Go's revolving door, live from ElevateDev

This show is supported by you. Stick around till the ad break to hear more about that. This is Cup o' Go for 06/10/2026. Keep up to date with the important happenings in the Go community in about twenty minutes per week. I'm Shay Nehmad.

And I'm Jonathan

Hall. And we're coming to you live physically in the same space for the first time

Ever.

Ever. And I'm going to prove it. I'm running away from my mic. And I'm speaking into Jonathan. And I'm back live from ElevateDev twenty twenty six. Yeah. What

is ElevateDev twenty It's 20

a conference we were invited to to speak because of the podcast, which is crazy. We really appreciate it. From a company called Genetech. We will have an interview with a Genetech employee coming to you soon, but let's start with some news. Alright. Should we keep old stuff around?

Yes. Done. What's next? So the proposal we're discussing is number seventy nine four fifty three. The proposal was to drop GCC Go. What do you think about that, Shay?

So I know GCC is like a compiler. Yep. And I know I like Go. Yep. But it's probably not the Go compiler, so I don't understand what it is even. I don't know. I don't think I've used it.

Probably not.

So what is it?

GCC Go. You may remember we I we probably talked very briefly about this on the episode with Ian Lance Taylor because he wrote GCC Go.

Yeah. Sure. I've been on the Go team since twenty o eight, actually, before the first public release. I mean, as we all know, Go was started by Rob, Ken, and Robert in twenty o seven as an internal project at Google.

So Ian Leslie Taylor wrote GCC Go, and I think its primary purpose was to be the other implementation of Go. And the value there is having two implementations of the same spec gives you sort of the ability to try things out different ways, see does this thing make sense in two different ways. And as this issue actually has talked about, there have been times when the spec and Go disagreed with each other. And so they would refer to GCC Go as sort of the tiebreaker to see which is correct. The problem is that GCC Go hasn't been updated for quite a while.

That's quite a while ago. We're on '26 coming on '27, so that's five years. Easy to count on.

Five years behind. Yeah. So the question here was, this hasn't been used for quite a while. I'm sorry. I made a mistake. I said it was the back end. It's the front end. It's the Go front end.

Well, it depends how you look at it. But

sure. If you're standing behind it, then it's the front end. If you're standing in front, it's the back whatever. The point is this is the Go support for GCC. It's well updated. The proposal was to drop support for it, sort of make it official that we're not doing anything with it. And there were a couple reasons. There were a couple things, not big things, but a couple things that are being maintained to support GCC Go. The question is In

the language, you mean.

In the language? Let me see here.

Alright. It's not it's not free to keep it around is what you're saying.

It's not free. Right. The conclusion was it's still useful to keep GCC Go around for the for the tiebreaker, if nothing else, a reason occasionally comes up. And also, Ian Lance Taylor himself says that there is a chance that it will catch up again, and he hopes he'll have time to work on it.

I hope so too, Ian. One of the things having GCC Go helps you with is the fact that you can compile Go to, like, more exotic architectures. Is there any chance I used it and I don't know, or will it have to be, like, old hardware, you know, whatever? I don't know if

it has to be old, but it's going to be less well supported hardware, or I shouldn't say less supported, but Less ubiquitous. Less common. Yeah. Less ubiquitous hardware. Because the GCC back end supports all sorts of architectures that the Go project itself doesn't want to maintain. So old versions of macOS might be an example.

Or maybe embedded stuff, or is that not the right

To some extent, although for a lot of embedded stuff, you're going want to use tiny Go.

Yeah. That's what I thought. Yep. And this is a sort of a weird proposal because it's negative. I would expect it to be proposal, keep GCC Go, and then it's like accepted.

Something like that? We're this will be episode one sixty one.

Okay. So episode three three three hundred twenty and two, something like that.

Very good.

Cool. I have a question for you.

Ask me.

Do you use AI coding agents?

I do. You actually have

a podcast coming up about that.

Right? Yes. I've recorded three episodes. I haven't released them yet. I need to get them edited. Filippo, I'll be reaching out to you soon.

But don't edit this part out.

So

which AI agents are you using?

I use Claude Code or not use OpenCode?

So I use Claude Code, Codex.

K.

I used another one for a little while. I don't remember which one. I think the cursor one. And today I'm starting out with AMP as well because someone recommended it to me.

AMP? AMP. AMP. Yeah.

We had Thorsten on the show. Remember? Mhmm. Thanks for having me. My name is Thorsten. I am software engineer, occasional writer. I've been riding Go since one point zero came out in 2011 or '12, something like that. I don't know. Yeah. It's by something he's working on, which is pretty cool. Anyway, how do you keep OpenCode and Claude, and in my case Codex and others, in sync with like AI agent specific configuration? Because Claude has ClaudeMD, but OpenCode may have like .opencode.

OpenCode supports ClaudeMD and agents. Md. I think it starts with agents and falls back to Claude if it doesn't exist. What's the question? How do I keep it in sync?

Yeah. Like skills, MCP servers that are installed and configured. And, you know, there is harness specific work, you know, like skills that doesn't look the same. Like Claude skills are at .claud and Codex skills are like, I think in .agents or .codex. And the config of like, even like what's the token backend, the API key that you use and stuff like that, in case you, like in my case, you actually get your tokens from Azure Foundry and not directly from the providers and stuff like that.

It does manually.

So that's a pain. And if you search for it, a lot of people are saying this. It kind of sucks. Like if you have Cloud Code, Cursor, Codex, oh, GitHub Copilot. Of course, I forgot about it.

Yeah.

But yeah, GAAL, supposedly you have one YAML file because that's everybody's favorite format.

And it has such good support in the standard library.

Don't even get me started. And it helps you, like, keep everything in sync. So you put one YAML file, and then you run this tool, and it'll make sure all the AI tools are in sync. What do you think about it?

It sounds like a good idea.

I think it's stupid. Okay. It's stupid that we need this tool. The tool exists because it solves a pain, but the pain is so stupid. Like these companies cannot agree on one name for the folder, maybe like .agent, and then just They have infinite tokens and they can develop infinite features, but they can't do this one because they're trying to stay like closed garden, walled garden, whatever.

Right.

That's fair. But like Skills, it's a bunch of empty files. Why can't we put them in the same folder? That's crazy. I think there is a good reason, like they're trying to do a walled garden. Like, they think people will be too lazy to move .claud to .agents. Are they really? You think people will stay on cloud just because the folder's there? So it's a stupid problem, but with a pretty nice solution. So check it out, g a a l. I'll try it, and I'll let you know if it's actually good.

Yeah. Let me know how it goes. Next up, a quick thing I just noticed dropped earlier today, new releases of Versus Code Go. Versus Code version 0.54 and the 0.55, which is identical, are released, 5.5 being the upcoming pre release and 5.4 being the stable release, add support for interactive refactorings. So certain refactorings will ask for user input before proceeding. A couple of other features, new debugger options in menu, and a package outline view.

Oh, nice. So

a couple of cool What

does interactive code actions like, I'm asking to implement an interface, it says? Or does that require something from me? Someone asked me about this today. Oh, if you have a class and you need to implement an interface in my session. At ElevateDev, which we're at. Anyway, why would you need a user input for that?

Okay. So the basic idea is that some refactorings require user input. It's not as simple as right click. Like, for example, right click, extract function is pretty straightforward. Like, there's only one thing to do.

guess when I refactor a thing, it moves it to a name like func one, and then I rename it. It's not like additional input. It felt to me like this feature existed.

Yeah. I can see that. Yep. Well, cool. So if you ever need to use Versus Code again instead of Claude or Codex.

My editor of choice now is Zed. It's so fast. Can't choose Versus Code. It's like Electron. It's too slow. I get like the typing delay. I like type a character, and then it doesn't show up in time. And I'm like, Crowd agrees with me, right? Yeah. No, no, don't want to shit on VSCO.

Yeah. So generic methods in Go, just as a reminder, we had generic functions for a long time, but no methods. Now we have generic methods, or I shouldn't say we now have them. They have been implemented. They will be released in 01/2027 coming in August, so we will have them soon. What's new? What's new is now Go sucks.

We saw yeah. Are other influential people in the Go world, except Jonathan and I. Specifically, there's a person on YouTube, Primogen. I hope I'm saying that incorrectly. Like a super, super famous influencer type of person. I watch his videos every now and then. Not exactly my cup of tea, but really cool. A lot of development type stuff. And he posted, I'm done with Go. And I was like, Oh my God, what did I miss?

So I had the same fear about generics, which was what, did we establish five years ago?

Yeah. Four and a half.

So my concern with generics was that this is everybody's going to start using generics and it's going be terrible. And I have seen a few libraries that use generics too much. In general, generics are useful, but they're easy to overuse if you want to. But I haven't seen that become a problem. So I think And

you see a lot of Go code bases. You're like a freelance it's not like this is based on one company. You've worked with many clients.

Yeah. Over the last five years, I've seen a large number of code bases, open source and closed source alike. And, you know, I haven't seen my fears come true at all. So I don't think these fears are founded. If they are, if they are well founded, we'll find out soon.

Yeah.

But I don't think that this is going to be the I don't think the predictions are going to come true.

I agree. I also think that as a like, if I'm just a developer, like a guy, I don't have a following, I don't have a show, I don't have a whatever, I can make these choices based on these predictions. But here it's like almost insider information on Polymarket, you know, the guy who ran into the NBA field, like a court in the finals, and then turns out he bet on himself in Polymarket that someone is going to break into Like the when someone with as influential as Primogen that has like hundreds of thousands of views per episode says, per video, I mean, has a statement that Go is bad because of X. Even if I can say that X is, I don't think is true and has been demonstrably false about generics in the past, it sort of creates this effect that it might become true just because someone famous enough said it. In your country, I think it's called the Trump effect, right?

Yeah. And I think there are many other examples of things we can do in Go that we don't, not just generics, not just this.

Yeah. People try to keep it pretty simple, I guess. That's why you picked this language. He didn't recommend towards the end a language I haven't even heard of, Odin. I think I heard of it in the context of game development, just like a few freelance game developers I follow.

No, I'm hungry and I hear the crowd outside eating food. Want to go, let's finish this up.

right. Let's go do the interview then. All right. We're here in the Genentech conference, ElevateDev twenty six. I guess we'll just grab someone random.

Hey, who are you? I'm somebody random.

Hey, somebody random. Was great to have you on the show.

Thank you. Happy to be here. Hi, Louie. I want you to introduce yourself to our listeners.

Yeah. My name is Luis Ciorone. I'm working in the R and D department here at Genentech. I've been an employee at Genentech for about ten years. And starting maybe back in 2024, I was given a team with the explicit objective of exploring opportunities in developing our code bases and Go going forward.

Cool. Cool. Cool. What is Genentech? We said this name like a million times, but if you can tell the listeners who maybe aren't into physical security.

Sure.

Genentech is I assume actually our crowd, our listeners being software developers, that many of them are actually into physical insecurity. I mean,

I'm sure I'll butcher this definition for the my coworkers and colleagues that are going listen, but I'll do my best. Genitech is a Montreal based company that handles physical security for kind of making sure that our customers have a one stop shop for the variety of security solutions. So we're talking access control, video surveillance, intrusion, license plate recognition, and etcetera.

Yeah. We saw some demos at the conference this morning with big, like, companies like Lowe's and whatever. It was really, really cool.

Yeah. I attended that one as well. I missed an earlier one with Home Depot that I heard a lot of great things about. So I'll probably try and check a recording of that if it's available.

Cool. The videos they shared were really some of them were really funny. Just guys showing up with a white box truck huge white box truck, trying to load some equipment from the store and then your system's detecting it. And then like local PD swooping in and they're like, Oh, I can't believe it. I came to a store that has cameras and they caught me. How is that possible?

Yeah. It looks straight out of a Bad Boys two movie. It's great. I mean, it's funny because we had other talks throughout the weekend, one of which was our product life cycle history. And part of that is to make those fake videos. So we have kind of actors that do like fake shootouts and fake stealing. If you, we were told if you see someone walking around the hallways kind of just leaving a knapsack, don't grab it. They're probably filming, something for PR reasons.

So now I know how to steal something from Genentech.

Yeah.

I just need to look like an actor.

Just just keep your lapel mic just keep your lapel mic on. Like, No, no, I'm filming, guys.

Yeah. It's like the walk, you know, as long as you're carrying a ladder, anybody will let you in.

Exactly. Is actually, y'all inviting us to the conference has been great already. Super good conference. I honestly can say I recommend working at Genentech. We'll get to that later, but it looks like an awesome place.

Yeah. So it's funny you say that a person's programming language kind of comes off as an identity, and being told to change that is is a panic because that's that's exactly how, my colleague and I felt when we were given this team. You know, we were kind of our our our reputation is built on on our ability to write .net c sharp, and then we were told, nope. You're doing Golang now. Good luck. And, we kinda panicked internally, And it was I was told

You're getting it already because that's how you do errors in it. You panicked. You don't you didn't throw an exception. Yeah.

We remained, so panic was fine. It was all you know, it was before the Christmas bakes. I I at the Christmas break rather, so I immediately bought, you know, a forty hour Pluralsight course. It was like, learn, go. And and of those forty hours, maybe two of it was relevant because I I realized that thirty eight was just kind of fluff that was common to any programming language and and to that two hour chunk was what was specific.

So that you're saying this is an attractive quality of Go. Like, you can get pretty proficient at it pretty quick. So if you're trying to, like migrate a company of one tech stack onto Go, rather, I should say, that's that's been working for you, saying the ramp up is pretty easy.

Yeah. Mean, helps kind of taper the fear that that comes. Because there's a there's a natural resistance to change that everybody has, and this kinda helped temper the fears that people were having. They're like, just spend a week. Look at it.

all project, component templates and project templates, you mean.

Exactly, yeah.

Yeah. You're not kidding about the branding, by the way. Jonathan and I, we just shared a photo of us from the conference floor. It's like a huge ass gopher statue. It's like as big as me.

I'm I'm really wondering what they're what they're gonna do with it. My my director No.

I'm taking it. I hope Eric Canada is gonna let me take it as a carry on.

Alright. Go for it. Yeah. Our director said the same thing about some of the stuff hanging in in the atrium. That where I

was that's

where I was saying. But oh, yeah. So the the the branding here. So it it it was kind of amazing for my team and I to see all this branding because there we were not the only team that was doing Go development. We were just the team that was told explicitly, do Go, tell us if it's good or bad, spread the words, convince us. And our our CEO was very invested too because he

So I'm I'm curious why you were tasked with learning Go. I guess two questions related. Why the interest at the corporate level or whatever to start doing Go and why did you get looped into this? Was it your choice or was it a mandate or voluntold? Voluntold is great.

Right. So I'll start with the easier of the two to answer and move on to the more difficult Why me specifically? It was really just the right place at the right time. We were in a transitionary period going through a minor restructuring in our department, and I was tasked with starting a new team with the objective to rewrite some of our kind of on prem design features to be cloud native.

Okay.

So they saw that as a, oh, they're writing a bunch of stuff from scratch, and it's gonna be on a cloud distributed platform that might be a good opportunity for Go. Start there. That's about how that went. Not because I had expressed previous, you know, knowledge of Go or anything like that. Mhmm.

That rate sounds like you're hosting the show next week, and you're you're designing the Go committee a week after that, and you're officially the compiler next month. So be ready.

I've been I've been the front man. I've been I've been the guy doing the talks and and doing the self deprecating jokes. To to get back to your question, the second question being why Go why are we exploring Go? It's it's definitely a multilayered question. I'll answer it to the best of my abilities.

I just wanna clarify because a lot I think a lot of our listeners are not, like, physical. They're like just software developers. Appliance, mean a physical thing in

means the physical box.

Yes. But we're talking about an edge device thing.

I'll share for Tom. This episode is only for Tom. If you're not Tom Vanderberg, could turn it off. But what you just said that happens is like, Oh, wait, Go, I brought it for this thing, but it's actually useful for this. If it hasn't happened already, what's going to happen is that internal CLIs are going to start be written in Go, because it's so easy to install them, like brew install from your local brew tap or wherever.

I mean, the LLM models, I mean, Go is pretty conducive to that. There's very you know, Go Go is a language where, I guess you guys repeat this under the podcast, but I guess I'll say it for for the record, you know, there's only really one way to do something in Go and using AI to kind of just five code stuff in Go is is a lot simpler than in other languages sometimes.

Yeah. We talked about what's his name? The terraform guy. Yoshimoto? Yeah, Mitchel Hashimoto. Jonathan's been looking at me. When we do the show not live, you don't have that Your eye roll doesn't have that much power. He said that he just got back into Go from Zig and he was like, Oh my God, Go is so fun with LLMs. They have GoPlizz and GoDoc on the command line and such like superpowers. It's really great.

Yeah. Short answer, all of the above. Difficulties on a personal programmer level was not the syntax, which is what we thought it was going to be. It wasn't, you know, lower spec devices. But the Not to say we ever shipped broken software because that never happens, but we had to quickly find some optimizations.

Hey, future Shay here. Sorry for the next part. We had some technical difficulties with the audio. So you're gonna hear Luis answer just out of the background noise of, my microphone. We tried our best to make it sound good, but that's how we got it. So sorry about that. Please continue to enjoy the rest of the interview. So thanks a lot for coming on the show. This is a good chance for you to like plug your your Twitter or your LinkedIn or Genentech itself. Are Genentech hiring?

Yeah. Genentech is always hiring. Were one of those companies who worked for AI. Was kind of cool. Maybe a lot of companies were hiring as fewer hired more. So if you're interested, we got multiple offices, mainly our head office in the Montreal office. But if you are curious, go to Genentech dot com. That's ateanttc, because there are kind of a lot of variations of spelling around those areas. You have to be careful, but I'm sure it'll be

Yeah. Link in the show notes. And honestly, being at the conference, I recommend it. I've never worked for Genentech, but this seems like a cool ass place to work, man, especially as a software developer. Like, this conference is one of the best conferences I've been to, not corporate conferences.

Yeah. I can see why that's a that's a stumper.

I don't know if that's

a good answer, but it's mostly setting up a build packages with other people. I noticed it's gonna be a controversial answer because I heard about Donovan talk regarding testing, though. So

we're

going testify. We saw

a deal

for a

lot of the factor standard package right now. But this is only because the journey that might be booked where ads has a data imposed a testing framework that was spinning around the India gate code omega. And the from some of the other teams that were doing go into a big pushback because they said it looked like spectral and not doing that. Which yeah. Fair enough.

That's one way to get Jonathan on your side. Well, Louis, thanks a lot for coming on the show and thanks all Genentech for hosting us. That wraps it up for this week. We're sorry if this episode was a little bit all over the place. It is live. I guess that's program exited. Goodbye.

Program exited. Goodbye.