GitHub Accelerator: our first cohort and what’s next | Hacker News Reader

GitHub Accelerator: our first cohort and what’s next

github.blog

233 pointssebg101 comments

> JessicaTegner/pypandoc Pypandoc provides a thin wrapper for pandoc, a universal document converter.

To be frank, that doesn’t seem like a project worthy of $20k to me. Not very many commits recently [0], it has 918 source lines of code in the main package, and 547 lines of test code — it even describes itself as a “thin” wrapper, and the only thing it does is converting some arguments to Pandoc flags and calling the Pandoc executable. How many more features could be added to this project? What is it missing now?

[0] https://github.com/JessicaTegner/pypandoc/commits/master

Looking at the developer's profile, clearly GitHub recognized this developer's right to positive affirmation.

Proud of GitHub for not caring it is a 918 line thin wrapper, this developer deserves to be set aside and carried to the top because of the challenging path they are on.

Who cares if the better project gets the money after all.

Edit: This post is so ambiguous that first I was discouraged when it got upvoted. :)

To be frank (and then polite and constructive), I think the relevant question is "What am I missing?" rather than "What is it missing?"

..and my guess would be that this is something like a cohort experiment. I'd bet they selected projects to have a broad sampling of kinds of projects, so they could see what working with them was like and what kind of impact they'd have - similar to a marketing trying out different audiences and types of ads, or startups running market experiments, or VCs shot-gunning to see what sticks, or developers doing 'probes' into problems and/or with technologies.

AFAICT, GitHub wanted to get some data around the questions "What's involved in us getting funding to open source projects / developers, and what kind of impact can that funding have?"

Even if your judgement of this specific project is apt (that it's weirdly small and inactive for receiving a grant), that may be the point.

Was there a grant application? Maybe they support the author to do something interesting, which is not evident from simply looking at the repo.

Of course there are reasons beyond technical ones for including a project in an accelerator cohort. I attempted to view the application to see what types of questions GitHub asked, but they've taken it offline now. On a similar token, one of the other projects is an open source Linktree clone... the MVP of that concept is so simple and there are so many of these projects, that it suggests technical complexity was not a highly weighted factor in the selection process.

That said, I have a few ideas:

- Although it has very little recent activity, it has dozens of contributors, was a Hacktoberfest project, etc.

- There is also often the desire to create diversity in an accelerator cohort. Every accelerators faces pressure for diversifying demographics. This also creates a flywheel effect to improve diversity in future batches.

- Another idea is that perhaps someone saw the potential for this to be turned into a SaaS. Non-technical people need to convert document formats every day. Maybe this does something Pandoc doesn't or that would make it a candidate for that?

I'm not saying this is a bad thing. But why $20k? It's obviously not a salary... It could pay for infrastructure if a project needed it, but if the project doesn't... I'm having a hard time understanding how it helps sustain the project. I personally wouldn't do both a full time job and work on an OSS project in all my spare time, for the money. If I really need that project I'm gonna do it regardless, and if I don't need that project, sacrificing all my free time isn't worth $20k.

$20k for 10 weeks of work seems like a good deal. It’s only 10 weeks so you can’t just quit your job; but if you are in between jobs, can go on leave, or (most likely case) work part-time and just don’t get paid enough, it’s a big help. And it’s certainly better than nothing which is what most OSS devs get unfortunately

Depends on the country of stay. If you are in Pakistan or India then those 20k is quite a lot. Not so much in Italy or Spain.

20k for 10 weeks is a pretty decent salary, even in the EU. Might not be super competitive for short term work but it's definitely enough to live on in most places.

Even in most (all?) Mediterranean countries, 2000/week is pretty good pay, would work out to 8000/month, which is on the upper-end when it comes to writing code for the usual web stuff. Pair it with that you get to work on FOSS, I'm having a hard time imagining a similar position with similar pay.

It'd depend on several factors, won't it? Those who are already working on OSS projects have found their balance (or willing to work despite challenges). This amount might make small or big difference for them. You should also consider that how much this is worth will depend on where they live too (for example, my yearly expense is less than $2000).

I view the money as a nice-to-have bonus. The big benefits of the program are:

- validation by Github of your open source project

- the signal boost associated w/ that validation

- the actual content of the course (licensing, managing issues, etc.)

- meeting other open source people to discuss what is working and what isn't

20k is nothing to sneeze at, but, as you say, most people aren't in the open source game for the money...

It definitely depends on the maturity of the project and the country you live in.

I was selected to be part of this first cohort with Mockoon. I'm from Luxembourg, so 20k is surely not a life changing amount of money. However, I quit my job two years ago to focus full-time on it. I think that I can make a living (probably a mix of paid services and sponsoring) while keeping the app open-source. But it's a long hard road and I am freelancing on the side which gave me the right balance but not complete peace of mind, financially speaking. So, while it's not a life changing amount, it will buy me some precious time. I also feel really motivated by this awesome recognition (after 5 years of work on this project) and I am looking forward the 10 weeks of coaching. So, all in all, it's a bit of a life changing event for me :)

(Edited a bit for clarity)

Congratulations! I hope it turns into a great experience for you.

Thank you :)

I maintain one of the selected projects. $20k means that I can focus on the project without taking any freelance work on the side for a while. Hopefully making the project more attractive to sponsors, hopefully snowballing into more OSS and less freelance.

By the way, https://github.com/sponsors/code-hike/

Yeah that's not a lot of runway. But it could be reasonable for a student or for someone living in a low cost of living location. Or if you have savings from your job and just need a little boost to get started. Or if you're okay with living frugally, which seems to be how people like Andreas Kling or Andrew Kelley get by. But yeah it's not really a sustainable model.

Middle income is $10-$20 per day[1]. $2k per week = $400 per day = 20x - 40x middle income. Minus taxes. And I guess you need to factor in that the figure includes people who are not employed. But still should leave you in the ~10x group.

[1] https://www.pewresearch.org/global/2015/07/08/a-global-middl...

Finally, software engineers can break out of the $10-20 per day income bracket!

$20k is one or two weeks in a HFT sweatshop vs. trading your project/dignity into M$?

Not seeing the value prop from GitHub here.

Because it's just 10 weeks. It's very clearly not a "salary" (as "salary" is usually measured by years), but more like a contracted work that lasts 10 weeks. $20k is more than most jobs posted on Upworks (etc).

(Yearly salary seems to me to be a US thing. Maybe it's an Anglophone thing?)

Don't forget that GitHub belongs to Microsoft. This is yet another attempt at controlling the OSS world.

(Not to mention control developer salaries.)

A rain of silent downvotes, how interesting...

FWIW, we (Dioxus) are using the $20K to fund an internship for one of our contributors. We're hoping the program will act like VC funding (marketing, really) and help us unlock larger corporate sponsorships.

[1] http://github.com/dioxusLabs/dioxus

> But why $20k?

One side note, is that it appears to be $20k per maintainer vs per project.

From the post:

> GitHub Accelerator is a 10-week program where open source maintainers receive an initial sponsorship of $20K to work on their project ...

> The 2023 cohort has 20 projects, with 32 participants ...

As others have noted, it's not a lot of runway, but maybe enough to not have to freelance on the side and that full-time focus for 3 months could be enough to snowball into a full-time "ramen-profitable" business (maybe). It'll be interesting to see how the first cohort plays out.

No, it appears to be $20k per project, as per https://accelerator.github.com/#faq --

"$20,000 stipend per project ($2,000 per week of participation) for the duration of the 10 week program. If there are multiple participants from a project, the stipend is divided evenly between them."

Thanks for catching this... that's a bummer.

dropping work for three months for a $n fellowship is the same thing as switching to a $4n job without benefits that you know will fire you in three months. $20k is not enough for that sort of thing

I'd say this is more about positioning themselves as "someone who cares about open source" more than anything else.

Their main product is proprietary, and they have a looong history of vendor lock-in and anti-competitiveness (including: Extend, Embrace, Extinguish). Giving out these grants lets them present themselves as someone who cares and funds open source, even if their core business stands strongly against it.

Perhaps people in tech mostly now GitHub's and MicroSoft's reputation. But if says "we love open source and even give out grants", its easy to understand why legislators would believe that both parts of the statement to be true.

Probably the cheapest way to advertise and build goodwill.

Looks like they peppered in some weird projects in there as well. I'd have expected some "hard software engineering" projects as well, but I guess not everyone is ready to be a spokesman for github's view of the world.

https://github.com/EddieHubCommunity/LinkFree : "Your profile will have links to your social media and content. You can also add your timeline, testimonials, and upcoming events that you are participating in."

Sounds lame. Is this really worth including in this list?

https://github.com/JessicaTegner/pypandoc: "Pypandoc provides a thin wrapper for pandoc, a universal document converter."

So random. Why not support pandoc instead?

It seems like a political statement by Microsoft. The list only has "trendy" stuff, carefully moves around any and all "AI" stuff, and none of it seems experimental at all. If I didn't know this was "Github Accelerator" and was shown the list, I'd have assumed this was some VC's list of prospective Open Source programs that could turn a penny or two down the road.

It seems pretty obvious to me that now that Github has cornered the "open source" market, it's on the "Extend" phase, looking to change how Open Source is approached.

https://en.wikipedia.org/wiki/Embrace,_extend,_and_extinguis...

The messaging was that an independent panel would select, and not Microsoft per se.

It says "Github accelerator" not "Github community accelerator" or something. Pretty sure micro$oft had a big hand in it, they wouldn't allow the pilot for their program to be so "unsupervised".

Great to see Dioxus there. I was looking at Rust GUI libraries and Dioxus seemed like one of the most interesting to me, but still felt pretty early in development. I did notice in their 0.3 release post they mentioned "This release represents an absolutely massive jump forward for the Dioxus ecosystem. We hope to ship future features more quickly into stable now that many of the desired breaking changes have been incorporated into the core library." so I am excited to see how the project progresses!

Parts of the Dioxus stack are also becoming framework-agnostic, like Blitz the “High performance HTML and CSS renderer powered by WGPU“:

https://github.com/DioxusLabs/blitz

A full list of projects that are being spun out as framework-agnostic:

- taffy: flexbox+css grid https://github.com/DioxusLabs/taffy

- blitz: html/css renderer https://github.com/DioxusLabs/blitz

- sledgehammer: faster than websys bindings https://github.com/demonthos/sledgehammer/

- native-core: salsa but for UI trees https://github.com/DioxusLabs/dioxus/tree/master/packages/na...

- tui: native-core+terminal rendering

We want to make sure every part of the stack is available for anyone.

We're super excited about the funding. It's letting us bring on some of the core team full time for the summer and part time for the year.

Our goal with Dioxus is to eat Electron, React Native, Flutter, Phoenix, Gatsby, and NextJS with one stack.

> Our goal with Dioxus is to eat Electron, React Native, Flutter, Phoenix, Gatsby, and NextJS with one stack.

What of Phoenix are you looking to tackle?

We love LiveView and we love the developer experience. Dioxus recently launched support for HTML-over-websocket (LiveView) and we're slowly working on wrapping axum/hyper in a Dioxus-focused API.

Thanks for working on LiveView! Next to Rust, one of my favorite ecosystems is .NET and Blazor has been a game-changer when writing desktop apps. I'm looking forward to having something similar in Rust!

Amen to that. It would be pretty amazing to write Rust for the UI everywhere.

Github star history graph of the batch:

https://star-history.com/#trpc/trpc&termux/termux-app&respon...

nuxt, trpc, termux, and responsively look pretty popular

Yeah, not pasting in my personal access token into a random website.

lol, yeah, I guess you can restrict scope but still, nice try

You can't restrict very much/very well. I think you still can't restrict to specific repos?

How much do we have to fundraise to give github to implement API tokens for which you can limit auth for more common use cases?

I know I have access to fine grained access tokens (beta) which let you get very granular on a per repo. I’m not sure if this is generally available or only because we are on Github Enterprise though.

You can use the new Fine-grained Personal Access Tokens [1] (as opposed to the original ones, now called "Personal Access Tokens (Classic)".

With the fine-grained tokens, you can literally grant this site:

- Repository access: none

- User permissions: none

- Repository permissions: none

[Or if you want to be really paranoid, make a burner account... ¯\_(ツ)_/¯.]

But also, star-history.com is not a random site — the repo [2] has 4k stars and has been around since 2015 and has been popular for over a year.

If you've ever tried to make a public app that depends on GitHub APIs, the limits without taking PATs are very, very low — totally insufficient for an app with any popularity, so there's not really an alternative.

[1]: https://github.blog/2022-10-18-introducing-fine-grained-pers...

[2]: https://github.com/star-history/star-history

its loading without a PAT now for me.. presumably because it was able to cache other people's requests

As the maintainer of one of the selected projects (https://codehike.org/), I really appreciate what GitHub is doing, putting the money and trying different approaches (last year it was $450K across 900 projects).

yay pomber! love code hike even tho i wasnt ultimately able to use it at my prev company, still want to use it some day because it is beautiful

Awesome to see Simon on there with datasette, and HTMX. Woot!

Best part of Simon being part of it, is that there will be a great record of it from his blog and TILs.

do you think that he has anything to gain from the accelerator? he should be teaching it given the crazy level of experience he has

Very cool, this looks like a great list of projects. I like that it's not a bunch of well-known and big names. It looks like stuff that is just getting traction and having a push with dedicated time to maintain could be a huge benefit for them.

I agree! At the same time its cool that they also picked a few proven projects like Nuxt or tRPC. Good mix :)

As a rubyist who loves ruby but isn't afraid to admit I see it declining in popularity... I can't help but notice there seems to be no ruby on the list. Plenty of python.

Maybe to avoid looking biased? Github already puts a lot of money towards ruby.

That looks awesome, when is the next batch coming? What are the conditions for the funding? My OSS project is a Dropbox like UI for S3, SFTP, FTP, NFS, SMB and pretty much every other file transfer protocol: https://github.com/mickael-kerjean/filestash Would that fit in your scope?

Any FOSS project you're open to work full-time is within scope, iirc.

Btw, what you're building sounds very similar to spacedrive: https://github.com/spacedriveapp/spacedrive

AFAIK Spacedrive hasn't delivered a release yet, so it's nice to see some competitors delivering.

Surprised that they haven't despite garnering interest and funding. Likely that it is not an easy problem to solve!

I've always been curious what sort of relationship https://nuxtjs.org/ has with Next.js and whether they will significantly diverge as their own entity (I haven't had the opportunity to use Nuxt 3 yet). Especially with all of the noise around projects like Remix.

Happy to see them on the list none-the-less.

Next is for React and Nuxt is for Vue...

Well obviously (I'm surprised two people have replied to state this fact), Nuxt's whole pitch was literally "Next for Vue".

I was talking about conceptually diverging instead of just being inspired by Next.js, given recent developments with Remix and https://astro.build/.

Nuxt is a separate entity from next. Nuxt is meant for same use cases as Next (or Remix) but their implementation is tied to how Vue works.

If they find remix API better, they are free to steal it.

looking forward to participating and accelerating htmx's old man cloud-yelling to levels not previously thought possible...

Congrats! I hope the money goes to even more memes and references. https://news.ycombinator.com/item?id=35532520

How do you hope that the sponsorship will help HTMX?

Its like Google summer of code but better.

GSoC was about bringing young people into FLOSS, this is helping people who actually maintain those projects.

I think they are vastly different, tho I am unsure which one would be more valuable in the long run.

I'm thrilled to see my project (https://mockoon.com) being part of this first cohort! It's a great opportunity but it's also good for open-source projects in general. It can only contribute to make our industry more sustainable.

Here is the list of the 20 projects that have been selected:

- https://github.com/analogjs/analog: Analog is a fullstack meta-framework to build applications and websites with Angular

- https://github.com/Atri-Labs/atrilabs-engine: The Python web framework to build production-grade apps

- https://github.com/bigskysoftware/htmx: Makes AJAX, Web Sockets, etc. available directly in HTML

- https://github.com/code-hike/codehike: Tools for building all types of code walkthroughs: blogs, docs, slides, tutorials, etc.

- https://github.com/DioxusLabs/dioxus: Friendly React-like GUI library for desktop, web, mobile, and more, written in Rust.

- https://github.com/EddieHubCommunity/LinkFree: Connecting with your audience with a single link, showcasing your content and projects.

- https://github.com/FashionFreedom/Seamly2D: Design CAD to democratize and de-centralize fashion design & production.

- https://github.com/fastai/nbdev: Increase developer productivity by 10x with a new exploratory programming workflow.

- https://github.com/formbricks/formbricks: Open-source Qualtrics alternative

- https://github.com/GyulyVGC/sniffnet: Cross-platform application to comfortably monitor and analyse network traffic

- https://github.com/JessicaTegner/pypandoc: Pypandoc provides a thin wrapper for pandoc, a universal document converter.

- https://github.com/mockoon/mockoon: Mockoon is the easiest and quickest way to design and run mock REST APIs.

- https://github.com/nuxt/nuxt: An intuitive framework for building web applications, built for the edge.

- https://github.com/responsively-org/responsively-app: A dev-tool for web developers that aid in faster responsive web page development.

- https://github.com/simonw/datasette: An open source multi-tool for exploring and publishing data

- https://github.com/spyder-ide/spyder: The scientific Python development environment - https://github.com/strawberry-graphql/strawberry: A GraphQL library for Python that leverages type annotations

- https://github.com/termux/termux-app: A terminal emulator for Android that provides 2000+ linux packages to code on phone

- https://github.com/TimothyStiles/poly: Poly is a fast, well tested Go package for engineering organisms.

- https://github.com/trpc/trpc: End-to-end typesafe APIs made easy. Never write another API contract again.

Edit: formatting and updated repo links

super cool congrats :)

Exciting to see both Termux and HTMX be selected, looking forward to what comes of it

Looks like Github linked to Nuxt's old repo and not https://github.com/nuxt/nuxt

Thanks for noticing, will check with owner and GitHub team.

So glad to see codehike here! Pomber has been building amazing open-source projects and codehike will be a game changer for documentation.

Thank you!

No problem man. I don't know if you remember, but you helped us add Codehike to our documentation some time back!

This list is great to see because as a supporter and user of open source projects, I am glad smaller projects are getting this too! Hopefully it keeps going!

Hyped to see htmx and datasette!

Awesome to see trpc and Mockoon on here!

Thanks! I am glad to be in this list too (Mockoon) :)

Microsoft has had a fraught relationship with the developer community since, well, forever. They seemed to think that buying a developer favorite like GitHub would mend those fences.

Nope.

The more they own, the less I like them.

VSCode, Github, TypeScript...

really looking forward for AnalogJs to compete with next/nuxt!

Amazing list of projects!

Reply on news.ycombinator.com