>Rather than becoming defensive, Masad and his team owned the problem. In fact, says Masad, within two days, they rolled out an automatic safety system that separates a user’s “practice” database from their “real” one. The way Masad describes it, it’s a little like having two versions of a website’s filing cabinet — the AI agent can experiment freely in a development database, but the production database, which is the real thing that users interact with, is completely walled off.
I gotta wonder who the median techcrunch reader is if the writer/editor felt it necessary to explain the point of having a staging and prod environment, and with such a pointless analogy. We surely cannot understand what a database is unless we're told it's like a filing cabinet, right?
You don't write for your median reader, you write for the vast majority of your readers.
That's a basic concept of writing. Journalism should be accessible, so even if you know what a database is and how to deploy it in different envs, you shouldn't write assuming that. If a large portion of your readers don't know what you're saying, you've failed as a writer. If your readership includes high school students, you write with that as the baseline.
Richard Feynman certainly didn't write as if he assumed the reader knew particle physics. Be like Richard Feynman.
Ironically, especially when you combine it with the em-dash, it really sounds like exactly the type of completely pointless and unilluminating analogy that LLMs love to generate. These analogies are essentially a bridge between two concepts, much like how a physical bridge connects two pieces of land separated by water, except in this case the 'water' is understanding and the bridge doesn't actually help you cross it.
They are merely quoting the analogy from Masad — for whom it makes sense if they are targeting nontechnical users and not professional developers anymore.
It's kind of a beautiful turn of phrase, in that the filing cabinet is entirely superfluous, you can use almost any noun. "it’s a little like having two versions of a website’s sub sandwich — the AI agent can experiment freely with a development sandwich, but the production sandwich, which is the real thing that users interact with, is completely walled off".
"When you click a button on our website, a request is sent across the internet to our servers, it's a little like if a sockeye salmon was sent across the internet to our servers."
personally, the evolution of replit makes me sad. i remember writing some of my very first ever programs in python using replit in middle school, as it wasn't blocked by the school network and it was the best way of running arbitrary code online back then. i used it to execute java code for AP computer science in high school, and i improved a ton at using the terminal as well. At some point, I stopped using the replit web editor and was coding by full-screening the built-in terminal and using vim. it was a formative experience and really helped me develop as a programmer even though all i had access to was a locked-down chromebook. but now, going back to the website and seeing the first thing it shows to you is how you can "build apps using AI", not even being able to even create an environment to run some python code without talking to an LLM, and the company focusing on ARR and becoming "AI-native" and creating value and all that jazz, and it feels like the magic of learning to code for the first time has been lost. luckily, kids these days are spoiled with webassembly and can run pretty much whatever they want in the browser, so i'm sure the next generation of young programmers will be alright
I've used Replit for educational materials when teaching beginners Python and JavaScript. They had a nice product called Teams for Education. It was even announced that it'll become fully free (the original blog post with the announcement was deleted from their website) [0], but soon after that the company had pivoted to AI and later discontinued the Teams for Education completely [1].
I also used Replit's embedded widgets for occasional lessons, but they kept changing the UI and behavior, making it difficult to write consistent reliable documentation for beginners.
I think by now it's clear that the product is not meant for educators, like it was originally, so that's ok.
Oh you brought back fond memories of using the very first education product Replit had. It was so good. I was teaching GCSE CS and having a blast giving kids python with turtle to play with. I would just give them some multicoloured octagon and more complex shape, and they had to draw it with turtle.
They enjoyed it, I enjoyed it, and the teacher UX was great.
Sounds like these guys kind of flailed around for almost a decade and then jumped on the AI bandwagon. Nothing wrong with that, I’m just surprised someone decided to bankroll them for so long. Also not sure how successful they’ll be long term, since they’re presumably a wrapper around one of the big three in a highly competitive space.
I don't think they can keep it. What they have has already been replicated by the big guys, with just maybe a few months delay. It seems like they're usually going in the right direction, even ahead of others, but it's usually not any massive breakthrough, they're just being good at executing quickly the natural next steps.
I've seen bots promoting Replit on Steam forums of indie games.
If this is the way they are marketing their product, I don't see it as having a future. What I've seen in the Dwarf Fortress forums alone makes me want to avoid the company with a 10' pole.
Hey, Replit employee here. I'm pretty sure this isn't us (definitely isn't our marketing team's MO AFAIK). Can you email me some examples at james @ replit dot com so I can look into this?
I find exactly one post about Replit in the DF Steam forums and it's definitely not from a bot account. Players who have just discovered vibe coding love to tell devs "just" to use something like they've been enlightened.
I took it for a spin spin to build a simple task manager. It worked for 45 minutes, built half the solution and then asked me for money. Compared to the other options in this space, it seems like an expensive solution.
If OpenAI and Anthropic's models are good enough at coding to make coding agents useful for non-technical employees Replit is targeting, why can't either of those companies use their AI to build a wrapper around their product as good or better than Replit? Has this guy never taken n -> infinity in a problem before?
We used to use it for coding interviews during COVID. It never struck me as anything special and I'm not convinced saddling it with AI improvements it's prospects. In fact it might make them worse given it's general applicability to learning which will surely be surpressed by the worries of an AI takeover.
I don't understand the appeal of Replit. Claude and other agents do what they do for much cheaper and efficiently. Is it for newcomers who don't know anything about software development, nd Replit fills the gap with basic user instructions and structure? Why is it even a thing while real agents are there?
> But the technical achievement wasn’t translating into revenue growth, and by last year, with the company at 130 employees and burning through cash, Masad said he had to make a painful decision. “I looked at our burn, and I looked at our progress on our revenue chart, and it just didn’t make any sense. The business wasn’t viable.” Replit cut its headcount by 50%, bringing it down to around 60 to 70 people at its lowest point.
> Plus, Replit has another unusual advantage for a startup: a $350 million war chest. Despite raising $100 million in 2023, the company “hadn’t touched” those funds by the time it raised this latest round, Masad told me. The company is capital efficient by design, though Masad joked that as an entrepreneur who grew up watching his refugee father struggle, “one thing I need to learn is to be less frugal and start spending money.”
Amjad is one of the only leaders in tech (along with Paul Graham) speaking out against Israel's apartheid and genocide. He also happens to be Palestinian. I will happily give him my money and invest in ethical platforms. Replit's a really good option for AI coding anyway!
I've used it for creating a template for an app. It usually creates a pretty accurate app with at least a basic UI and some interactions. The biggest issue is that once you've created the template, it's pretty tough to add some more complicated features using the AI.
Too much money for a business that really won't find a way to justify the investment. Could have been a terrific lifestyle business and become extremely successful.
This guy is exactly like the get quick rich scheme guys.
I believe he initially had a vision for the product but now the whole schpiel is quotes like "a kid made $180k on an app in 4 weeks" like he was on and on about in podcasts.
Ok if its so easy, why don't you use your own website to do a bunch of $180k / m apps? How many companies can have employees produce 12 different apps per year and a combined revenue of $2.1m per employee? Also everyone can do it so you can pay them minimum wage you don't need developers.
But of course nobody is going to make any successful apps other than the people in the marketing materials.
44 comments
[ 4.4 ms ] story [ 45.7 ms ] threadI gotta wonder who the median techcrunch reader is if the writer/editor felt it necessary to explain the point of having a staging and prod environment, and with such a pointless analogy. We surely cannot understand what a database is unless we're told it's like a filing cabinet, right?
That's a basic concept of writing. Journalism should be accessible, so even if you know what a database is and how to deploy it in different envs, you shouldn't write assuming that. If a large portion of your readers don't know what you're saying, you've failed as a writer. If your readership includes high school students, you write with that as the baseline.
Richard Feynman certainly didn't write as if he assumed the reader knew particle physics. Be like Richard Feynman.
Oh the things he did to filing cabinets, especially "secure" ones...
The bigger question here is why prod/staging wasn't an obvious design choice in the first place!
Is probably a consumer tech enthusiast and not a software developer.
"When you click a button on our website, a request is sent across the internet to our servers, it's a little like if a sockeye salmon was sent across the internet to our servers."
I also used Replit's embedded widgets for occasional lessons, but they kept changing the UI and behavior, making it difficult to write consistent reliable documentation for beginners.
I think by now it's clear that the product is not meant for educators, like it was originally, so that's ok.
[0]: https://web.archive.org/web/20240924020257/https://blog.repl...
[1]: https://blog.replit.com/update-on-teams-for-education
They enjoyed it, I enjoyed it, and the teacher UX was great.
If this is the way they are marketing their product, I don't see it as having a future. What I've seen in the Dwarf Fortress forums alone makes me want to avoid the company with a 10' pole.
Tells you everything you need to know about the company and its leadership.
[1] https://news.ycombinator.com/item?id=27424195
> Plus, Replit has another unusual advantage for a startup: a $350 million war chest. Despite raising $100 million in 2023, the company “hadn’t touched” those funds by the time it raised this latest round, Masad told me. The company is capital efficient by design, though Masad joked that as an entrepreneur who grew up watching his refugee father struggle, “one thing I need to learn is to be less frugal and start spending money.”
HN: It sucks now
As an online IDE they would have had a chance. But when they pivoted into AI, they decided to enter a highly crowded place with very strong players.
Recently OpenAI and now Anthropic are developing mobile clients as well:
https://www.testingcatalog.com/anthropic-prepares-claude-cod...
I believe he initially had a vision for the product but now the whole schpiel is quotes like "a kid made $180k on an app in 4 weeks" like he was on and on about in podcasts.
Ok if its so easy, why don't you use your own website to do a bunch of $180k / m apps? How many companies can have employees produce 12 different apps per year and a combined revenue of $2.1m per employee? Also everyone can do it so you can pay them minimum wage you don't need developers.
But of course nobody is going to make any successful apps other than the people in the marketing materials.
https://intuitiveexplanations.com/tech/replit/
Well, that’s highly unusual. I recon it is because replit covers the whole production pipeline, from specification to implementation to deployment?