Hacker News Reader

> But all forms of coding and media editing (the main tasks that are truly CPU/GPU-bound, the area where something like Mighty would help) have remained as local desktop app affairs.

Media creation maybe, but I feel like professional software engineering might move soon.

At the beginning of the year I moved from a company where I worked entirely locally, running a local stack, developing in a local editor, etc, to a company where I work almost entirely non-locally. That happens to be on a box under my desk, but I only use it over SSH. It's beefy, and I could get an even beefier VM somewhere if I wanted. All my editing is now in a web-based VSCode instance which has been much closer to desktop VSCode I was using before than I expected. All my builds happen remotely. It's honestly an amazing experience. I think things like GitHub Codespaces have so much potential here.

I agree overall, but I don't think it'll be VS Code in the browser, I think VS Code's remote features with the client-server model (client: VS Code's UI running locally, and server: the code executing and file-system running on the server).

I've also tried VS Code in the browser, and personally I find it an absolute unusable mess, the key binding space is just way to overloaded for a complex app like VS Code, and the browser itself to co-exist (e.g., many of VS Code's important bindings get eaten by the browser itself). I think that all that really matters is that the UI for complex apps runs locally.

(But I could be wrong here, I don't make the mistake of extrapolating my own experiences to other users. E.g., I also find VS Code to be so slow I avoid, but most users couldn't care less. https://news.ycombinator.com/item?id=27360494)

Good point, remote dev is probably the more appropriate tech for most orgs.

FWIW, I'm incredibly picky about shortcuts, and as long as you use Chrome to "install" the web app, it basically works fine.

VS code has the best remote features I have ever seen, good enough that I once got annoyed VS code had freezed just because the network died, then _realized I was doing the work over the network_. I hope it will stay a local app with powerful remote capabilities.

What hardware do you use to talk to the box under your desk? How do you work remotely with this setup, and what if the network VPN goes down?

Working in this way depends a lot on the stack you’re working with. For languages like Java or C#, where you can’t really work productively without an IDE, browser-based VSCode won’t cut it for most people. (I’d prefer a JetBrains IDE over VSCode too.) If your language doesn’t have a good remote debugging story, a remote-first stack won’t cut it.

This setup with a box under your desk doesn’t sound reasonable to me from a financial side either. According to your profile, you’re at Google, so I assume the hardware you use to access the box is a Chromebook. Sure, those are dirt cheap, but is a desktop + Chromebook combo really cheaper than one reasonably priced PC laptop with specs similar to the desktop?

I think I’m still happy with my beefy (spec-wise and kilogram-wise) laptop, having the ability to do everything directly on the machine (with no network round-trip for every operation), and still being able to do things if the VPN or the network goes down (with limitations, of course).

> What hardware do you use to talk to the box under your desk? How do you work remotely with this setup, and what if the network VPN goes down?

I use a MacBook Pro. I just connect to the machine. We don't use a VPN (see BeyondCorp) but something like Tailscale would probably work well. The fact the machine is under my desk and not in a data center is due to my laziness in returning it, nothing more.

I'm using VSCode for mostly Java. It's not perfect, but I prefer it to JetBrains products. JetBrains products are more capable for now, but I think their new IDE does remote development, and VSCode gets more IDE-like all the time. I think that's the direction things are going.

I'm using a MacBook Pro because I've been using Macs my whole life and have too much muscle memory to switch to something else.

As for having a beefy laptop, that only goes so far. Right now laptops tend to top-out at 32GB RAM if you still want plenty of choice, you make sacrifices to get more than that. Also many companies have policies or regulation restricting what can be on laptops.

If you are at home, your box is in the data center, then a BeyondCorp outage makes it impossible to work, right?

I agree that laptops can’t always do everything, but many things can be done comfortably well. At a previous job, I had 32 GB of physical RAM and 20 GB of swap (thanks to 5-10 RAM-hungry JVM apps), and things were quite usable for standard dev workflows.

> a BeyondCorp outage makes it impossible to work

Yes. So does a VPN outage if you're using that instead. If your laptop gets chewed up by a lawnmower, it's also impossible to work.

I can get things done with a VPN, $CloudCodeHosting, or network outage. Sure, not everything would be possible, but I can still write code and run it.

> a BeyondCorp outage makes it impossible to work

Err, sort of. Beyond Corp isn't really one single service that can be running or down, it's more of an approach, a bunch of different pieces, etc. I haven't seen an outage yet, but I don't think it would look like "the VPN is down", probably more like "this web app thinks I'm logged out".

I do recommend reading the BeyondCorp paper, although it's been a long time since I read it.

What's Figma got to with any of this? Figma is local compute just as much as any native app.

This is not about web vs. native but client-side vs. server-side rendering & compute.

Figma is really slow on older computers. If your laptop is (say) 4 years old, it's entirely likely that Figma is not usable on it and so something like Mighty might make sense. But then when your power supply goes out and you replace the old laptop with even the cheapest Mac, suddenly you don't need Mighty anymore.

You say "what really happened" four times with four, to me, unrelated conclusions.

Trying again, there's one point: What really made Figma successful is design moving from being made in "professional software" (like say Photoshop, Premiere, Visual Studio, Maya) to "collaboration software" (like Slack, Google Sheets), and that this was made possible because "flat design" is less technically demanding than the "skeumorphic" design it replaced.

(Sorry I wasn't clearer, I struggle on how to make points like this succinctly without is sounding a bit disjointed.)

I agree with one of your points (that where Figma really succeeded is that it understood better than anyone else the collaborative nature of web design), but I disagree that flat design vs. skeuomorphic makes much of a difference. I've seen more "realistic" designs where the actual iconography/images are created in something like Photoshop and then just imported to Figma, and Figma handles it fine.

Point being that I think that even if all apps still had the design aesthetic of 2008 iPhone apps that Figma still would have succeeded.

All the heavy lifting of figma happens on your local computer. The delta between doing that in a browser in a native app is only shrinking over time.

I think the delta you're talking about is already gone. Figma is generally considered more performant with its WASM rendering approach than its desktop counterparts.

But I don't think the let's say "other deltas" that are now keeping other high-powered apps from becoming web apps are moving at all. You can check my other comments in this thread for more details, but a specific point I'm making with the original comment is that there are other areas the Figma hasn't dented at all that prevent more powerful apps from being adapted to the browser. And that the fact that design no longer needed those features is what paved the way for its success.

Figma as part of an app family, is way way more similar to Google Slides (i.e., office suite) than it is to Photoshop (i.e., "professional" software).

> Figma as part of an app family, is way way more similar to Google Slides (i.e., office suite) than it is to Photoshop (i.e., "professional" software).

Digging into my own comment here, part of the reason I harp so much on the skeuomorphic to flat design change is it's such a bizarre thing to happen to an industry, for its defining requirements to change in such a way that the technology gets so much simpler.

E.g., can you imagine if realistic light rendering suddenly was no longer desirable in 3D software? Of course that would leave a gigantic opening for new players!

Multiplayer is such a game changer that all editing tools will eventually feature that, so the Figma model is going to win.

Once that is true, you want as many people to collaborate as possible and the browser is the perfect delivery model. Plus you ensure a consistent version across all clients.

Couldn’t agree more!

And liveblocks.io is here to make that happen at scale.

I don't disagree with that point as an idea. But I do disagree that that's the direction the market is going. E.g., we're here discussing a startup pivoting away from a bet on exactly the idea you're expressing here.

Today the onus is on people who still believe in that idea to express why they think Mighty failed (without falling into the well-worn "year of the desktop linux" trap).

Personally I don't care about the idea either way, I'm just looking at is where the market is going (e.g., how the market share of various software packages are trending). Ideas that sound good fail all the time in the market.

And so far, outside of Figma, the idea you're expressing looks to me like a failure? And personally, I've started to look at other things different about Figma that might have accounted for its success.

In the code space, would have to disagree I think, given how both JetBrains [1][2] and VS Code have embraced the client-server model, and how products like GitPod have used these to a full extent with excellent native integrations.

[1] https://www.jetbrains.com/remote-development/gateway/

[2] https://www.jetbrains.com/fleet/

Yeah I addressed this here https://news.ycombinator.com/item?id=33584880

I think the distinction is that the UI needs to run locally so that shortcut-rich powerful apps don't have to fight for the keybinding space with the browser.

I'd say that the migration to the client-server approach for professional software is already underway, with VS Code of course being the canonical example, but my (limited) understanding is that Blackmagic's Da Vince Resolve (which is currently eating Premiere and Media Composer's lunch) also uses a similar model where data can be stored remotely but the UI runs locally.

> This means what really powered Figma's takeover was that flat design took over, which is less technically demanding.

Absolutely 100%. The other thing is that flat design brought a lot of low quality designers into the market because you could just add a few recommended spacing grids and flat colours and get a design out. But because the quality often was low, it forced collaboration for cross checking and that caused Figma and other software to come to the forefront.

Why stop at creative cloud… go after autodesk. Hire guys writing shaders to build productivity software better in browser. Figma showed it can work.

Microsoft’s code editor killed everything in the market by using this client-server approach.

It is so seamless that you forget that the code is not in your machine.