Seems very limited for an IDE. Only certain web frameworks permitted for now, python and go support is still not there, and Flutter is the only choice for cross-platform apps. Android and iOS simulators aren't built in yet either.
Given the blog title and the blog itself, I was expecting a bit more and something less piecemeal rather than what is actually presented here which is an online IDE mixed with google AI.
Alternatively, they make IDEs and so does everyone else. They even have a web IDE powered by VSCode as an offering in GCP, like GitHub codespace.
This seems tailored to being good at doing front end stuff (multiple emulators, multi-browser previews, etc). Which is probably a problem google deals with themselves.
GCP's on-boarding process, especially setting up local entertainment, used to be a little complex¹. Maybe having the whole IDE in the cloud would simplify it. It would be cool to have ”Bard” create a role based on my code or refresh an expired token.
1. A bunch of really smart folks asked me for help with it on a few different projects. I tried GCP's VSCode extension but it wasn't particularly easy to use.
This is Google at its 1990s COMDEX/Microsoft Vapourware worst. There is no product to try, no code samples, no documentation. Just an animated GIF of what the experience may be like and a signup link.
This announcement feels like a "well, we're not close to shipping and we don't know if we have the resources to continue, so let's issue a press release and see if we can muster some interest to gauge if we should keep funding this". Or else it's just an announcement to help show that Google is invested in AI.
It is weird they would announce it without any real screenshots or demos. This is usable by Googlers. So it does exist, but I'll leave it to the team to hopefully actually demo it at some point.
EDIT: to add, the blog post does mention this: it is built on-top of VSCode, so it's a UI people are used to.
Maybe at Google everything is an endless sea of complexity. It sure isn't for me, I generally can have a dev environment up & going in a few minutes locally.
The "full stack" Project IDX is suggesting certainly has an endless sea of complexity: Angular, Flutter, Firebase, and Google Cloud, all together in one "full stack application"? Yeah, that is a ton of Google-created complexity to smash together.
Google: We've made our tech stack so complex you need an AI to get started working with it.
This has "promo packet driven development" written all over it. Zero chance something as random as this is going to get continued support from the company.
So yeah, regard it as an interesting plaything; certainly not something that you should get too emotionally or intellectually invested in. Especially given Google's penchant for killing cool things.
Yes! We have exciting news to share! You get $o.99 off an annual YouTube premium subscription. And we use your code to raise the craft of developers worldwide.
It's very strange how lately new Google product launches always seem to be gated behind waitlists or slow incremental rollouts. Back in the Gmail days that happened for genuine reasons like very real machine capacity constraints, the explosive popularity of the product and so on. But these days Google has so much machine capacity they sell it to other people, and an experimental IDE isn't the sort of thing that's going to experience explosive consumer-scale growth. Presumably the waitlist is for product management reasons but it's an odd choice. Usually products get a spike of interest when first launched which then rapidly evaporates (see Threads, but also e.g. Chrome and Wave had the same issue), and so it's not a good idea to waste that first surge of interest.
I would love if this IDE had first class support for Bazel projects. IDE support is one of Bazel's weaknesses. I feel an IDE focused on polylingual monorepos and their build tooling (Bazel) would do well, as no single IDE works well for that use case. Seems like it would be possible to whip something up using the facilities provided by the language server protocol and Google's Bazel expertise.
It sounds like whatever project you build will be in a GitHub repo, so maybe you could enjoy trying it out while it lasts and switch to some other development environment if needed?
> While Google’s been working on making multiplatform app development easier for years – from Angular and Flutter to Google Cloud and Firebase
Yea, and those products are not doing well. It would be nice to see Google focusing on the ship that's already sinking rather than trying to play with a siny new toy.
> paired with the universal access that comes with being hosted in the cloud, in a datacenter near you.
How is a locked down vm, that I don’t control, “near me” more convenient than the fast, portable computer in front of me that I can build apps anywhere on, even when not networked?
You don’t need a fast computer near you? Plenty of people like this model of development.
I personally love being able to compile massive projects while sitting in front of a MacBook Air or iPad. It sips battery while churning away at a data center elsewhere.
> You don’t need a fast computer near you? Plenty of people like this model of development.
Not really, honestly, and I've tried it in the past. Maybe it's a lifestyle thing, I travel around a fair bit on the train and value being able to do everything from my laptop. With an M2 MacBook Pro I can also churn through big Swift/Rust/JS projects and with a local LLM installed and some offline docs it's pretty productive.
If the Android and iOS simulators are lightweight compared with their native counterparts I'm all in. Android Studio's adb stopped working for me recently and the Android Studio emulator is just too heavy for my 2013 Macbook Pro which works just fine for everything else, eg. JS, Python, Clojure and Golang.
As the CEO and cofounder of Coherence (withcoherence.com) - it's cool to see this project launch. At Coherence, we've been working fully in the cloud for the last 2 years - and it's better than I thought it would be. But it's hard to get more conservative teams to jump into a new workflow. While we've done a lot to support integration with local tools and development, I'm looking forward to seeing more innovation, and am curious to see Google invest here more directly. For us, all of this innovation is about delivering the best developer experience for end users that we can, let's see what direct IDX goes as well.
Other products to check out: codesphere, stackblitz, Replit
The ignorance of many comments in this thread astound me.
Sure, setting up a development environment for your small pet project is easy. But as soon as you start working on software that has been maintained for a few years, dependencies start to grow. A typical old Java project depends on Gradle (or Maven, or even Ant), an outdated JDK, Eclipse, Protobuf tooling, XML tooling, custom tooling, you name it. Next, your project may require linting, formatting, and will be checked by some third party services. Recently, Docker and friends have joined the party, and a web based frontend requires TypeScript, a framework, Webpack, and many other libraries and tools.
Joining a team with such a project is not an easy task. In a typical 500,000 line code project, it may take a whole day or more to get the basic system up and running. All the dependencies may break in a myriad of ways, and can result in a lot of unnecessary problem hunting.
Sure, there are tools to automate setting up the build environment. But even those require maintenance, and setting things up for quick onboarding requires a lot of time, and if done wrong, will take even more time to get right later.
It would save a lot of money and effort if I could have a reproducible development environment, and I would not mind if that is in the cloud or not.
> A typical old Java project depends on Gradle (or Maven, or even Ant), an outdated JDK, Eclipse, Protobuf tooling, XML tooling, custom tooling, you name it. Next, your project may require linting, formatting, and will be checked by some third party services. Recently, Docker and friends have joined the party, and a web based frontend requires TypeScript, a framework, Webpack, and many other libraries and tools.
You say that as if it were a God-given truth that that is how things must be, or maybe even as if it were a good thing that that is how things are. Which could hardly be further from the truth. The solution isn't to cover all that shit up with yet another layer of cruft, but to get rid of most of this fricking cruft.
This sounds like it could become a nice way to try out Flutter since you have to install Android Studio and/or XCode to get started on mobile development. Android emulators seem to require messing with bios settings for unclear reasons. It gets obscure pretty quickly. Flutter depends on a lot of baroque complexity that I guess mobile app developers just accept and it’s not entirely covered up.
Other development environments might not benefit so much:
For Go, the SDK is easy to install and VS Code works.
I’m out of date on Python and assumed package management would be kind of a pain to deal with, but I tried miniconda recently and it it’s pretty nice.
Web development environments continue to evolve, but there are some easy to use ones. I like Deno so far.
It would be nice if we didn't need this. It would be nice if we had a native(!) low-level GTK-like API on MacOS, Linux, iOS, and Android. Windows has it but that's because Windows is ancient and backward compatibility is important, and no one has figured out how to ruin it without breaking backward compatibility yet.
69 comments
[ 0.18 ms ] story [ 141 ms ] threadGiven the blog title and the blog itself, I was expecting a bit more and something less piecemeal rather than what is actually presented here which is an online IDE mixed with google AI.
This seems tailored to being good at doing front end stuff (multiple emulators, multi-browser previews, etc). Which is probably a problem google deals with themselves.
1. A bunch of really smart folks asked me for help with it on a few different projects. I tried GCP's VSCode extension but it wasn't particularly easy to use.
Shut down due to low adoption < - > Low adoption due to fear of it being shutdown
EDIT: to add, the blog post does mention this: it is built on-top of VSCode, so it's a UI people are used to.
(googler, opinions are my own).
Privacy, speed, lock-in, etc.
Especially with google's track record.
Google: We've made our tech stack so complex you need an AI to get started working with it.
So yeah, regard it as an interesting plaything; certainly not something that you should get too emotionally or intellectually invested in. Especially given Google's penchant for killing cool things.
One could also make the opposite argument that having beta testers is a way to build hype. It all depends on the situation.
Yea, and those products are not doing well. It would be nice to see Google focusing on the ship that's already sinking rather than trying to play with a siny new toy.
How is a locked down vm, that I don’t control, “near me” more convenient than the fast, portable computer in front of me that I can build apps anywhere on, even when not networked?
I personally love being able to compile massive projects while sitting in front of a MacBook Air or iPad. It sips battery while churning away at a data center elsewhere.
Not really, honestly, and I've tried it in the past. Maybe it's a lifestyle thing, I travel around a fair bit on the train and value being able to do everything from my laptop. With an M2 MacBook Pro I can also churn through big Swift/Rust/JS projects and with a local LLM installed and some offline docs it's pretty productive.
Wasn't his point that he already has a quite-fast-enough computer much nearer; like, right in front of his nose? As do I; and, I wager, you.
Other products to check out: codesphere, stackblitz, Replit
Sure, setting up a development environment for your small pet project is easy. But as soon as you start working on software that has been maintained for a few years, dependencies start to grow. A typical old Java project depends on Gradle (or Maven, or even Ant), an outdated JDK, Eclipse, Protobuf tooling, XML tooling, custom tooling, you name it. Next, your project may require linting, formatting, and will be checked by some third party services. Recently, Docker and friends have joined the party, and a web based frontend requires TypeScript, a framework, Webpack, and many other libraries and tools.
Joining a team with such a project is not an easy task. In a typical 500,000 line code project, it may take a whole day or more to get the basic system up and running. All the dependencies may break in a myriad of ways, and can result in a lot of unnecessary problem hunting.
Sure, there are tools to automate setting up the build environment. But even those require maintenance, and setting things up for quick onboarding requires a lot of time, and if done wrong, will take even more time to get right later.
It would save a lot of money and effort if I could have a reproducible development environment, and I would not mind if that is in the cloud or not.
You say that as if it were a God-given truth that that is how things must be, or maybe even as if it were a good thing that that is how things are. Which could hardly be further from the truth. The solution isn't to cover all that shit up with yet another layer of cruft, but to get rid of most of this fricking cruft.
Other development environments might not benefit so much:
For Go, the SDK is easy to install and VS Code works.
I’m out of date on Python and assumed package management would be kind of a pain to deal with, but I tried miniconda recently and it it’s pretty nice.
Web development environments continue to evolve, but there are some easy to use ones. I like Deno so far.
Eh... GTK? (Or Qt?)