As a high school CS teacher, replit was a godsend.
Aside from letting us do an end run around the downtrodden IT department and their dreaded policies on new software, the screen sharing was a huge help for locked down classrooms where we had to remain distanced.
The product really came into its own for collaboration. Google Docs but for code. Let’s look at some Haskell? No problem. Need a quick graph? X11-VNC stuff worked its magic to give us matplotlib in seconds. The turtle graphics for the younger pupils was so fast too. Scratch is great but Python is better, not the least reason being that syntax errors are impossible in Scratch. Debugging and manipulating text are huge parts of learning the craft as they become teenagers.
Amazing technology plumbed together into a fantastic product. We didn’t pay a penny for it mind you — hopefully they have good sales connections to school districts with deeper pockets than ours.
Even if their sole revenue comes from 100$ of teacher training a year, that’s $500M ARR (4M teachers in the US * 20% are teaching something akin to core CS * a global market that is 5x the size of the US.)
They could grow to 300 employees and that’s still pretty good revenue, but I’m waving my hands a lot on these numbers.
(Nit: I wish I could git pull and push to my account of 1000+ repls.)
At the end of the day SaaS exploits organizations that prefer opex to capex. Even in SaaS you have to deal with downtrodden IT departments and their policies.
"20% are teaching something akin to core CS" seems to be a huge over estimate (off by 100x?). Most high schools would just have 1 CS teacher at most, right?
It’s a nod to the future for sure but we had 80 teaching staff, 3 full time CS teachers, 2 more doing design and technology which incorporated some low level C, 1 more from Physics doing STEM extras and 1 in Geography who was 50% time doing lab work for other subjects.
Until CS is a core subject that’s going to be it for staff levels who do coding, but the elephant in the room is the Mathematics team. They are enormous in any school, and if they incorporate discrete math into their core with one semester of coding you now have another 8-10 teachers who would all really benefit from Replit.
So yes, 20% is a stretch, but CS taught as a core subject is coming.
Thanks for the glot.io recommendation. I had the same problem with repl.it.
I could either use nodejs or python on the terminal or keep waiting for repl.it to initialize them on the browser. Glot at first glace seems to be instantaneous.
The account is not the problem for me. It's being forced to create and name a new project for something that I intended to be a throwaway (like a python playground). Now I just open the same project over and over, deleting whatever the file contents were from the last time.
"Inspire creativity and generate value for creators through community."
I find this one to be the most interesting.
The creators of Replit seem well-aware that they're not going to replace the modern IDE and they don't intend to. It seems like they're pursuing a slightly different goal, one adjacent enough that a lot of people commenting seem to miss the point.
I'll take a broad example: When I look at the cryptocurrency community, I find nearly everything about it contemptible, except the community-driven aspects of it. That's where it really shines.
The way people collaborate and work with each other to build a DAO over the weekend is fascinating. It's an interesting phenomena that I don't think has been properly captured in tech, except in bits and pieces.
For example, Terence Tao talks about leveraging the power of community to solve mathematical problems that needed a lot of disparate insights during the Polymath Project. Together a group of random mathematicians on the Internet were able to essentially turn one of the math problems into a competition: "to improve the bounds for small gaps between primes" where the gap initially began at around 70 million and eventually (through different mathematical breakthroughs from many different contributors) they were able to decrease the gap number down to 246.
It's an amazing feat and Terence remarks that only certain problems are amenable to community work, if the incentives are high enough.
So what does this have to do with Replit?
There's a lot of problems out there that are difficult for any one individual to solve. And certain problems, especially in software, that would be best addressed by groups.
I can imagine a place where you can meet with other programmers who share skills you find useful to be a very valuable space.
It may not seem like a massive leap, but I find anything that reduces friction for software engineers tends to have a 10X effect on productivity.
"Replit (YC W18) Is Hiring to Bring the Next Billion Software Creators Online"
Yeah right, the main problem stopping people from writing python scripts is that the one-click installer isn't convenient enough ;) Or is it all those software developers who refuse to use a keyboard and insist on using a smart phone touchscreen?
Their product is a code snippet repl console. And it's not even the best, first, fastest, or cheapest one.
Their self-presentation to me seems so over the top that I feel like they will turn out to be a scam.
I get that. But, some things just deserve an instant blacklist, and, a CEO punching down at an intern is at least pretty close to that in my mind. We're not talking about something like when TripleByte was going to opt everyone with an account in to sharing a bunch of data they hadn't originally agreed to be shared, then reversed course before implementing it after seeing public feedback. We're literally talking about a well funded company going after a college student and former employee. One thing I've learned over time is that if someone will do bad things to someone else, there's a good chance they'd do bad things to you, if given half a reason to. That makes me wary of dealing with repl.it in any way, but, especially as a potential employee.
These kinds of things are sometime useful if you need to quickly hack on a language you don't normally use. If you don't normally use that language then you are not used to the install, the package system, the environment variables and hacks you need to get it to work properly, the troubleshooting decision trees, etc.
I agree if python is your main bread and butter language doing REPL locally (or debugging with VS Code) would be much nicer.
I used to use replit for this, but I much prefer my vscode setup so I’ve switched to using docker (e.g., running `haskell`). But I agree that hacking in a new language is a good use case of repl.it
Vscode runs in browser and we will see a new startup in this space doing better than replit, with less effort by outsourcing ide to Microsoft which does great ides.
I used Replit to build a Python battlesnake for https://play.battlesnake.com/. It was nice and easy to get started with. Unfortunately the latency/performance isn't reliable enough for my battlesnake - my snake moves between about 50 and 150 in the global league just depending on the Replit server performance.
Are you using the default Battlesnake region? Depending on your location and where your replit ends up hosted, switching regions might reduce your latency a bit.
$80M seems like a lot for a developer tool with a team of 40. Why would they need so much? Besides general growth, the only thing I can see would be for cloud computation costs for AI training, but even that seems excessive. Perhaps they're going for a long runway?
The founder of a successful startup I know used to tell his first funding story all the time. Over a game of tennis, he asked for $2mm, the VC agreed before the ball soared back over the net. He always would say, “I should have asked for $4mm!”
I have a feeling that YC companies (and other well connected) get money thrown in their general direction now a days, all they have to do is ask
> I have a feeling that YC companies (and other well connected) get money thrown in their general direction now a days, all they have to do is ask.
If twitter is any indicator, it seems like the Replit co-founders are actively mentored / coached by pg. Amjad, the founder CEO, even called pg "his manager" at one point. This leads me to:
As someone who went through YC and met PG while he was still actively running it, I can tell you that at least in our batch he was 100% spot on about who was going to do well. He had a tendency to spend his free time with the same individuals who ended up doing phenomenally well. It would be easy to be dismissive about this and say something about doubling down on his best investments, etc. But during our batch many of those companies had not yet become the clear cut winners that they are today, and instead only turned into them a year or so down the road. One of the best companies from our batch is Flexport, and PG had been super impressed by Ryan Petersen even before he was officially accepted into YC.
It's pretty clear in hindsight that taking note of who PG was hanging out with was a very investable strategy. In fact, one of my batchmates had some money to put into other companies in our batch, and when he asked me for my opinion of who to invest in, that's exactly what I told him. I just did some quick math and would estimate that his ROI today is somewhere in the 1,000x range.
Do VCs invest to give companies a "long runway"? Would be news to me, though it does seem plausible. But I'd expect a business strategy of: 1) Invest at a loss to become #1 2) ? 3) Profit.
Does anyone have any thoughts on the blog post [0] linked at the end of the announcement? Is this actually how venture capitalists evaluate these sorts of things? That post feels like it approaches "conspiracy theory" levels of leaps [1] about the evolution of the product and its potential impacts from a company whose product offering doesn't really seem to have changed or expanded much in the past few years, and which feels from an outsider's perspective like it's generally being left in the dust by more modern competitors like GH's Codespaces.
[1] e.g., the CEO making vague tweets about "web3" = "Replit will be a vortex for talent, creativity, and economic activity online. It might become the operating system for the 21st Century, the platform on top of which much of the internet is built, remixed and monetized."
> That post feels like it approaches "conspiracy theory" levels of leaps
While I don't agree that there is conspiracy going on, Replit's co-founder, amasad's post on entrepreneurship from 2016 kind of explains why his tweets outlining his outlook for the future of his company may come off as that:
I view entrepreneurship as means of reconciling the infinite and finite. You venture into your imagination, gather knowledge, and dream about a better world. But you have to bring some of that back to earth. You take a step -- no matter how small -- towards your imagined world in the real world.
This is not a one time thing, it's a recursive process. If your imagined possibilities became real then that's your new "finite". The process restarts and that's how we make progress.
...
I see this as the perfect framework for the work we do in technology. There is always a tension between what is and what could be. There are people and organizations who are stuck in the status quo (the finite). On the other hand, there are folks who are stuck in the what could be without any actionability (the infinite).
Taking a step into the infinite could be hard, but the hardest thing of all is bringing something back to the finite. The real art is coming up with the smallest possible task that can be done here and now.
Can't believe I have to add "vortex" and "remix" to my marketing buzzword list.
Jokes aside, that post is at least insightful about their future trajectory? It definitely gives off the vibe that they're going for a platform/creator economy kind of monetization. I'm with you in that I'm skeptical it'll pay off, especially because the direct comparison they use (Roblox) is very different: it's a game that kids play on, not an "operating system".
They do mention concrete numbers.. doubled users to 10m in one year.
The valuation is still high for where the company is now. I personally think churn is going to be a huge issue for them and there is a huge uncertainty if they would be able to counter it, but I think it is unfair of you not to highlight that they do mention concrete numbers and show solid user tractions, which certainly counters your assertions that "nothing changed" in the last few years.
If Repl.it comes out with a way to sync an offline mode to their cloud offerings I will jump for joy, become a paying user, and maybe even apply to work there.
Just a reminder that there is skulpt if you want a completely free Python 3-ish alternative that runs in the browser. I'm not sure it is suitable for a production environment, but it's great for teaching kids how to code.
Replit was recently in the news when the CEO used expensive company lawyers to attack a an individual contributor on an open source project to force the project to close even though they had done nothing wrong just because it was impossible to pay the legal fees.
Yes.. I thought that incident made him look very, uh, disgraceful.
He likes to predend Repl.it is some marvel of technology when it is unironically a weekend project (which the intern he attacked did better). Bruised egos are dangerous.
I bet the investors took this on a positive note, as jerks tend to be successful CEO in Venture-Tech space. The irony is that every employer wants to hire people who share their vision of the company. But when you quit the job, you should also quit that passion.
Apologies don't undo a history of a mal-intent towards employees who helped build the company. Similar to how apologizing doesn't undo a bad first impression.
Yes, Replit runs on phones and tablets. You can code & run any repl, share repls & edit with other people. We recently started rolling out CodeMirror as the backbone of the editor and we'll be making a ton more enhancements there.
They still using Monaco by default (I just signed in and checked). For what it is worth, Monaco used to have zero support for mobile, but recently got much better: https://github.com/microsoft/monaco-editor/issues/246#issuec...
I tested out Code Server (vscode via the web) on an ipad recently, and it worked surprisingly well.
Yes, Jupyter notebook has used CodeMirror for nearly a decade. The official Jupyter notebook has never used Monaco, but the Google version (colab) uses Monaco in order to very nicely leverage LSP support (similar to VS Code). There is preliminary work under way right now to switch from CodeMirror 5 to CodeMirror 6 in the official Jupyter project, but that's going to be significant work (see https://github.com/jupyterlab/jupyterlab/pull/11638).
It doesn't seem like there is much overlap between Replit and Jupyter notebooks right now, as far as I can tell...
Yes it can. One example is the Juno app in the Apple App Store, which provides a very mobile friendly interface to hosted Jupyter notebook services. The same developer also has a similar app that provides offline support. I’ve also put effort into making my implementation of Jupyter in Cocalc somewhat mobile friendly…
In the earlier emails, Amjad Masad (the CEO) offered to hire the intern, was pleasant, etc. And then just a few threads later, when things turned sour, he insulted the former intern, calling him the "most demanding intern we've ever had."
It's deeply telling about a person's character when they pull this kind of move -- kind & understanding one moment, negging & insulting the next, even if it contradicts their earlier kindness.
What really bugged me about this was actually how he handled the backlash - he initially doubled down, saying he was doing the right thing. Only as the whole thing escalated more, he back-paddled.
Made me feel like he only did it to avoid bad publicity and stuff like this will keep happening.
Ever since, the way Replit represents itself (as the "revolutionary" new platform) is quite off-putting for me and I've stopped using/suggesting them.
They're doing the exact same thing when accessibility is concerned.
In my few interactions with them, they either claimed that they had it figured out (which is not true), or they promised that it will get fixed really soon. I've reached out to them about it a couple years ago, and little has been done to address the issue since then.
Their accessibility is weird, all the modals, menus and so on are perfectly accessible, you can see that somebody put a lot of thought into them, but core components of their platform (like the terminal) don't have any accessibility at all, even though flipping one switch in the library they use would get them 90% of the way there.
Considering they're deep in the education space, this is a really big issue. People might fail their classes because of this, and it might even possibly be illegal in some jurisdictions. If there's one sector where you really need to do accessibility right, it's education.
If you don’t believe your company is going to take over the world you probably shouldn’t be a startup founder. At least not one where you’re raising tens of millions of dollars of VC capital.
It's not revolutionary – no offense intended. There have been a ton of companies that built products like they did that are far more sophisticated with comparably simple getting started experiences.
For example, Cloud9 IDE (c9.io) had (and has) a fully functioning (custom) IDE in the browser, with syntax highlighting, autocompletion, debugger w/ breakpoints, and a terminal/REPL: https://web.archive.org/web/20170201175009/https://c9.io/ - fully functional in 2016, and earlier. I was part of the team at AWS that acquired the company that year [1], which is now an AWS product: https://aws.amazon.com/cloud9/ - they made it trivial to click a button and boot a workspace with a sample Python, Ruby, JavaScript app, etc., and then start hacking on it using the IDE or web-based terminal/CLI. (The latter alone is a very difficult thing to build in a web browser! Not to mention syntax highlighting for many languages, autocompletion, etc. – though Language Server Protocol implementations were a help for adding basic auto completion IIRC).
The Cloud9 IDE was sufficiently mature that the Cloud9 team used it to build their own product.
And there are a number of other companies that provide similar products.
Maybe Repl.it hit some sweet spot of user experience that the others haven't, but – I mean no offense – I don't see what's innovative about it compared to the numerous other products I came across while doing market research back then. (I don't recall all the names.) It looks more like an IDE now than it did a year or two ago, but it seems to be behind Cloud9's capabilities as of 2016, not to mention the present version of AWS Cloud9 and CodeSpaces.
I feel I might be "that guy" who's calling Dropbox a dumb idea, but I'm skeptical that a revenue stream will materialize to justify the valuation. Tools like Repl.it, Cloud9 IDE, and CodeSpaces seem to me like features, not products.
Large corporations aren't going to want their proprietary codebase uploaded to a third party cloud service. GitHub (Microsoft) and Amazon have overcome that trust hump; and both have many developer-oriented products where these online coding tools are one feature among a large product suite, where it can be connected to continuous build and integration systems that deploy code to their clouds.
The largest addressable market for a standalone product that software development shops aren't actually going to use for real software development might be teaching students, which is not a large market. Cloud9 IDE tried that play and the market wasn't large enough to justify a startup, which is why I assume the founders & board of Cloud9 approved AWS's acquisition offer.
I haven't used Repl.it's product recently. What's the differentiator? What's the moat? I'm sorry to be a downer in a forum where we're meant to cheer startups (especially an HN startup); I'm only trying to offer rational analysis. Repl.it must be doing something unique right to be growing enough to garner such an investment. (Wasn't Repl.it just like ~3-5 people a year or two ago?)
On the other hand, this could be the VCs and founders making a play hoping one of the tech giants will acquire them. But Microsoft and Amazon already have equivalents and would likely not be interested. Maybe Salesforce or Google? I don't see this growing to become a public company without substantial diversification of the feature into a product suite that's sticky with enterprises. Maybe they've got a roadmap... would be interesting to see the pitch deck for this investment round.
[2] VS Code is an Electron web app compiled to run natively, which is presumably how GitHub got it running in the browser – especially given that as a Microsoft subsidiary they could presumably tap Developer Division's help.
> [Amjad Masad] initially doubled down, saying he was doing the right thing. Only as the whole thing escalated more, he back-paddled.
Folks should be allowed to change their mind. How is that a bad thing? Given more input / data, it only makes sense a rational person changes their opinion / outlook.
> Ever since, the way Replit represents itself (as the "revolutionary" new platform) is quite off-putting for me and I've stopped using/suggesting them.
For better or worse, they're going after the next wave of developers. Not you and I:
"Replit wants to build the future, and to build the future, you don't rebuild the old thing for the old customers. You build something new that starts out looking like a toy, and you build for the underserved, not just for moral reasons, but for strategic ones. It's disruptive innovation meets the compounding power of young users."
...
"For example, it means not targeting experienced developers to start, even though they're the more valuable users today. If they've already written their first line of code elsewhere, they definitionally can't write their first line of code in Replit."
Funny how the mind works. I have a bad memory normally, but I am yeah "heard of this company on HN before and it was something not too awesome. I didn't remember the details but the flag "something bad about X" stuck.
Especially since if you offer generic compute without any form of registration, you're basically inviting people to mine crypto with your AWS bills, which is something they've experienced: https://news.ycombinator.com/item?id=26682237
At https://CoCalc.com we offer generic compute without any form of registration required (see https://cocalc.com/auth/try). And indeed, crypto mining is a periodic headache for us. I think Replit used to have a model that made it much easier for them to offer something without registration, because they supported client side Python, etc., via Emscripten. However, they eventually seem to have moved away from that.
I realize that this is a car-and-mouse game that you might not want to talk about in public, but are there techniques that you’ve found to work? Is just throttling the CPU for anon accounts sufficient to make it not worth the trouble?
Yes, there are many techniques we've tried for around a decade, with each being a filter that helps a certain amount, and nothing being a magic bullet. This doesn't really answer your question, but maybe gives a sense of perspective at least. I've run a range of sites that provide free online compute to people for over 20 years, starting with sites related to number theory that I hosted at Berkeley back in the late 1990s. Cryptocurrency mining wasn't a problem for the first 10 years, since cryptocurrency didn't exist. Before cryptocurrency, the problems were mainly spammers sending out spam emails, people launching attacks or abusively trying to exhaust all resources (disk, cpu, memory, etc.). People using up all resources became a lot easier to deal with the advent of cgroups and Docker (VM's were not a solution since they are too expensive to allocate for each free user).
The next useful filter is to completely disable the network, except for paying customers: this dramatically reduces malicious use, since it's not possible to use CoCalc to launch outgoing DDOS attacks, run port scanners for weaknesses, etc. Disabling the network also discourages crypto miners, since the default crypto mining software typically communicates over the network with a mining pool. Of course, some small percentage of crypto miners will: (1) use a stolen credit card, or (2) manually upload data from a mining pool and manually copy their results out, and similar issues. You then have to deal with each of these problems, and they do cost you time and money. The goal is that they cost you less than the value that comes from having non-paying users, and so far that is the case for us. For (1), using Stripe helps a lot with stolen credit cards, since Stripe is quite good at detecting and preventing use of stolen cards, but they definitely aren't 100%. Regarding (2), there's really no way around having to develop heuristics and scans of your infrastructure to look for malicious and costly behavior, and manually look into it. Indeed, some technique involve silently throttling abusers, so they hopefully don't even realize their attempt to waste your resources is failing.
All that said, providing compute as a free service to mathematicians at least, isn't hopeless! People appreciate what we do, and the jerks who try to wrecks things for the rest of us haven't succeeded.
> All that said, providing compute as a free service to mathematicians at least, isn't hopeless! People appreciate what we do, and the jerks who try to wrecks things for the rest of us haven't succeeded.
Indeed, the ability to quickly open up a Sage session and check my understanding of something was very useful as a math undergrad! Thanks for the detailed response and cheers for what you do.
literally LOL'ed at their blog picture/headliner page. they are borrowing from the festival world where the bigger the headliner (aka investor/check?), the higher and bigger they are displayed.
wonder if there are parallels between VC's and DJ's and their desire to be the biggest and first.
Yes, and if you use it to work on an open source project that Repl.it's founder thinks might compete with his company, then said founder will send you increasingly unhinged emails insulting you and threatening to sue[1].
Aside from letting us do an end run around the downtrodden IT department and their dreaded policies on new software, the screen sharing was a huge help for locked down classrooms where we had to remain distanced.
The product really came into its own for collaboration. Google Docs but for code. Let’s look at some Haskell? No problem. Need a quick graph? X11-VNC stuff worked its magic to give us matplotlib in seconds. The turtle graphics for the younger pupils was so fast too. Scratch is great but Python is better, not the least reason being that syntax errors are impossible in Scratch. Debugging and manipulating text are huge parts of learning the craft as they become teenagers.
Amazing technology plumbed together into a fantastic product. We didn’t pay a penny for it mind you — hopefully they have good sales connections to school districts with deeper pockets than ours.
Even if their sole revenue comes from 100$ of teacher training a year, that’s $500M ARR (4M teachers in the US * 20% are teaching something akin to core CS * a global market that is 5x the size of the US.)
They could grow to 300 employees and that’s still pretty good revenue, but I’m waving my hands a lot on these numbers.
(Nit: I wish I could git pull and push to my account of 1000+ repls.)