proper decision whether it's worth considering, regarding regulation, risk and compliance. I wish you didn't had to go through the whole sales game to find these kind of information out.
To be fair, I eventually got there (I think) but I don’t use Next.js and the banner of “building the next web” threw me off. It’s essentially web hosting, maybe on the PAAS end of the spectrum like Heroku. Seems like they turned up the collaboration dial, which gives them an entry point to team sales. That’s the best I could tell.
Every single thread on HN is filled with the same cynicism. It's getting boring and I consider myself to be a cynic.
What are you even complaining about in this comment? You are complaining that you can't learn what a company does in a single sentence? Or are you claiming that vercel is a bullshit company (because it isn't)
Also, if you scroll down just a little bit in that page you learn exactly what vercel does.
I like vercel and use it for all my projects, but in light of the goal of helping everyone build the next web, I think it would make a lot of sense to enable drag and drop deploy, like netlify. I recently tried to show someone who used windows and had neither node nor git installed how to deploy a static website with vercel, but it was such a pain that we ended up using netlify.
A very happy customer of Vercel here, and a Next.js user.
Vercel makes it really easy to deploy isomorphic JavaScript applications, to the point where I barely have to think about the runtime context. Thinking back to the ~2015 era, it feels like magic in comparison.
I'm trying to decipher what this is. So far I've learned there's a thing called "Jamstack" that seems to be about static websites. How does one manage databases with this technology?
I think the idea is that instead of generating the original http content based on your DB, you have static webpages which have JavaScript call out to your dB-backed API to modify the content in an SPA. This can then be made more efficient with automatic server side rendering.
I run a super small site, so all the data is compiled at build time into JSON files and served using serverless/lambda (e.g. short server-side functions). The data I'm getting from Notion and Airtable, so I don't have to run a "real" DB.
For commenting I'm experimenting with services like Fauna and Supabase
You either generate pages at build time from the data or you build a static page that will fetch the data from the client.
This means you have zero server side rendering, you're just serving up static files, which of course is faster and can utilize CDNs and such.
SSR is also still possible, that would basically give you a classic model of fetching data on request and then rendering the page. My impression is that this is meant more as a transition than a preferred option.
The SSR features are also optimised to the point that the data fetching and rendering is decoupled. So a client side (JS) navigation only needs to fetch the new page's data. The DX around this is super smooth (just expose an async function...)
207 comments
[ 1.9 ms ] story [ 169 ms ] threadNext, from our limited experimentation, has much better ergonomics and it looks like the company has a stronger grasp on go to market.
Crazy how fast that happened
My team has just started a Next.js app. I chose Next.js over Gatsby because on the face of it, Gatsby looked like a lot more configuration. I fully expect we'll "disagree" with Next.js at some point but for now I am really liking that it is opinionated, as it it letting us build really quickly.
I love Hugo and I wish there was a middle ground that was more approachable. I might have to try next.
For Gatsby there is gatsby-image, which works super well. For Next.js there is next-optimized-images which is not very mature. The latest version of Next.js also provides next/image, but as far as I have understood, it requires you to run the Next.js backend server, so it is not an option for 100% static websites.
https://nextjs.org/docs/basic-features/image-optimization
> Instead of optimizing images at build time, Next.js optimizes images on-demand, as users request them. Unlike static site generators and static-only solutions, your build times aren't increased, whether shipping 10 images or 10 million images.
Sometimes I also bumped into issues with Next.js where the framework was trying to send some code to run on the client browser instead of running it during page generation. The division between server and client seemed a bit too hazy for my taste and felt magical in the wrong ways.
I found the plugin system confusing, and poorly documented so it was hard to know how to get the most out of a plugin and adjust it to whatever our own funny use case was.
The graphql thing was annoying, we don't use graphql and don't know it and didn't find any advantage in having it around.
It's been a while since I touched it, but every time I need to a put it off because it's not a nice env to work in compared to our "from scratch" react apps
No need to be aggressive. I don't have anything against Next, I just felt much more comfortable with Gatsby.
I was using Next.js to build a staticly-generated website (so within the remit of Gatsby) but eventually i outgrew the limitations of static site generation (23k pages and counting...) it was so easy to switch to a traditional server-generated model with Next.js. I'm incredibly greatful for picking up and sticking with Next.js for this.
https://nextjs.org/blog/next-9-5#stable-incremental-static-r...
(edit: whoops this was for the parent comment)
Was it your intention to completely talk over everyone's head by using arcane or invented abstractions? Trying to make investors feel they are onto something really advanced?
Or just caught up in VC/SV speak? Because products that nobody understands have a harder time taking root in the commons.
Edit: Literally there is a question above (top at this time) 'What is Vercel?' - this is my point, a communications failure. If people Googling 'Vercel' and not getting the answer very quickly, then it's a communications problem.
Speak in plain English, describe 'what it is' and the direct material value.
> Vercel is a cloud platform for static sites and Serverless Functions that fits perfectly with your workflow. It enables developers to host Jamstack websites and web services that deploy instantly, scale automatically, and requires no supervision, all with no configuration.
I don't mean this in a negative way, and concede that Marketing often fails to even get their users and buyers to understand.
Vercel is failing on communications a little bit.
They don't have some highly specialized tech that 'nobody should understand' - they're literally making tech for the front-end masses, basically 'tech commoners'.
There are literally a handful of people on this page saying 'what does this do?' - how often do you see that on an HN page?
Anyone in tech, should be able to plainly understand what every other tech companies does, within a few sentences.
Some of it might need to be abstracted, and some of it may not fit very well into a sentence - but at least the Wikipedia version should be clear.
Explaining Next - the frame work in slightly more mechanical terms, and then, separating the notion of 'hosted service' would be great. (If that's what it is, because frankly I'm not even sure).
If you go to vercel.com - nowhere does it really say 'what it is'.
The introduction is 'Vercel combines the best developer experience with an obsessive focus on end-user performance. Our platform enables frontend teams to do their best work.'
That is 'bad communication' because it's completely not informative.
Sometimes marketing speak lingers on 'why it's great' - fair enough, but that statement above isn't even that, it's secondary jargon.
'We enable teams to do their best work' <- no place in communications unless you really need to fill a panel with something.
It needs some work.
Next is made by Vercel.
They also have an option to staticly rendered certain pages.
They netlify but for next. Also next usually requires a node server so technically can also act as an api.
Can generate a server less bundle.
It also provides serverless API endpoints capabilities with other languages (Python, Node, Go, Ruby). Most of this comes with a great free tier (including bandwidth which is unlimited).
Team accounts allow your team to collaborate, deploy, run analytics etc seamlessly.
It's great, they've done a great job. I might be forgetting a couple of things, but this is a big part of why they are so successful.
In terms of actual hosting (removing the build/deployment pipeline), Vercel is a subset of AWS Lambda.
Using their own Next.js as an example: everything in /pages/ is client app, everything in /api/ is serverless app.
Client talks to the /api/ endpoints, /api/ endpoints talk to your domain services.
Vercel also provide edge-location hosting and caching.
"Serveress" = you aren't managing a logical server
I.e. you don't need to concern yourself with provisioning, nor maintaining, a server as that part of the process has been abstracted away from you.
They have a suite of complementary software products and services, including NextJS (a "batteries included" React framework with robust support for SSR and SSG), Now (hosting / CDN / runtime in "serverless" paradigm), Micro, etc.
World-class DX and remarkably powerful tools that somehow combine simplicity and flexibility.
No affiliation, just a longtime grateful fan.
FWIW I liked "Zeit" better than "Vercel" (shrug). But preferences aside, I was annoyed that the new redirects for "now.sh" URLs were -- uncharacteristically for a team famous for attention to detail -- mishandled, e.g. breaking a fantastic preso I'd bookmarked (which grauch used in his React Conf 2017 talks) and which I had cause to reference today in promoting NextJS with a client. It's the only real "footgun" move I've seen them make. Here's hoping it's an anomaly!
It vaguely seemed like a replacement for react but it’s hard to believe you could raise $40M for that so something else must be going on. I generally was under the impression that the industry was moving away from client side rendering since it’s so slow. At least I browse the web with JavaScript off by default.
I’ll be releasing my solution as a config wrapper (or offer as a PR for their Preact example) before I launch my site so other folks don’t need to go through the same process.
For my simple sites it feels like hosting is just solved. Every time I rev my site I just push to git and I’m done.
If you want to deploy a JS frontend (not just static files, though you can), this automatically handles a lot of things like SSR, SSL, etc https://vercel.com/pricing
You can also use their serverless functions; which add "backend" functionality to your existing frontends. You can deploy these serverless functions in a bunch of languages (see docs) but personally I've used them in Nextjs's baked-in functionality of api routes to basically add an endpoint for mutating/querying to/from my DB for example.
Vercel(Frontend -> Serverless functions) -> DB
- they develop Next.js which is one of the two primary React-based solutions for static/server/client rendered websites (the other being Gatsby)
- they provide low friction deployment solutions for same (like heroku)
- they provide tools for specific markets attracted to one or both of the above, e.g. ecommerce
- probably a fair bit more, but those are their standout offerings
Worth noting that Next.js is a mix of custom offerings (they have their own solutions for things like static site generation and image optimization) and configurations for existing tools (a lot of what works well is a predefined set of Babel and Webpack configs, but they can be overridden unlike create-react-app, and there’s broad support for that).
Basically node-js servers running your code with an integrated CDN, but they also support Go, Python and Ruby for running server-side logic.
I have a set of critiques just like any infrastructure user, but by and large Vercel is moving in the right direction and I'm looking forward to seeing them succeed!
[1] My bill running CollegeAI.com on Kubernetes went from $600/mo to $20/mo with Vercel. It required a large refactor to serverless and the initial week of deployment the performance was so much better I got a massive traffic spike that increased my traffic from tens to thousands of daily signups. The bill that month was over $1000 because I didn't have the serverless parameters set up totally correctly.
Basically I was polling an API endpoint on our SPA client. Maybe one hit every second. We were getting over 10,000 (non-signup) users a day, so it ended up being a lot of requests. Our functions were configured with the default amount of RAM and CPU allocation. When I saw the bill shoot up, I reduced the rate of polling and adjusted the serverless functions to have a lower amount of RAM. This got the price down to $50/mo. Later, Vercel would make pricing per developer seat so the price dropped even more.
The traffic dropped to more reasonable levels after that. But it was a very good week and the traffic remained nominally 2x-5x. I'm sure google's algorithm played some part in the traffic spike.
(also, I am not an employee of Vercel, just a customer :)
I think my problems were always more pronounced with Kubernetes fwiw :)
We are super happy with how it turned out vs dealing with full Shopify front end/backend stores.
[1] https://littleastronauts.com
Any gotchas when using Shopify as just the backend? Did you use the Nextjs e-commerce starter template?
We developed from scratch, I frankly wasn’t aware of the template when we started but also wanted a very specific look and feel (and the whole planet section on the bottom of the welcome page obviously had lots of work).
Not aware of any issues. We did quite a tight integration with Shopify such as the entire website shop is 99% generated from the Shopify items (which in turn are generated from the Print On Demand catalogues. It’s turtles all the way down).
You can add to cart, then add the discount, then redirect to checkout all within one link (working on an app to help merchants generate them).. I believe that is the proper flow but I cannot remember off the top of my head I just know that is all currently possible with a single link.
That said, I have been made aware that the checkout experience will be changing next year. I'm not sure what that entails yet just that it is coming.
Also, having worked with Shopify rather heavily since August, I love it.
[1] https://community.shopify.com/c/Shopify-Design/Cart-Use-perm...
After a quick look in the profiler I found that you're eager loading full-res images from shopify, and it's adding quite a bit of time before the page fully paints, with a lot of reflow. I'd recommend lazyloading them after the fact, with a super low res version as a background-image and adding a min-height to the images so that the whole page doesn't reflow as much.
Otherwise a great site and good looking products, If I had kids I'd buy a shirt.
Just a note from a fan - I hope you all can determine a way to approach the future with longevity and corporate independence. The web really needs your entire teams dedication towards performance, end user experience, and developer experience in the coming years. I would love to see Vercel as an independent entity a decade from now and witness how your philosophy changes the web in that time!
Don't think that's possible after raising such amounts of money. Their way out is now either getting aquired for loads of cash or going public. Going public also means loosing parts or all of your control over the company. Those investors will want to see their money back some time in the future.
not a very intellectually honest statement to make considering founder supervoting shares are extremely common in tech.
I don’t actually know, but I was struck by the hyperbole in a comment essentially complaining about hyperbole.
If trillions of tracking dollars with a measurable boost on valuation and trading [1] and thus a measurable reduction in the cost of capital [2] is “nothing else,” then yes.
[1] https://www.sciencedirect.com/science/article/pii/S138641810...
[2] https://opus.lib.uts.edu.au/handle/10453/130246
only if all shares have equal voting rights.
This is a semantically null sentence for the purposes of your article. Skip it and you'd have the same meaning without sounding like you're sucking up to a company.
> Hashicorp pushed back on the dogmatism of pure static. Their team is laser-focused on the user experience. They don’t care if an application is CSR, SSR, SSG, etc. as long as their end-user is delighted.
This has no meaning to me, personally. If Hashicorp doesn't care about pure static as long as their "end-user is delighted," why did they push back? Was there some kind of drama you're revealing? How is "end-user delight" related to any of the acronyms? Are you suggesting pure static pages "delight" fewer users?
The writing is pretty bad. Funnily enough, on the webpage they link from the "CSR, SSR, SSG" string doesn't contain the words "CSR" or "SSG", so it doesn't explain much unless you have the time to read the whole page.
re: hashicorp - jeff escalante their web team lead gave a talk on their move recently: https://www.youtube.com/watch?v=QlNP3PSLT2Q
I love Vercel - it runs several of my Gatsby sites and an Angular website for free. It also builds out my PRs so I can see how the website is changed. If ever I need to use their paid tier, I'll do it.
The very nature of an innovative idea is that few people see it.
I have a bunch of apps on Vercel right now and zero worry that I will be able to deploy them somewhere else should they drop out for whatever reason. They are just Node/React apps.
In the meantime you get faster, free hosting, better builds and spend zero time on ops.
I pushed a new proof of concept to Github yesterday and deployed it in literally 3 clicks on Vercel.
Please show how to do this with AWS.
Then take a look at the features like preview deployments, CDN, environment variables etc. Does your solution cover all of them with the same ease?
I think people are crazy to accept abstraction on the machine layer but keep reinventing the wheel for the infra in the middle. These services let's you focus on the actual product.
Their API is great: we were able to automate deployments of branded status dashboards very quickly. There are some rough edges around domain name management but overall the experience is very developer-friendly, analogous to Heroku but for a different architecture.
Just to elaborate on that, Vercel is awesome for stateless apps - specially front-end apps. Heroku excels on (small scale) statefull apps - like the back-end app connected to a database (add-on).
In my previous company we used both Vercel and Heroku, which was a great match. Both of them have Review Apps, for example.
Along the way, showing how OSS businesses can win: Slowly but steadily (looking at you mesosphere).
Congrats to everyone at Vercel! God knows it has taken very many pivots to finally reach here.
[0] https://news.ycombinator.com/item?id=12320489
[1] https://news.ycombinator.com/item?id=18407503
[2] https://news.ycombinator.com/item?id=12788922
[3] https://news.ycombinator.com/item?id=22933479
Mesosphere was started by Berkeley PhD interned at Google Borg. They worked on some exploratory project to expand the capabilities of Borg platform and wrote some paper.
Unfortunately, that project was on a slightly skewed direction of the technical trend. The philosophy and direction embodied in that project was neither useful inside Google, nor it matches the industry trend. But nonetheless an interesting improvement.
With the reputation of the group and relationship they decided to start a company.
And once docker is out, and kubernetrd next, in 2014 many have seen that mesos the technology is bound to irrelevance.
That prediction verified itself rather quickly, when mesosphere rebranded itself d2iq.
“Vercel combines the best developer experience with an obsessive focus on end-user performance. Our platform enables frontend teams to do their best work.”
Nice, inspiring evocative words, but I care to find what the fuck Vercel IS, and this is not an answer.
Maybe if I reword the above it will become clear:
Vercel lets developers have a really great time concentrating super hard looking super close at how fast things are. The thing we do lets you do the best thing you do.
Ah, now it’s crystal clear what Vercel is.
It was pretty easy to just click "Documentation" and see more: https://vercel.com/docs
> Vercel is a cloud platform for static sites and Serverless Functions that fits perfectly with your workflow. It enables developers to host Jamstack websites and web services that deploy instantly, scale automatically, and requires no supervision, all with no configuration.
Seems fairly clear what it does.
This one is better than most zero information PR lines. For what it is worth the statement can't be written by GPT-3.
https://www.toolshero.com/communication-skills/ladder-of-abs...
YOU aren't the audience for this press release because the people they are trying to attract are more interested in abstract concepts than concrete ones. E.g. "I'm a great visionary CTO of a 100 person dev team and I use Vercel". A developer doesn't care that Vercel raised $40M, all they want to know is "if I push code to your stack, will it host my JAMstack".
NOW, I would argue that a good press release for a devchain based company can talk to BOTH audiences, and I think the criticism there is fair.
TBH I think the linked blog is targeted at a dev audience, and if it isn't then "They don’t care if an application is CSR, SSR, SSG, etc. as long as their end-user is delighted." (with those acryonyms linking to a Next docs page) is probably not supposed to be there.
Highly recommend for the techies here who want to hone their writing skills.
> Vercel is the best place to deploy any frontend app. Start by deploying with zero configuration to our global edge network. Scale dynamically to millions of pages without breaking a sweat.
which seems rather clear to me.
One of the first sentences on the home once you read past the heading.
Their only goal is to get someone high level in a big company feeling the sentence is cool so that he lets his email in the "get demo" page.
This is just normal enterprise sales, and this is probably why they managed to raise so much compared to most similar services. They really have a strong business focus and a sales team
The problem here is “I want to avoid increasing the number of backend and SRE engineers for the customer-facing portions of the business and I have a team bought into next.js already or are platform-agnostic.”
People are not Stumbling upon vercel. The developers are coming through next.js. That said, I think this is evidence they could have better copy on the front page.
https://news.ycombinator.com/newsguidelines.html
What are you even complaining about in this comment? You are complaining that you can't learn what a company does in a single sentence? Or are you claiming that vercel is a bullshit company (because it isn't)
Also, if you scroll down just a little bit in that page you learn exactly what vercel does.
Microsoft/GitHub could be a good landing spot for them if they are to sell.
Vercel makes it really easy to deploy isomorphic JavaScript applications, to the point where I barely have to think about the runtime context. Thinking back to the ~2015 era, it feels like magic in comparison.
For commenting I'm experimenting with services like Fauna and Supabase
This means you have zero server side rendering, you're just serving up static files, which of course is faster and can utilize CDNs and such.
SSR is also still possible, that would basically give you a classic model of fetching data on request and then rendering the page. My impression is that this is meant more as a transition than a preferred option.
The SSR features are also optimised to the point that the data fetching and rendering is decoupled. So a client side (JS) navigation only needs to fetch the new page's data. The DX around this is super smooth (just expose an async function...)
How am I going to educate my child on the value of /value/, when this is a term with serious meaning?