Darklang Goes Open Source
blog.darklang.com165 points by stachudotnet 15 days ago
165 points by stachudotnet 15 days ago
https://blog.darklang.com/goodbye-dark-inc-welcome-darklang-... includes this, which is a really interesting pattern that I don't remember hearing about before for this kind of company:
> In conversation with our investors and the board, we believed that the best way forward was to shut down the company, as it was clear that an 8 year old product with no traction was not going to attract new investment. In our discussions, we agreed that continuity of the product was in the best interest of the users and the community (and of both founders and investors, who do not enjoy being blamed for shutting down tools they can no longer afford to run), and we agreed that this could best be achieved by selling it to the employees.
Any other examples of that? I'm particularly interested in that for this kind of software product.
> Any other examples of that? I'm particularly interested in that for this kind of software product.
As far as I know, this pattern is not uncommon among traditional businesses. King Arthur Flour Company is the largest one that comes to my mind, but on a local level; grocery stores, restaurants, mechanic shops, plumbing businesses, etc very often "change ownership" this way.
In software, it's pretty common in informal OSS project to transition ownership this way when the original owner/author loses interest or is otherwise unable to maintain the project.
In terms of commercial sortware, something like SketchUp comes to mind, though it's not exact path. It was a startup, acquired by Google, then spun off again with its employees
I wish the owners of Komoot would have done this.
They sold the company because they didn't see a future of growth, and the employees were notified of the sale of the company just a couple of days before.
The new owner then fired most of the employees, it's an Italian "tech company" (Bending Spoons) which already bough companies like Evernote, Brightcove or WeTransfer, and has nothing to do with the outdoors.
Komoot was the best outdoor-community app in Germany and very popular in Europe, made mostly for hiking and biking.
You can see in this really moving video, made by the employees after they got fired, how much they loved their team:
That seems like an exemplary way to handle a failed rocketship that nevertheless produced something useful to certain customers. Big thumbs up to those who made it happen.
> an 8 year old product with no traction ... and we agreed that this could best be achieved by selling it to the employees.
Can someone with more business sense than me explain this? Why would employees want to buy an 8 year old product with no traction? At face value this sounds like a "holding the bag" scenario, not?
They're not buying the product. They're buying the company so they can pivot and implement their vision.
Basically, the investors lost interest but the team is passionate and see a path to success. They won't be maintaining the old product, they're going in new direction.
It's actually the opposite. They're buying the assets (the software), not the company.
> To ensure continuity for users and fans, as well as to continue building what we regard as an important technology, Dark Inc has sold the assets – the Darklang language, the blog, the hosted service, the Discord, etc, darklang.com, etc – to a new company started by Dark Inc's former employees.
(from https://blog.darklang.com/goodbye-dark-inc-welcome-darklang-...)
In other words, there's a new company (Darklang Inc.) that has purchased the assets of Dark Inc. (for probably relatively little money). This clears the cap table, making it easier for the former employees (now founders) to raise money for the new corporation.
So, they pretty much bought the entire thing without transferring company ownership? It's weird to split the information on multiple posts.
They want to maintain the community they have because it's the real asset for any open-source project, which is the direction they're betting on.
> investors lost interest
The founder lost interest, he started a new company and he is the CEO of it!
Have you ever been to a great restaurant that happened to be on the wrong corner? Or been at a company where one change in execution made or broke the company? My guess: the founder lost interest but the employees still believed in the [impressive] tech. Because of the lack of traction: the cost of the tech wasn't prohibitive for the employees?
> "holding the bag" scenario, not?
Only if they are buying it for what the investors had already put into it, which is not likely. They most likely discussed how much the investors values physical assets and trademarks the company holds (like how much they are likely to get back in a bankruptcy) plus whatever makes a deal fair and maintain a happy cordial relationship with said investors for future endeavors.
Corrected by @justincormack. Post was about Docker Inc.
Thats not what happened. There was no new company, the company continues to be the same. They didnt sell off the business for money raised.
Really? So it was just a recap?
Who are we talking about here? Looks like a comment was edited but now it seems like you're talking about Darklang?
They were referring to Docker, after some claims regarding some corporate activity (forget details)
I mentioned docker was recapped but also wrongly assumed the existing entity was sold off and they formed a new entity in that process. Justin kindly corrected me.
Was previously "source available" but is now Apache 2. Good choice IMHO!
Also looks like it required their cloud setup to run, you previously couldn't run it locally. Now you can, so I think it's moving in the right direction!
I guess darklang was too far ahead in their thinking in some areas and choose the wrong path for other. I really liked the deployless idea, but would have loved in even more on-premise. No way to get the data to stay in Europe.
Making hard connections between the editor and the lang was interesting also. Seems like they have moved away from that.
Hope there is a easy way to set it up locally, i was really intrigued when they first launched
Yes, the next version will be able to set up locally - you'll be able to install a single Darklang binary and run any darklang program without any further steps. See the explanations on https://darklang.com homepage.
The issue with the hard connection between the editor and language is that each change becomes a massive undertaking. Making a language improvement was much much much simpler than making the editor change to support it.
It was a bit hard to understand what is coming and what deprecated. As soon as i went into documentation I was send to "darklang classic".
How are the deployless senario now? Where you first serve only yourself then your team, then beta, then everyone... Or something similar to that. I really liked how that story was told and how much complexity it removed
> We're now building Darklang to run locally as a CLI
Dark's structure editor looked promising. I'm really disappointed that the project moved away from this because a hosted visual programming environment felt like the whole value proposition in the first place.
Was it the pivot to AI that killed this, was it issues with the design of the language or was the structure editor just not as useful as it seemed?
Stachu is really interested in bringing back the hosted programming environment in some form fyi. Just need to get the basics working first.
I've been following Dark since its inception and found the idea inspiring. I'm happy about today's announcements and look forward to seeing what comes next.
On a personal note, I'm curious around the move to F# as the implementing language and wonder if there will be ports to other languages now that it's open source.
F# from OCaml isn't much of a leap, though. F# is basically OCaml modified to fit .NET datatypes.
I was following this since it started. I dislike modern code experiences because of many of the things, especially devops, they fixed. But it being not fully open source always held us back and we developed our own solution with common lisp and some typing sauce that keeps us free from all the modern dev and deploy crap. Everything runs easily on your laptop as well as in the cloud without any difference and, outside running a simple script to set up some vps or cloud hoster. Happy darklang went apache license and looking forward how their deployment looks now as not self hosting is just never ever an option: all companies become frauds in the end so we need to be able to move any time.
Epic's new language, Verse, is also well poised for the "immutable" future of AI agent coding. Verse uses an effects system, and the <transacts> effect is required for any function to change a mutable variable. These changes are transactional, so if you have a failure in your <transacts> function, any changes it made are rolled back at exit. Code still looks imperative-ish, but it's both safe and pure.
(Or, it works something like this. The documentation is hard to understand; I'm working mostly on memory from their keynote)
Verse is totally proprietary and is still tied to the engine. Trust me that it sucks
I’ve watched the demo video, and gone through the discussion here but I’m still not sure what the core use case for Darklang is. It feels like I'm missing something obvious.
Can someone explain the practical problems this is solving better than existing tools like Python or other backend stacks?
Also, genuine question: how was the team able to work on this for 2+ years without revenue or traction? Was there still funding left, or was this a side project during that time?
The theory is that because Dark offers a nice interactive environment when you have a request come in you can inspect the data, "fix up" your code, rerun it quite smoothly. Completely removing the deploy question. Really good for glue projects IMO. Think stuff like google scripts (though google scripts are much more miserable to ues)
Many moons ago I had tried to play with Dark and unfortunately the surrounding language stuff really didn't work out for me though. I really liked the concept but ultimately I think I would have preferred if they had just built out the environment on some other language.
Even just lua would work decently well for their use cases, if paired with a good "standard" library.
The UI was fun to use, in any case.
EDIT: to be clear I'm a bit of a type weenie but for the stuff Dark could be good with you really sometimes just want to look at the JSON and poke at it, and your type systems aren't going to be helping you _that much_
I don't understand. Hot code reloading on my dev laptop gets me this kind of quick iteration when I'm solving problems. But why do I want that in production?
A user send a request, and gets a response you don't grasp. In Darklang-like envs that request data is held onto for a bit, you can replay it and tweak things. And you don't have to think about environment differences because you're working _in production_.
You could be taking webhook values from some external service that starts passing in unknown values. That service's documentation might just be outright wrong about its behavior. But you don't have to play post-mortem forensics, as you can hold onto information.
This can be valuable for one-offs where you don't mind breakage, with relatively low request coutns and "time to fix" is what you're optimizing rather than general uptime/downtime. So some silly discord bot could work off of this, but you probably don't want your B2B SaaS being in cowboy mode.
Sometimes you want a thing that just does something and you don't care about how busted it is, but you want it to be _very easy_ to fix up issues you do care about.
Because the title at the top links to the blog (not the homepage) I was a bit puzzled as to what Darklang actually _is_. One more click on a similar logo reveals "Darklang puts everything in one box, so you can build CLIs and cloud apps with no bullshit, just code."
What are the pros and cons of Dark Lang vs Unison?
In theory, Dark and associated infrastructure for running Dark apps is the perfect companion to LLM based vibe coding… I think, and I am just understanding this now. The goals of Darklang were always “no this, no that, not that either.” And so the focus was not on targeting 3rd party stuff of questionable design, but rather a single integrated set of patterns that abstracted the messy bits away.
Turns out, the messy bits are the things that turn your vibe coded Twitter clone into a full time operations job…
Been following from the sidelines for years. Wish Paul and team invest in a person good at docs, some web design and copy writing. The website, docs, visuals, examples, typography, are just very confusing and feel amateurish.
Not blaming. Not everyone is good at everything or wants to make time for it.
But a good, well structured landing page with great, real life, examples and good hierarchy backed by awesome docs will make a ton of difference adoption wise. I hope.
congrats Paul for landing the plane. i respect your dedication to the cause you see more pressing and have burned bridges for and hope T4P contributes toward the permanent peace we all want.
Am I the only one that thinks this is nonsense? Can someone please explain why I would want this over python or others?
Congrats on Dark for making it this far!
Relevent timeline:
https://blog.darklang.com/dark-announces-3-5m-in-seed-financ... (2019)
https://blog.darklang.com/dark-and-the-long-term/ (2020 - in which the team is fired to extend runway I guess to today)
TL;DR: We’re taking a longer term approach to building Dark. As part of this, we’ve made the difficult decision to shrink Dark’s team, and to change how we build both the product and the company."
So where do we go from here? Right now, the team is just me. I am committed to realizing the full vision of what Dark should be. Dark is financially healthy for many years, and there is time to think and to plan. I plan to involve the community much more in Dark’s growth, and slowly rebuild the team at a pace appropriate to the product’s maturity, focusing on a small, tight team that can wear many hats.
Start of a new rewrite: https://blog.darklang.com/dark-v2-roadmap/ (2020)
Two years later: https://blog.darklang.com/backend-rewrite-complete/ (2022)
seemingly a new pivot to "all in" on AI?: https://blog.darklang.com/gpt/ (2023)
No news, one year later https://blog.darklang.com/an-overdue-status-update/ (2024)
Would be interesting to the Dark team to revisit this post, which is a look at PL funding models:
https://blog.darklang.com/how-to-fund-caramel/
Building programming languages is hard especially when you're not backed by a company. I think Eve (I worked on that one) and Dark were the two major VC funded languages, and at this point I don't think that's a good model for funding this kind of thing. You need waaaaay more that 2-3 million; most of that is funneled directly in to SF landords pockets. Something more like the Mojo people have gotten is what it takes (they've raised upwards of 100 million).
Anyway I can't wait to see where Dark goes in the future, and what their funding model will be going forward.
> You need waaaaay more that 2-3 million; most of that is funneled directly in to SF landords pockets
Which is why you should build your team in Denver, Minneapolis, Chicago, Detroit, etc. There's a competitive advantage to hiring outside the SF tech bubble today. Over the last 5 years the network effects in SF have begun to evaporate.
Agreed in hindsight, but at the same time there was no place else where a couple of 20-somethings could grab a cup of coffee with a VC and walk away with a handshake deal for $2 million dollars. That just didn't happen in Denver et al in 2014.
Does that still happen today? Anecdotally there has appeared to have been a massive funding crunch for pretty much anything that isn't virtual healthcare or AI since COVID passed, though I'll be the first to admit I'm not in the know on a lot of these things.
My anecdata is the opposite. I know people getting funding for non-AI projects. The most "nontraditional" one I can think of is Nautilus [0].
We (darklang) are fully remote! One person in Vermont, one in Algeria. :)
This is a pretty weird take. Talent in Denver, Minneapolis, Chicago etc. is not a whole lot cheaper than in the Bay Area. Employees are getting a large (majority) of their comp as options or RSUs, so that makes the delta even smaller, you're just talking base salary.
If that's "make or break" for you, then something is wrong. There are plenty of reasons to want to have a distributed workforce (larger talent pool in general, passionate employees) but saving money is the least important one here.
> You need waaaaay more that 2-3 million
Mozilla alone invested an eight digit amount in Rust.
Meanwhile Elixir had no such backer.
What do you consider the financial and developer contributions of Dashbit née Plataformatec, Dockyard, and Fly.io?
Dashbit and Fly.io were announced in 2020, Dockyard started using elixir big time in 2015~.
Elixir started in 2012. Phoenix in 2015. Certainly none of those companies had millions to spend on Elixir at the time.
My point was that José and Chris being continuously employed working on the language and Phoenix respectively (by the companies I named) are probably considered a form of commercial backing by most people. If you want to split hairs about specific timing or magnitude of investment, that's kind of nitpicking IMO. It's not grassroots at basically any point.
José founded Dashbit after Plataformatec got acquired but up until ~2020 Plat directly funded the creation and iteration of the language, with it originating as an internal research project.
https://elixir-lang.org/development.html
José Valim created Elixir in 2012 as a Research and Development project inside Plataformatec.
Dashbit José Valim has openened a new company to help startups and enterprises adopt and run Elixir in production.
Those companies employ a useful fraction of the language core team, too.
Ericsson did a lot of the heavy lifting with BEAM.
The same could be said about Darklang's underling language, the browser, and javascript.
I don't think open sourcing is going to fix their adoption issue. Like the other comments mention, you need to be worth the time investment to gain traction. If Dark was truly as revolutionary as it was marketed as, it wouldn't have had problems staying source available, IMO. Folks will pay or put up with whatever it takes to be in the ecosystem (such as CUDA).
I agree it won't fit it, but IMO it will remove one of the barriers to adoption. The problem with doing something revolutionary, is that it's only going to be revolutionary in some ways, and it has to compete with things that are mature in ways you are not. And the original version (now called Darklang-Classic) was quite immature in an awful lot of ways that made it difficult to build on.
That's being addressed with the new version of course!
Mojo was also less ambitious in a lot of ways. It blows my mind the Eve and Darklang guys raised so much money without a lot of momentum. I'd think you'd go the other way, start an Open Source project, spend 10+ years gaining a community and refining it, then raise money.
In both of the above cases, the founders just got bored of their project before they found PMF.
You just have to look at the landscape at the time. There was a lot of money to be had if you promised the sun and moon, because $2 million wasn't a lot compared to the potential upside. The problem was, and this is what Paul found out too, they wanted to see typical startup metrics before they'd put more money in, and it was always going to take more than $2-3 mil. You just can't demonstrate those with a concept of a language.
Do you think Unison will suffer the same fate?
No, I mean these low bus factor languages don't really die as long as the BDFL keeps working on it. Biggar keeps Dark going through thick and thin. Chiusano likewise with Unison. Even if their Unison public benefit corp runs out of money, Chiusano could probably do what Dark did and buy the IP. With my programming language I'm making sure that there is no IP and therefore nothing to own. It would probably be easy for Biggar to just wash his hand of Dark as well but it takes guts to keep going in a direction you know it right, and so I'm happy to see the project continue.
I'm glad to hear it as I'm very interested in Unison. (I've been watching them from the sidelines for ages.) For those not in the loop, it's a language where your code is stored on disk in AST form, not textually. The AST representation is smart in many ways - for example, renaming a function is an O(1) operation, regardless of how often the function is used. They also have a way to serialize Unison functions and send them over the wire to other Unison programs, which is pretty sick. Their site is here: https://www.unison-lang.org/
(I'm mostly interested in it because I think it would be an ideal language for videogame scripting & modding)
FWIW we also store ASTs, not text. We offer many of the same benefits. Some day (soon?) I'll write up a full comparison between the two, because it seems a common ask.
I'd love to see where BABLR falls in that comparison. We're trying to take the same kinds of tricks and bring them to every programming language at the same time
I can’t see teams adopting Unison (or similar languages) without a way to store code in Git.
Maybe the editor can load text and do structured editing. Maybe the runtime can send functions across the network. Great. But not using Git for storage and review is just too alien for most teams to even consider.
> it's a language where your code is stored on disk in AST form, not textually.
Doesn't this result in vendor lock-in for editing the code?
Paul had created Circle CI, so I can see how investors would at least be trusting. Rightly so, I think, as he's not just talented but he knows talent.
Mojo's promise is the same code, but faster.
They were planning some language extensions but it's more like a compiler project than a programming language project.
The truth is, most developers don't want to learn a new language.
They will jump through extra hoops just to use their favorite one (e.g. Airflow).
Successful languages appear when there is an extreme market demand (C++ providing OOP over C) or, more commonly, a hot new platform that people want to get in on (JavaScript, Swift, Kotlin, C#, ...)
For most people, new syntax / semantics is considered a negative and there needs to be some massive upside to overcome that.
There was also Wing cloud (fka Monada) and there’s Mojo by Modular (https://www.modular.com/mojo.)
Feels like two types of companies raised money: - Companies trying to couple the cloud with a programming language. - More recently, companies trying to couple GPUs with a programming language/alternative to CUDA.
Will be curious how this generation goes.
Very cool. Now make a page that clearly explains what it is!
[flagged]
[flagged]
This is a generic tangent, and it's against the guidelines to do this. Additionally, the statement "I stopped reading" is just the kind of thing we want to avoid on HN, as it's empty signalling, rather than substantive commentary about the topic. Please avoid commenting like this on HN.
What an interesting project - once this rolls out, absolutely want to try it. There is something delightful about the fundamental idea of a "canvas" as the code editor.