Replit permanently moves to paid hosting after 7 years of free service
noreplit.com
noreplit.com
So, free users are sad because it’s not the free they want it to be, except the free they want is essentially free everything forever. That doesn’t really work.
Registering a whole unique domain and taking the time to make this really rubs me the wrong way. If you dislike it, move on, maybe tell your friends.
I’ll also say that if something means so much to you that you paid for a domain but didn’t pay for the thing you paid for the domain to complain about, why didn’t you pay for the thing you complain about?
If something means something to you, adds value to your life, or saves you time, maybe it’s worth considering paying for it if the value provided is worth the cost. Demanding free stuff is entitled and silly and needs to stop.
At risk of derailing this conversation, but a common problem today is that oftentimes the companies offering a free service simply won't let you pay for it: long before Elon and Twitter Blue, I wanted to pay for Twitter so I could have an ad-free experience with some kind of "personal-grade" API access so I could use my own clients - but that was never an option.
Similarly, I would like to pay Google for an ad-free search experience (and to filter out content-farm websites from my search results...), instead Google's attitude is pushing me towards https://kagi.com/ - which is their loss, I suppose.
There's also the worst-of-both-worlds: when you pay for something, and it still comes with ads, and introduces user-hostile changes to the UX - or otherwise leaves you with the feeling that you aren't valued as a customer (e.g. Windows 11, Reddit, post-Elon Twitter, etc).
---
We're told that as customers in a free-market it's up to us to vote with our wallets, but clearly even that isn't working to effect the change we want to see.
But bnce the code is written to handle it then it's there for everyone to use (though yes, billing infrastructure does requires continual upkeep) - it could even be an internal garage/passion-project/hackathon thing. Companies like Twitter want to attact "the best", and people like that will want to build products that they themselves want to use - why stop them?
Plus there might be other intangible benefits to supporting power-users: it means the power-users will talk about how great something is, and you can't buy word-of-mouth influence like that.
> Me, you, and most of engineers are in extreme minority for our willingness to pay for stuff to avoid ads.
People said the same thing about paying-for-YouTube, and yet YouTube Premium is here to stay, and I'm glad that Google was willing to experiment with the concept, and it worked - but what about all the companies that aren't willing to at-least try to find out?
A passion project is great until the passionate project flounder moves on, then it’s unmaintainable junk.
It’s kinda like how Microsoft handled easter-eggs in the 1990s (before they were banned completely around ~2003), where the egg is just another engineering feature-project, with a spec-doc, a dev timeline, a test plan, and accountability - even if the end-feature is designed to be hidden.
And the Peacock streaming service.
I did use GSuite for my own email/cloud storage setup though have since moved to other providers to try to mitigate risk, eg if Google doesn’t like me for literally any reason, they could disable my account and I’d lose access to mail, files, etc.
I’ll often pay for products I use a lot even though I don’t get that much value out, eg Discord/Reddit, though those two companies seem to recently have changed course somewhat so I’ve stopped paying for them. It was mostly to support a product I cared about.
A lot of people aren’t fortunate enough to be able to pay for stuff, which I was once not, and realize that it’s not always so simple of a problem.
In this case, though, if you have access to the Internet and a semi-okay device, you can achieve a lot
Like, it’s understandable that you get a little burned the first time it happens, but then you’ve learned that this is how it all works. “Free” in a SAAS context has always and will always mean “no need to pay us until we decide otherwise, end of negotiation.”
Every time I look at a free account for some product now, I ask myself if I’m willing to pay for it at some point. If the answer is “no”, then sometimes I just don’t even do it.
People can’t even be bothered to think critically about the product situations they put themselves, and I’m sure these people are intelligent in many other aspects of their lives, but this is such a simple concept that I don’t understand how people, especially tech professionals, struggle with.
I think with software it’s even harder for some people to appreciate the costs because they don’t have an intuitive understanding of what’s required to make it - even though most people don’t know how a lot of their goods and services are made, they know eg a guy in a factory made it and some other guys transported it, and they can kind of put a face to a name. They might know that someone had to design the page they’re looking at, but they have no idea a backend even exists (why do you think we have the term “cloud”?) and no way to estimate what it costs to run, so of course they don’t also know about CVEs and web standards and wipeout/takeout/data residency and multitenancy and releases. They just see a web page that used to work and now doesn’t.
It doesn’t help that so much of the software they use is free, because it’s being used as a funnel/delivery mechanism for paid stuff like hardware or monetized through ads, or paid for by their employer, or because it’s just cheap to maintain. It’s kind of reasonable to assume that all that software stuff you use is free because it’s easy to make. Even though replit is software-for-software, it seems like the type of thing that’s used mostly by beginners who still probably don’t understand what it takes to maintain and operate paid software to a reasonable degree of quality, so the same situation applies.
If that is the case, imo they'd be better served looking into the immense amount of resources available to students: I know when I was a kid I had $0 to spend on software and hardware, just sending an email from a .edu account got people willing to give
Today it's even easier to get those kinds of resources: https://aws.amazon.com/education/awseducate/
There is some liability on the host some free things your put out there for example "free cdn hosted" JavaScript libraries. If your not at a significant scale that has a business model that lets you commit to continuing to host things you should perhaps not set up free hosting for things.
Not saying it applies in Replit case; they are I imagine a company trying to show revenue growth so they can continue to exist and does not sound like they are breaking production with these changes that are announced ahead of time. Users can migrate to another thing as the article is outlining.
In my particular case there were a few pain points where I couldn’t help but sympathize with the complainers. For starters, they’re still subject to all the problems stemming from vendor lock-in, so migrating away may require not just a lift and shift but a partial or total rewrite. Another common problem is that they wrote some software for eg a for-fun website or side project a long time ago and don’t check up on it often - they may not even use or have access to the email and other contact info you have at your disposal to tell them that their stuff will break if they don’t pay by a certain date. So a lot of people don’t know about the change until their stuff breaks or they check in one some old project and find it broken or gone.
This was less of a problem for me than it probably is for Replit, but students are particularly impacted because they usually have little money relative to their investment or interest in the thing. Students under 18 without any credit or debit card may not even have any way to pay at all. And also they are still figuring out how things work so they don’t necessarily understand without being told that eg free tiers are primarily meant make it easy for people to try the software before paying for it (the company doesn’t want you to use it forever without paying), subject to a lot of annoying things like abuse, and just generally require a decent amount of operational maintenance to keep running. And of course they haven’t had corporate communication styles beaten into their brains so they’re pretty rude about expressing their discontent.
Basically, I think it’s wrong to feel entitled to receiving a free thing that obviously costs someone money to provide for forever, and that you should just pay for it if it adds value and is reasonably priced. But it’s also inherently frustrating to be on the receiving end of these changes, so grumbling and discontent are only natural.
That's a failure of the product/marketing planners then - there's a great deal of positive value that comes from offering free (or heavily discounted) plans for verified students, for example. There's a reason why Adobe and Autodesk tolerate, to a certain extent, unlicensed/piracy of their tools, especially in markets where even their educational-discounts are still too expensive for many to afford.
So yes, they dislike it and want to let people know it sucks. Replit is welcome to charge what they want. Former users are welcome to post a public website of what they think of the changes.
I remember them showing up on here. I knew this wasn't going to last.
Cloud Flare will be in a similiar spot in the future.
lots of hosting providers with cheap options these days.
Lightsail instances each include at least 1TB of transfer each month, which is the same as DigitalOcean. Vultr and Linode also charge network overage fees on a per gigabyte basis if you transfer too much. Absolutely, everyone should set up billing alerts on any of these services.
I'm not here to say Lightsail is the absolute best option, just an option.
In which case, I’m not using any of them outside of an LLC.
I suspect the bandwidth is lower, but that is fair enough.
Started using them for almost 15 years ago when running an SMTP server on home-tier internet connections became less feasible.
The impression I get is that most $5 VPSes are basically the same, although I've really only used Linode. They give you some useful free stuff like an external firewall/router and monitoring (with only 24 hours of history, but that's plenty). Their docs and scripts are really good too if, like me, you're only a dilettante sysadmin.
Although my current one only has 512MB RAM and isn’t enough for Rust to compile Scryer Prolog. Prolog, a language first run on the computers of 1971 and standardised around 1998.
Anyways, we both learnt a lot (htop tmux etc) . I'm always jealous that he got to learn everything earlier than me. But if he's not better than me then I consider myself a failure wrt being an older brother.
The only drawback is that this doesn't work if you want to do ai stuff. For those use cases I rent a machine on paper space for a cheap hourly rate.
I'm a little sad I never was. I started with the Linode "hardening linux guide" and so had a firewall and disabled ssh passwords from day 1. I still have fun looking at the failed attempts on 22 and 443. My server gets so many weird requests, and they used to crash the server. A few iterations and that stopped happening.
Oh, another thing that's worth learning: how to acquire and refresh a Lets Encrypt TLS cert via the ACME protocol. Doing this requires interesting confluence of skills and tools - you must carve out a vestigial http route in your server, and also configure certbot and cron. And working out the bugs takes a few iterations. (You could install Caddy, but where's the fun in that?!)
Making it all work, from scratch, made me feel happy in the same way that when I watch people rebuild carburetors or who build bookshelves from scratch makes them feel. It's not new, it's not innovative, but its good. And it's always more interesting than you'd ever suspect.
Learning how linux operates, ops is done, networks are configured, firewalls are hardened, processes are monitored, resources are allocated, disks are backed up, ips are assigned, routing tables work, symlinks work, and code is deployed is never going to be an obsolete skill.
Pick whatever amazing heroku, fly, render, replit, next, k8s, nomad, docker, whatever platform you think has abstracted things away to make those skills obsolete and you'll find an entire cohort of our industry dedicated to using those skills to keep those services online, improving how they operate, and writing the literal software that powers them. Do you think docker or k8s could exist without Ops knowledge or systems programmers?
My view is that there will be a few very smart people at the companies you mention, doing those things. But most jobs in the tech industry will not require those skills. That's the point of specialization. It is efficient.
You claim that linux, disk backup and routing tables will never be obsolete. Really?
Being able to do various forms of maths (e.g. geometry) has been a useful skill for literally millennia.
Being able to drive has been a useful skill for about a century. The number of people who need this skill has probably peaked in developed countries, but is rising globally.
A lot of basic computer and networking technologies have been around since the 70s or even earlier and the pace of change has slowed a lot since then.
It is also useful to have a deeper understanding of underlying and related systems to you own - it makes it easier to understand problems, to communicate with specialists, etc.
You mentioned you are an economist. Do you think it would be a good idea for someone interested in macroeconomics to decide never to learn any microeconomics at all? Specialisation should be additional knowledge built on a broader base.
> You claim that linux, disk backup and routing tables will never be obsolete. Really?
I would interpret never is "not in your working life" which is a reasonable claim.
Also, underlying skills. We will always need OSes (even if Linux because obsolete), we will always need to backup data, and we will always need to route information. If you have learned one OS, or have experience of backups, or setting up networks it will be much easier to update to the new version.
If data exists, backing it up will never go away, and we still use magnetic tape drives for enterprise backup so no, disk backup isn’t going away, not within our lifetimes, and since your concern is the person wasting their time, I’ll assume our lifetimes is a fine timeline to look at.
Linux, and the open source world that exists because of it and GNU, aren’t becoming obsolete. With the steam deck and EVs, and android auto, Linux is deploying more places than ever, and only increasing. Every container you run is Linux, etc. Even windows supports running Linux.
Packet switching, which routing tables are a part of, is literally how the modern day internet is able to exist, IP protocol and all.
Everything you suggested are core to all modern day tech and won’t soon be obsolete, and you’re probably well aware. Perfect troll response with the selected tech you mentioned. Someone else will have to feed you.
Very simple, but what works, works.
If you don’t like it, there are plenty of alternatives I imagine. And if you find the alternatives more lacking or more difficult to use. Well hey, that’s the value add that replit is charging money for.
I think most of the confusion here comes from some users not realizing the difference between the previous hosting solution “Always On” and the new one “Deployments”. Always On was a quick way to keep a dev machine running, but was difficult to consider for production use cases since the dev machine was what you were serving from. Deployments decoupled the two and enabled separate machines (including dynamic scaling) to be used for development and production. So now users have on-demand dev machines with our cloud IDE in addition to persistent deployments. There are a lot of benefits from this approach, but it meant that Always On wasn’t as useful.
If you're tired of shifty providers, why not vote with your wallet and pay people money for services instead of forcing them to take on increasing amounts of investment to subsidize your usage?
It just comes back to bite you when said investments need to make a return and they get forced into being shifty.
_
I get money can be tight when it's just a hobby project.. a) consider taking money for it b) consider you will not find many hobbies where the "tools of the trade" as cheap as paying for compute can be.
This is a great point. Even my cheapest hobby, running, costs more than a $5/mo VPS in shoes alone.
Heroku (I work there) eco is $5 a month + $5 database ($120 a year).
I understand that the use cases don’t perfectly overlap, so I’m curious about your use case.
I would also suspect that for a free product, nonpaying users also consume a disproportionate amount of resources, both in terms of hosting and customer support - based on my experience working on the acquisition of a similar developer product by AWS, Cloud9 IDE (2019). Students and universities and learning, etc., isn't where the money is.
> The final blow came when Replit announced that your repls could only stay online when you are actively within the workspace, now Replit Deployments stand as Replit's only hosting option. This decision seemed to counter Replit's user-friendly ethos, sparking widespread confusion and shock within the community.
Honestly, that sounds a bit entitled to me. Maybe they reversed course, and that sucks, but are people really expecting Replit to provide free hosting? (I didn't even know the product was capable of that). Free tiers for a product like this can be a massive pain in the neck: a lot of cost for little upside. It's probably worthwhile for GitHub for various reasons, but likely not for a product like this, especially for anything involving code hosting.
For Replit to survive as a business they likely need to fast track into a successful B2B model. A lot of what they built, Cloud9 IDE had developed as of 2016: a web-based IDE backed by a container that you could launch on a dime. https://web.archive.org/web/20160403092511/http://c9.io/ - perhaps it was ahead of its time.
I do think that there's a growing market now, with offerings like GitHub Codespaces; and AWS continues to offer it as AWS Cloud9 post-acquistiion, but ... I see it being very difficult for a company to break into that space with a standalone offering, while competing with integrated offerings like primarily GitHub Codespaces (that build on existing GitHub and VS Code ecosystems) or to a lesser extent AWS Cloud9 (that integrates with AWS). There is a market for cloud-hosted IDEs, and probably web-based ones, but they need to super tightly integrate with everything else to provide a better UX than local IDEs.
Fact check: there is still a free tier (0.5GB RAM, 0.5 vCPU, 10GB storage).
Fact check: you get one always-on instance with your "Hacker" plan.
So this post has a totally misleading title.
I don't even understand the use-case that requires these 5 always-on instances, and the complaint site doesn't make it clear.
If this changed, let's get a link to an official post. My downvote will follow this post.
You can read all about our hosting changes here: https://blog.replit.com/hosting-changes.
TL;DR
- deprecating always on in favor of new deployments product
- migrate workspace URL to replit.dev from repl.co
- replit.dev is the new development URL. only available when someone is in the editor.
- deployments are the only way to host things on replit
From the blog:
> We remain committed to providing a powerful free development experience to anyone who wants to code. This post is only about the hosting experience, which we are migrating to our new Deployments product.
Which is only paid, correct?
Then immediately follows with specs and prices:
> Replit.com - $6.40/month
> 2.199GHz/1GB RAM/10GB transfer
vs
> DigitalOcean.com - $6.00/month
> 2.494GHz/1GB RAM/1TB transfer
EDIT: I originally misread 10GB as 10TB - I'll still maintain that it likely doesn't matter. It's a managed VM where deployments are handled for you, for $0.40 more, and if the author is burning through 10GB of bandwidth monthly on hobby VMs I'd be surprised.
I think you mixed up your units. 1TB — 10GB == 990GB. In terms of transfer cost (at least according to those numbers, which I have not verified) DO is orders of magnitude less expensive per GB. Surely that deserves “way cheaper”?
So, assuming your own quote is correct, DO is indeed way cheaper for anything that is more than a toy.
I totally agree and would say that openly criticising such decisions is important, so that companies are a bit more careful with introducing usecase-breaking changes - including changing their business model in a way that makes it not viable anymore for many users. Such things really make a company seem unreliable, but unfortunately it seems to have become common with the "ship early, break often" mentality especially in the start-up world.
That said, I try to avoid free services now. Kagi unlimited at $10/mo replaced Google for me, I pay for Google Workspace (though only out of a desire not to take the time and energy to migrate), Dropbox, etc. If a service is a hobby for a company and not its core product, I try to avoid that, too.
- I've currently moved everything to a cheap DigitalOcean droplet that works wonders for me, and is far cheaper than what Replit could offer
- Comments about me & other noreplit.com contributors being "entitled" is odd to me. Nothing is ever free, but a service we paid for being switched to something that costs tons more than the competition, and suddenly, isn't great. - Replit has stopped community events such as their Reps program, all hackathons (not AI related, at least), and have abandoned their 16k+ member Discord server in exchange for a barely active discourse site.
- "That’s the impression I got looking at the dev behind this site. They clearly have development experience, they’re just mad and trying to make as much noise about it as they can." That was never my intention. Many people are mad about it, but the only thing I've personally done recently is created the website with the help of a few other folks. Forums such as this one aren't created by me.
- I'm not broke. I have about ~$5k in my bank account at the moment, but I'm not willing to let that go to waste. Beforehand, I was fine to pay $7/month for my silly little projects to have "good enough" uptime, as I think should be true for all hobby projects. It doesn't make sense for me to spend tons more for a little more reliability when I can just spend a day or two migrating, that I did.
- The domain was $15 for the year and isn't on auto-renew.
Replit simply changed their business model to not favour youngins that don't have a slightest clue what a vCPU even means, after they seemed they wanted to support that. So we're a bit pissed having to move our shit, but all is well in the end.
We recently added autoscale (scale-to-zero) & static deployments to our plans. It should be pretty easy for you to host simple sites if you're a sub. Hacker starts at $7/mo. Can read more at replit.com/site/deployments.
Alternatively, you can also just add a card and pay for autoscale overages directly.
Then do that, nobody is stopping you.
Kids, a lot of kids.
OpenAI is $20/month. Or are they comparing Ghostwriter to what you can knock together yourself on top of the OpenAI API?
(Disclaimer, I work at Replit on the AI team, but don't represent the company)
Ironic that another feature at Replit is now being advertised as "free"..
This whole HN thread is about how Replit has reversed many of it's free offerings.