🪖 Gab's Notes on this week's Go news and an interview with Matt Boyle of Ona and ByteSizeGo

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Gabriel Augendre (00:00):
This show is supported by you.

Shay Neymar (00:07):
This

is CopaGo for September 19 2025. Keep up to
date with the importanthappenings in the Go community
in about fifteen minutes perweek. I'm Gabriel Lojean.

And I'm Shay Nehmad Hello.

Hello.

John, you sound extra fancy today. Oh, no. Wait.
Gabriel, how about you introduceyourself to the people?

Thank you for having me. I'm Gabriel. I'm a
technical lead at OVH Cloud, andI work with Go day to day with
Go Codebases. Nice. So

Welcome to the show. Thank you so much. Thank
you for covering for John, who'slike in in South America today
or some, having a drinkingcoconuts on the beach and
wondering what's up with Go thisweek. So let us help him and
tell him what's going on with Gothis week. How about you start?
You mentioned the JSON blog. Wementioned it in the past and now

(01:10):
you finally dove into it.

Yeah. The Go team published a blog post about
the JSON v two that they haveadded to Go 1.25.

So it's not by default, right? Sounds like I I
I use it by default if I haveone twenty five.

That's correct. It's a v two package
and it's not visible by default.You need to enable a Go
Experiment flag to have itavailable in your code base.

So GoExperiment is something I do at runtime, at
build time. Is it it's like anenvironment variable, right?

It's an environment variable. I believe
you need to set it at buildtime.

Yeah. It seems like it used to be with the loop bar
thing, so I assume it's thesame. So what's the best place
to learn what's up with this newV2 package?

Go. Dev, the blog from the the Go team, they
just published an an article,like, ten days ago.

I will respectfully disagree. I think, as usual, the
best blog post to learn aboutthings in Go is actually Anton's
Anton Givano's blog. Anton, wehad him on the show, of course.

My name is Anton. I do some open source
stuff, and I write interactive,maybe I can call them guides or
books, and interactive articleson my blog. That's mostly what I
do in my free time. So that'sit.

So, yeah, Anton has all these runnable examples you
can run directly in the in theUI, which shows, like, the
difference between v one and vtwo.

Yeah. That's correct. And the the Go team
even linked to Anton's blog postthis time. So I believe they he
made made it quite popular.

He made it. Mom, look at me. I'm on the Go blog.
You know, I wonder if they linkto Cup of Go at any point in the
official Go documentation. We'llwe'll we'll need to check.
So what are what is like yourfavorite difference from the new
version?

I believe the the case insensitive
unmarshalling, which is well,it's case insensitive in v one,
and v two is case sensitive. Itbit me a few times because you
could in v one, you could havetwo different keys in the same

(03:40):
JSON with different case, whichis valid JSON, but the behavior
can be surprising in the v one.

Yeah. It was definitely not what I expected
it to to run. I love the new I Imean, there there are a few
things that are nice. I kind oflike it might be a bit stupid,
but I really like that you haveMarshall Encode and Marshall,
Marshall Wright and MarshallReed. So you don't have to,
like, create a new encoder andblah blah blah.

(04:11):
You can just do, like,json.marshallwright. Like, it's
an operation I do so often in,small scripts and whatever. It's
very nice to extremely quicklybe able to read and write JSON
without, like Yeah. A a wholelot of ceremony. You know what I
mean?

I agree. Yeah.

Even though it's probably the least substantial
difference in terms of syntacticsugar, it's, like, the best.
Would you recommend people tochange to the v two To, like,
turn on this experimental flag?Obviously, it'll be helpful for
the Go team as, like, feedback.Right?

I haven't tried it myself. So I don't know
if I'm ready to encourage peopleto try it. But, well, people, I
believe, should try it for thebenefit of the future proofing
their code maybe. And in theend, the v one will be
reimplemented in terms of v two.So all v one functions will call

(05:08):
the v two just with a differentset of options.
So the they seem quite confidentin the implementation they did.

Cool. Well, go check out the dev blog. It's in
the show notes. The link isgonna be in the show notes. And
obviously, we'll, directly linkto Anton's, blog as well because
I honestly I I legitimatelythink it's better.
Anton with another, really good,thing going on. I wanted to
mention a task you have to donext week. I know we just met,

(05:41):
Gabriella, and you've beenlistening, but I'm already
giving you tasks.

You're already giving me work?

No. Homework, homework. I want the survey. I
want the survey completed on mydesk by eight 8AM.

On Monday?

Yeah.

Dang. But 8AM in in The US, right? Oh,

so gives you you have the whole day. Yeah. So
yeah, the new GO survey is outfor 2025. It should take you
about ten, twenty minutes andwe've seen in the past how how
much the survey informs the GOteam on like what to do and what
to focus on. Obviously they havetheir own agenda and it's driven

(06:20):
by Google and one of the veryvisible things we see is like
they talk about AI a lot andthey release Genkit, we know
we'll talk about it maybe nextweek's episode, we're just not,
we haven't tried it out yet.
But they are still, like Go isstill a very much a community
driven language and I think thisis one of the, other than
actually contributing issues andpull requests and things like

(06:42):
that, I think this is one of themost direct ways to impact way
the Yeah. Language

And maybe one of the most easiest as well.

Yeah. It just takes like ten minutes, fill out a
survey. Not that not thatdifficult. I actually recommend,
if you're working Gabriel, youmentioned you work at the OVH
Cloud as a tech lead, right?Yep.
So you you and you work withGoCode at the OVH Cloud? Yep. So
my recommendation, I don't knowif you've ever done something
like this, but the next, like,you know, like engineering sync

(07:12):
or design meeting or whatever,do, like, the last ten minutes,
fill out the survey together.That that's usually a lot of
fun. One of the most funquestions to answer in a group
is, how long have you been usingGo?
For me this year, this was likea whole thing. But wait, how
long have you been using Go?

I've been using Go for maybe two to three
years.

Oh, so you're gonna be in the initial. Yeah. They
don't have a bracket for meanymore, bro. I'm in the 10.
Wow.
That's it. I'm I'm officiallyold as fuck. That's the oh,
Filipov blipped that out. Sorry.I'm officially older there.
Yeah. I I had to answer, for forthe first time that I've been
using it for more than tenyears. Although to be fair,

(07:57):
like, I don't know how you countit exactly because I've done Go
ever since I've I've starteddoing Go. I haven't stopped
doing Go, but I haven't done Gofull time a 100 of the time. So
I used to do Go, when I startedwith it, like 2014 and then I
worked at a company calledGardicore and a thing called
Infection Monkey that was likePython for a year and a half or

(08:20):
something.
So I did go in the backgroundbecause I liked it, but it
wasn't like my full time job. SoI think honest if I'm speaking
like, you know, net, how much,time I've been doing Go, it's
probably like six years of likefull time Go, not ten. And it
was just like four years in themiddle where I did other
languages. So I don't know howyou I don't know how you count

(08:41):
it.

I'd say you can count. The the question is,
how long have you been using Go?It's not how long have you been
using what? Using it 100% of thetime or professionally

Yeah. Or I always I always had something going on,
whether it's a podcast or a sideproject or whatever. But, the
the survey this year, other thanthis question, which is fun,
there's a lot of questions aboutAI obviously. Are you using Go
with AI? Which curse like whicheditor are you using?
Blah blah blah. And I think oneof the interesting questions
they leave open is what is thebiggest challenge you personally

(09:15):
face using Go today? To me, thiswas like an interesting answer
because I'm at a startup andwe're not using Go at the
moment, and it's like figuringout when the right moment is to
transition from like shitty, youknow- TypeScript. Transpilot.
Yeah.
Whatever to to Go. But I assume,like, it's such a big question.

(09:37):
People who are actually usingGo, maybe it's like, oh, the
libraries are not good enough.People who who are like me,
like, want to use Go more. It'slike, oh, getting people to
actually believe that Go is morethan just a server language.
I don't know. I I I'd be reallyinterested to see the responses
for this one. I believe it's Sodo you have it on your calendar
when to do the survey?

I marked it Monday, eight to ten, so I can
send it to you.

In all seriousness, we really encourage our
listeners to do it. The the linkis in the show notes, should
take about ten minutes. And,yeah, go go let your opinion be
known.

So, Shai, I believe you you had Red One on
the show in the previousepisode?

Yeah. That is true.

Redwan (10:25):
I'm Redwan . I'm currently working at Vault,
which is kind of like a Europeansister concern of DoorDash in
The US. So, yeah, I've beendoing open source for a long
time and working here foraround, like, two years. And I
primarily work with Go.

So, yeah, we had him on the show. He writes
really good blog posts.

Yeah. And he he published another one earlier
this week about tests. You No.Use your code,

I just I use a a programming thing called bugless
oriented programming, and thatjust saves on all the No, I'm
just kidding. Obviously, I usetests. The main tests I'm using
right now are, end to end testswhere I set up the server and
the database and I actually run,API requests because I found

(11:18):
that at least in the setup I'mworking with right now, mocking
is kind of weak. Like I haveunit tests for just like pure
functions and then most of mytesting is end to end.

And what frustrates you about your end to
end tests? Do you have

some What frustrates me is that they're
slow because they need to set upa Postgres. And if I could make
them work with mocks, I would. Ijust didn't have the time yet.

Okay. So Red One advises not to use mocks
because they could lead you theycould lead your tests to be less
readable and less maintainable.And you may end up, if you're
not careful with especially withlibraries that generates mocks

(12:05):
based on an interface, you mayend up with tests that verify
that a function has been calledonce, twice with a given set of
arguments. And RedOne arguesthat this is not the best way to
test your code because you'll bemostly testing your mock, and it
will be fragile. It will it maysurvive real bugs, which is

(12:29):
something that you do not wantin your tests.
And so he advises instead toeither write your own fake
implementation of a database, oras you do, do the real test with
a real database if you if youcan if you can spare the time.

The the this blog post is not specifically about
Go, but the the examples in theblog post are in Go. I think in
Go, this problem is a bit moreobvious because, you know, when
you use mocking tools likemockery and things like that,
you have to generate the mockimplementation. So you have to

(13:07):
go through a build step. Whereasif you use like py like Python
mocking, like with monkeypatching or or just mocking in
TypeScript, this happens atruntime. So you don't have to do
anything and and it'll sort ofstill work, which is even more
insidious.
Like, it's even more fragile.

Python, especially, I believe, if I
remember my Python days isespecially, in my opinion, nasty
because it allows you to monkeypatch something directly without
it being intended to be patched.

Yeah. It doesn't have to be part of any
interface. You can just it's alltext. You can just rewrite text
in these text files. And therewas a discussion with, Josh
Bleecker Snyder, friend of theshow, and Red One in our Slack
channel.
The the this is like a aninteresting blog post, you know,
and an interesting discussion inisolation, but it is not

(14:01):
happening in isolation. The theinstigator for this conversation
is, Red One's frustration with azillion lines of AI generated
code, that's, that does nothing,which seems like he's been
dealing with. Not a good time tomention, like, where he works
probably. But have you been,like, feeling this, frustration?

(14:27):
Is this something that you'veseen, like, people giving you
pull requests with AI code thatdoes nothing, tests that are
irrelevant, things like that?

Don't work on very popular open source
packages or open source codebases. So I haven't had the
chance yet in public work. Andin our company, we well, in the
in the company, we are workingas a team, and AI as a coding

(14:57):
companion, let's say, hasn'treally been popular yet. So we
haven't had these pull requestsfor now.

I've seen

Have you had to deal with them?

No. No. We're the tiny startup, so everybody's,
like, super accountable. I'vehad to deal with, like, you
know, I'm running, like, cursorbackground agent or cloud code.
I've been trying to experimentwith all these things.
I've been I've I've had them,like, run. What what I'll do is
I ride to in a very Frenchmanner and not not in a very non
American way, I like ride mybike to work. And what I'll do

(15:30):
is I'll try to fire off likethree or four background agents
on my tasks for the day, put mykid in kindergarten, get on the
bike, get to work, open mylaptop and see what happened. So
far it hasn't like been givingvery good results, but if I like
paid less attention and justopened these pull requests and
send them to, you know,coworkers, they might have, they

(15:51):
might have been like, shy, whatthe fuck? Like, this is this
doesn't really work.
I've I have seen in a verysimilar vein, there's a bug
bounty program for curl, likemany other tools and curl is
like super important so theytake Yep. Security reports very
seriously. I've seen somethingon LinkedIn today, I'll see if I
can pull it up, but if not, I'lljust I'll, like, explain it. We

(16:15):
see an obviously AI generatedsecurity report that's, like,
super detailed, lots of code,whatever, that doesn't actually
call curl at any point. It likeimports the c code and then
generates some vulnerable codewith curl structs, but the
vulnerability doesn't actuallylike, it doesn't call any
function in curl at any point.

(16:36):
Yeah. And then the maintainerthe maintainer has to read
through this, like, hugesecurity report, waste a ton of
their time. This is, like, superexpensive people time, you know
what I mean? It's like themaintainer of cURL. Dude is
literally holding the Interneton their shoulders.
Then the he responds like,you're not even calling curl

(16:57):
code. And then they respond,you're absolutely right. And
it's like Oh.

Yeah. I hate that.

I feel like what Red One is pointing at is like,
this is a mistake a person coulddo. But the fact that LLMs make
this code more easy to write,easier to write, I should say,
doesn't take away from thepersonal responsibility someone
should have. Like, you're stillaccountable for the code your AI

(17:23):
tool generates.

I believe so. Yeah.

I think that's You the actual

should be you should be accountable.

Like, oh, it's just Claude. No, man, it's you, like,
you need to review it. I think alot of people are learning it
right now because they haven'tbeen team leaders yet. But when
you're a team leader and yourlike, one of your team members
does something wrong, you're notlike, oh, it's it's that guy's
fault. It's you're the teamleader, it's your fault.
So people are living with thiswith AI agents now. But still,

(17:49):
technically, LLMs aside,technically, this is a good blog
post. Like, I would introducethis to any coding standard,
like use mocks sparingly, knowwhen to use a mock, know when to
use a fake, know when to use thereal implementation, right? It's
a good balance.

Yeah, the it's the what's introduced the post,
but it can be applied to

anything. Cool.

So, Shai, we've been talking for about
twenty minutes now. What, whatabout a quick break?

That sounds great. As Gabriel mentioned at the top
of the show, this show issupported by you. The easiest
way to support us and the mostdirect way is just give us money
through Patreon. Thanks a lotfor Sarangan Thuraiman. No way

(18:37):
I'm saying that correctly.
Joined us on Patreon this week.Yeah, this is a hobby, but we do
it and we do it for fun, but itcosts us a bit of money, you
know, time at work and, hostingfees and editing costs and
things like that. So Patreon isthe easiest way to, make this
hobby a little less expensivefor us. You can find the link to

(18:58):
Patreon and all the rest of thethings like, our Slack channel
where a lot of the discussionswe had this week actually came
from, and our shop and our emailand past episodes with
transcripts. You can find all ofthat at cupogo.dev.
That is cupogo.dev. Youmentioned you checked out the
show, Gabriel.

Yeah. I checked out the the show, and I
checked out the the Cupogostore. I see the, a sticker that
I really like, and I may needone for my laptop, which is
quite crowded already.

It's not that big. It's like a it's a good size. I
have one. I have it. You youwon't be able to see it like
listeners right now, but Gabrielis able to see my wallet.
So it fits on like a wallet.
We're not gonna have room.

No. But I I put them on top of the others,
so the the next one replaces theprevious.

I legitimately worry that your laptop is
heavier than it should becauseof all these stickers. But yeah,
we actually got quite a few,swag, orders this week. It was a
refreshing thing in my email.It's like order, receive, order,
receive. That was cool.
I like it when people get theshow swag. If you have any swag
requests or ideas, drop them inthe Slack channel. It's like

(20:17):
with a reprinter, so we can dowhatever idea you have. Other
than getting swag or joining ourPatreon, the best way you can
support the show is to leave areview on Spotify, Apple
Podcasts, or wherever you listento your podcast, or just share
this show with like coworkers,friends, or co students, or
whoever you know in your lifewho's doing Go right now. That's

(20:38):
it for the ad break.
Let's do a quick lightning roundand wrap this one up. Stick
around after the lightning roundbecause we also have an
interesting interview with MattBoyle from Bite Size Go and Ona.
Super interesting. You you gottahear that conversation.
Lightning round.
My thing for the lightning roundis a very French thing. Go team

(21:02):
is taking a week, of a quietweek, which is something I
wasn't aware of, but it'sactually in the wiki that
periodically through the year,the Go team holds quiet weeks
with no important conversationshappening to give people time to
focus on deep work. So not greatfor the podcast because we are
supposed to report on theimportant happenings with no

(21:24):
important conversations. Howcould that happen? But seems
like a no meetings week wherepeople are encouraged to take a
vacation, but they don't haveto.
They can also do really deepwork. So that's great. You can
expect people to not respond onthe Golang issue tracker and to
not have meeting minutes andthings like that. But yeah, I'm

(21:45):
down for them to do deep workand actually solve like hard
problems if we give them sometime. That's a really cool idea.
Alright. The show is notplanning to take a quite week,
but we'll we'll do something.What do you have for the
lightning round?

My item for the lightning round is an
article by Philmore Labs,entitled The Day the Linter
Broke My Code, in which theauthor describes, Well, they
implemented a feature in theircodebase, and they ran some

(22:19):
linter over it, and the lintersuggested a fix. They even
applied the auto fix that theLinter provides. But this fix
breaks the author's code insubtle ways that you may not see
if you're not testing properly.So, well, we won't dive into too

(22:42):
much details and not spoil thearticle, but it will be in the
show notes for all of the story.The your linters are not silver
bullets.
Yeah. For helpful.

We had a linters discussion in the the Slack
channel as well that's we'lllink in the show notes, where
Ken Smith mentioned like becausewe talked about naked returns,
in one of the latest episodes,we we talked about like, should
you introduce a Golang SeaIsland? And generally, were

(23:13):
like, don't use all the lintersin Golang Sea Island.

Oh, no.

And I was like, yeah, you you should probably
you you shouldn't probably useall of them, but my opinion is
add them a la carte, like startwith very few and then every
week or so, add one or two, thegood ones, right? Like there's a
lot of value in incrementallyadding those because you have
time to learn what they do,etcetera, etcetera. Yeah. Agree.

(23:37):
But in Ken's case, the linterswere okay.
They didn't introduce bugs.That's pretty rare that that
happens.

It's rare, but, yeah, it can happen.

So Yeah. Never accepted blindly for sure. No.
All right. Well, cool stuff.
Gabriel, thanks a lot forjoining. I hope you had Thank
you

a lot for having me. Yeah. I had fun a
lot.

If, people wanna find you online, where where can
they find you?

The best place to find me would be on my
website, my blog, gabnotes.org.

Link in the show notes, of course.

Awesome. There's an about page, and you
can find ways to contact me ifyou want.

Thanks, man.

Thank you.

Oh, this was great. We'll see you all next week. I'm
not sure about the dates becauseit's Jewish high holidays. My
parents are visiting The US.It's gonna be a hectic time for
me, but I'm sure we'll be ableto figure it out.
In the meanwhile, for our Jewishlisteners, which is like happy
New Year, because it's JewishNew Year, and enjoy your quiet
week. Maybe you can show yourbosses at work that Go has a

(24:44):
quiet weekend because you're aGo developer, no meetings for
you as well. Otherwise, yourcode won't compile or something.

Oh, man. I would love that.

You know what? Try it and report back next week
because Gabriel is looking forwork. Alright. Thanks a lot for
listening, everybody. Alright,man.
I'm so hungry. I could really gofor some some bite sized snacks
right now. If only there wassomeone on the call who could

(25:14):
help me with that. Oh, hey,Matt.

Matt Boyle (25:17):
Best introduction I've ever heard. I love that.

I mean, it's better than from the top ropes with a
steel chair, it's Matt I

Matt Boyle (25:30):
mean, that's I don't know which one to pick as the
best one now. They're bothpretty up there. Amazing.

They're both on the record. Everybody, it's Matt
Boyle. Matt, please introduceyourself.

Matt Boyle (25:40):
Well, I don't know how to follow the introduction,
honestly. But, yeah, I'm I'mMatt. I'm currently the head of
engineering at a company calledOna, where we are building an AI
platform using Go. I also amvery, very active to the best of
my ability in the in the Gocommunity. I run a website
called bitesizego.com, where Itry and teach people post
beginner lessons on on on on Goand spend a ton of time, running

(26:03):
a Go community on Twitter aswell, where I try and encourage
people to help people learn Goas well.
So two two sides to me, Isuppose.

Well, everybody's, multifaceted in a way. I'm sure
people listen to this show andthey're like, oh, Shy, he's the
Go guy. They what they actuallydon't know is that I suck at all
languages, just so you know.

Matt Boyle (26:22):
Yeah. And battle them all equally. Yeah.

Yeah. But let's start let's start with let's
start with Bite Size Go becauseI think I assume that most
people, if they have heard ofit, and that's how we got
connected, is through Bite SizeGo. And immediately, the thing
that's interesting isintermediate. You're like gating
it to not absolute beginners.And I'm interested in why, like

(26:46):
why would you do that in in thespace where most people are
beginners?
Like if you look at the at thesurveys and just in general,
normally most people arebeginners, not, intermediates or
seniors. So why not cater your,thing to the widest audience
possible?

Matt Boyle (27:02):
Makes sense. I think, I mean, my my general,
thoughts were maybe some peopleare still in that beginner
bracket because there's no wayout. Like, there's no clear path
to to get out to the to makethem feel like they're they're
leveling up and they're at theintermediate stage. Right? This
actually stopped this the wholeproject started.
The reason I started working atBiteSycho is, I was I was
working at Cloudflare for anumber of years there, and we

(27:22):
actually really struggled tohire Go engineers. Honestly,
There just wasn't a whole bunchof Go engineers out there. There
was people who kind of done thebeginner pieces, lots of people
who'd learned Java or stuff likethat. And we were we were
finding we were tended to hirepeople who hadn't wrote Go and
teaching them on the job, whichworked fine. It's very, very, as
we all know, like Go isincredibly productive.

(27:42):
It's very, very easy to get upto speed on. But I wondered if,
you know, I could do my bit, Isuppose, to help people get to
the next level outside ofoutside of work so that then
maybe they had more confidenceto go and apply for their go
jobs that were open and to bemore successful in it. So I
wasn't necessarily interested incatering to the to the widest
audience. I think it was moreinteresting to try and just

(28:03):
make, the content that I wish Ihad while I was learning is kind
of how I framed it to myselfwhen I was when I was doing
this. I was very fortunate thatI got some you know, I got to
work at, a company called Curvewhere I got to write Go for
production.
And then obviously, I got tospend time at Cloudflare where I
got to write Go for production.But not everybody gets to have
the opportunity. Right? And, sopart of thing I wanted to do is
give people the chance to maybehave the opportunity.

So it's sort of, what you would like to have had
when you were doing the thattransition. Think that's the
When, best like, people try tomake content for content's sake
instead of doing just thingsthat they are interested in in
doing at the moment, it's veryapparent. Especially now that
generating content is, well, atleast mediocre content is pretty

(28:46):
free.

Matt Boyle (28:47):
You know what I mean? Yeah. Absolutely. It's
definitely getting harder, Ithink, to, tell what's been AI
generated and what hasn't. I'mthere's some stuff I read and
I'm, like, doubting it now.
And I almost feel bad doing thatbecause, like, like, someone
spent a lot of time on this and,like, I you know, I am like, is
this was this generated byChatGPT? And it turns out it
wasn't, then you feel kindaguilty, but you're suspicious of
all content now. Right? You'realways like, there probably was

(29:08):
AI involved somewhere here.

I had a real, falling out with our this might
be a bit too much insidebaseball here for my startup,
but our designer is a properAmerican person and is writing
in proper American English, sothey use em dashes. And she's
writing really proper sentencesjust because she she cares. And
I was like, this is some CAgenerated. She's like, no,

(29:32):
Amdashes aren't correct.

Matt Boyle (29:34):
Yeah. I saw Paul Graham on Twitter and also there
was like a whole countermovement for a while. It's like,
let's let's claim back, youknow, the Amdash. Like, it
belongs to the English language.You should be allowed to use it
without being accused ofgenerating content.
Like, let's bring it back.

Yeah. For sure. So if I go to ByteSizeGo, what what
can I find that will help me,you know, level up as a as an
intermediate or a senior Goengineer?

Matt Boyle (29:58):
Yes. The first course I ever created was
actually on debugging. So very,very niche. It's like, okay. So
you start from your local,application and, you know, you
you log.
Like, you log to the console.That's how everyone starts out
debugging. Right? But there'sall these tools you can build on
top of that that, again, wasvery fortunate to get to
experience, as I kind ofnavigated my career in terms of

(30:18):
firstly using the debugger, thengetting into, like, Grafana,
using metrics, distributetracing, how to, profile the Go
application and figure outwhere, you know, it's could be
faster or why you're seeingslowdown. It's a kind of a
progressive walk through likelocal, don't really know how to
debug all the way through tothis is how incredibly large
companies save CPU by, you know,kind of looking around at some

(30:40):
of this stuff.
So that was the first thing Iever did. That landed really
well despite my The

coding is like a is like a an art. Whenever I
interview someone and, you know,we have a coding exercise and
instead of adding a ton ofprints because the our coding
exercise is on a real code base.Well, real. Like, it's on a you
give them, like, a toy codebase, but it has some bugs.
Whenever they add a breakpointinstead of just adding a ton of

(31:06):
prints, it's immediately, like,plus 15 points on my evaluation.

Matt Boyle (31:10):
Interesting. So I actually there was a a very
small amount of controversyaround my my debug calls from
some folks who said debugging isusing the debugger is is a waste
of time. Like you shouldn't youshouldn't be using it. If you if
you're using a debugger, you'vemade your program too
complicated. You should be ableto hold in your mind the, like,
the control flow and it'spowerful to have to walk through
it.
So we actually was like a like abit of back and forth between

(31:32):
folks on on X about, know,whether you need to debug or not
and I was really surprised byit. It caught me somewhat off
guard because to me I just don'tunderstand why you would not use
such a a useful and and powerfultool. I I love it. I still use
the debugger to to this dayfairly regularly even though, I
also use AI quite a bit todiscover the the first few,
issues I'm seeing, you

know. Mhmm. I'm not sure what's the case of let's
not use, you know, let let's notuse a debugger. I mean, it's a
tool. Using it doesn't mean Idon't know.
That sounds like an interestingdiscussion, but it's it's a bit
too much of a rabbit hole for usto to fully flesh out. If we
actually, if we flesh out everydiscussion in your x threads,

(32:15):
then we're probably we canprobably spin off a whole
podcast just on, Matt is arguingwith people on Twitter.

Matt Boyle (32:21):
I like the name of that podcast. Maybe we should
make it.

Okay. So there's a debugging course. I see there's
also, like gRPC courses, CLIcourses, and recently AI
engineering, which is likeprobably pretty hot.

Matt Boyle (32:35):
It is. So the the CLI course, partnered with one
of my friends, Marion, who's asenior engineer at Netflix. Like
And that's the thing I alwaystry to do with these courses is
I went and found people, well Ijust happened to have in my
network. Honestly throughGopherCon nearly everybody who
made a call or has worked withme I've met at GopherCon so you
know shout out to going toGopherCon and meeting people
take the hallway track speak atthese conferences. You meet

(32:58):
great people.
But wanted to meet people who'ddone these things in industry,
right, and could could bring theenterprise angle to them. So
Mary did one on CLIs and how tothink about them with her
experience in Netflix. My friendChris did one on gRPC with his
experience at Cloudflare andGoogle and, like, how to think
about it from there. And thenthe latest one on AI,
engineering is is is from myfriend who works at Plane, which

(33:22):
is a like an up and comingcustomer support startup, but
he's he's truly excellent. He'dactually done a bunch of courses
before for, I think it was Udemyor Udacity, and, generally is a
great educator and has had roleseducating before.
So his AI engineering course isis fantastic, actually. I I
really enjoyed working with himon that one.

Cool. I'm wondering if, the how much the AI
engineering course is, like, aproper AI engineering course? Or
how much is it you just, tookthe onboarding from Ona and
packaged it to the course?Because it seems like super
relevant. But before we beforewe get to Ona, is there like
who's the who's the persona,like, you you think should visit

(34:03):
Bite Size Go?
So it's like a team lead lookingto upscale their team or someone
preparing to pass, a seniorinterview. Is that the sort of
people you would imaginejoining, like, this community?

Matt Boyle (34:14):
Yeah. So there's there's two personas, I think.
So one is you've done thebeginner content and you're
looking for the what's next.And, I have a blog somewhere. I
need to update it, to be honest.
But there's a there's a blogarticle I wrote, which is really
popular, which is called I thinkit's called learning Go in 2025.
And basically, I I take it fromwherever you are today. And I
say these are things you shouldlisten to, these are things you

(34:34):
should read, these are thingsyou should learn, I'll get you
to the next level. Some of itincludes my courses, some of it
means go elsewhere, go and reada blog, go and, you know, listen
to old Go Time episodes, may itrest in peace, and, and, you
know, some of the blogs andstuff that I really enjoyed. So
you should be able to find apath from this is where I am
today, so this is where I'mgoing next.
I also at one point published a,it sounds similar to your coding

(34:55):
exercise actually that you do atyour company, which is great. I
have a GitHub repo where Ipublished a semi working Go
project that's got a bunch oferrors in it, and you can use
that to try and figure out how,like, where you are at being
able to debug someone else'scode. Like, can you see issues
in it? It's got security issuesas well as business logic errors
in it. So, you know, you can seeif you could line up to that.

(35:16):
And I give a rubric of this iswhat an intermediate answer
looks like. This is what I, youknow, expect a senior engineer
to be able to answer to it too.Team leads could come and say
like, oh, I I want my team tolearn this topic we're about to
use. Like, gRPC is one that'squite foundational. You might be
like, we're about to use this atwork.
Hey, everyone go and watch this.This is how, you know, they use
gRPC at Google Cloudflare. Like,there's some good insights here.
So probably kind of both ofthose audiences. I hope are

(35:37):
catered for, but let me know.
I also did a small thing lastyear, which kind of goes against
my original promise, but seemsso valuable I did it anyway,
which there is a course forfree. I did with, I did it with
JetBrains and it's calledlearning Go with Goland. And, I
take you through, starting fromyou can write a line of Go all
the way through to building,like, two or three projects.

(35:57):
It's completely free. And if youcomplete the course, get a free
year's license to to Goland aswell.

Oh, that's that's actually pretty cool.

Matt Boyle (36:04):
Absolutely. So it was a no brainer. Was like, of
course, I'll do this. This makessense. Like, why why would I
not?

The question is do you use, Goland? But I assume
now the answer is you use ONA.

Matt Boyle (36:15):
Yeah. So, ONA's pretty interesting. Let's jump
to that. So

so I'll say as a as an outside observer, I'm not I'm
not a a user, but if Iunderstand it correctly, it's
stepping into a pretty red oceanof, AI coding thingies that I'm
afraid already to use like theword agent because it's been

(36:41):
devoidable. Like, it's the lifehave been sucked out of that
word. It's meaningless now. I'mafraid of using the word
assistant because it sounds likeit's gonna talk to me. I don't
know what to call it even.

Matt Boyle (36:53):
Yeah. Makes

sense. It's an AI coding thing thingy.

Matt Boyle (36:56):
Yeah. So let me let

me Maybe I can direct it a bit more.

Matt Boyle (36:59):
Let me demystify the thingy. So what we built, what
we built to ONA was, we builtwhat we call mission control for
AI agents. So we've all, we'reall using AI agents, We're using
Cursor, we're using Codecs,we're using CloCode. These
things run on your localmachine. They make changes to
file systems.
They don't have a whole bunch ofcontrols. So we've all got that

(37:20):
annoying thing where we have toallow every single command to
happen. We also have topotentially risk the fact that
it could go ahead and deletefiles, etcetera. There's also
this whole other category ofpeople called background agents
that you've got, cursorbackground agents, you've got
codecs. These things are allawesome.
I I love all of them. But all ofthese things require you to kind
of change the way you work andare very hard for very large

(37:44):
enterprises to adopt due to riskappetite reasons. So this is
where kind of ONA fits in. Sothe way the ONA platform works
is we deploy it into largeenterprises VPCs. So we deploy
into the AWS or their GCPaccount.
And when you go into owner, youselect one of your projects that
you're working on, and you clickstart. Opens a remote

(38:07):
development environment insidethat VPC that's completely
sandboxed and ephemeral, and allof the agent interactions
happens inside the sandbox. Soif the agent does anything that
you're unhappy with, you justblow it away. So it's very, very
controlled. It all happenswithin the network perimeter.
And it also means that if youwant to kind of have tighter
restrictions around yournetworking, you can do that so
that the agent has limitedability to do things. And we

(38:29):
also have guardrails, which arelike, agents aren't allowed to
do this command. Agents aren'tallowed to use this MTP server.
So it gives, companies thatcouldn't necessarily use all
these AI tools that you can useat home on the weekend. We give
them a way to be able to usethem, in their enterprise, which
means that, you know, they canhave the same productivity boost
that that we get.

Awesome. And just like in a glance as the as the
head of engineering, this you'resaying this is written in Go,
like this platform. Right?

Matt Boyle (38:55):
The whole thing's written in Go, yeah.

So all But I thought Go was bad for AI. I
thought you needed to use Pythonfor AI.

Matt Boyle (39:00):
I mean, that's, you know, that's that's what people
have to believe, you know? Like,I think one one thing that
That's always

what they want you to Yeah.

Matt Boyle (39:09):
I don't know. I always, so Go is, as we all
know, and I'm preaching to thechoir on this podcast,
incredibly versatile. I thinkthere are some disadvantages to
using it for writing, AI agents,mostly just around ecosystem.
Like in Python, there really isa package for everything and
everything's been done before.If you write an AI agent or or
platform in something like Go,you do have to write a lot more

(39:32):
from scratch.
I don't necessarily think that'sa bad thing. It means we're
very, careful about thedependencies we we do pull in,
and it also means we cancontribute back if we need to
where it makes sense. So we werewe're actually discussing today
probably gonna have to make a acommit to the standard library
to, we were struggling with somethe MCP support. It's not quite
the same as the, the TypeScriptMCP support. So we're probably

(39:55):
gonna contribute back.
But I love everything about thatsentence. Like, I want us to be
known for for building somethinginteresting in Go. And if we can
help forward, like, the, youknow, Go Go is a place to build
AI, we'll we'll do that too. SoI'm actually excited about the
the fact that it's notnecessarily popular yet, but one
of my personal missions is tomake it so it is. So, you can
let's see where we're at intwelve months.
Maybe I can come back on thispodcast and we can review

(40:16):
whether as a company we weresuccessful in in bringing some
AI adoption to building AItooling.

We actually mentioned in the, like, pre roll
to the show that, it's alsohelpful. In in some sense, it's
helpful that, packages are notas willy nilly in go as in NPM
just because this week. Well,this week as we're recording,
I'm not exactly sure when theinterview is gonna come out. But
there was I'm sure that it'sgonna my sentence is gonna be

(40:43):
true every week. This week, NPM,security vulnerability with
supply chain has been exposed.
Some developer has been pwnedand now all companies are at
risk when they upgrade. I don'tknow, shift left or ANSI colors
or something like super basic.

Matt Boyle (40:58):
That's it. And you're speaking to two really
important things here. One isthe Go ecosystem generally is is
is very good with this stuff.And we have this sort of
principle of not pulling independencies unless we actually
have to. And I think with, howeasy it is to generate code
that's actually pretty good now,there's even more reason you
probably shouldn't pull in awhole bunch of dependencies.
A little a little copying isbetter than a little dependency

(41:19):
as we say. But the other thingis what you just spoke about
about that dependency risk isexactly why large enterprises
are a bit nervous around, youknow, fully adopting AI agents
because things like that arevery easy to introduce,
especially if you're sort ofvibe coding something and not
paying too much attention towhat's happening. So having
guardrails and controls is isone of our main focus areas, to

(41:40):
be honest. It's like, what doesit look like for an AI agent to
still be productive, but alsooperate within the risk appetite
of of of companies that, havestrict regulation around them.
Right?
Like, they don't have the sameability to just let engineers
run whatever they want on thelaptop.

Mhmm. So how did these guardrails in general, I
understand that, you know, likea bank or like a hospital can't
have their engineers just runwhatever because they're under
regulation. But in practice, howcan you make AI agents safer?
You mentioned you're deployingto the customer's cloud, so
that's like data residency. Butis it just like a system prompt

(42:19):
that's like, please, please,please, please don't
hallucinate.
Please, I beg you. My entireproduct page is all about don't
leak the keys or is this is itsomething a bit more
sophisticated?

Matt Boyle (42:29):
I mean, it's it's all it's it's that plus a lot
more. I mean, the system promptis surprisingly effective.
Honestly, I won't I won'tpretend that we don't have a
system prompt that sayssomething like, you know, do do
smart things please, misteragent. Like, it's it it does
work. But it's it's that's layerone defense.
Right? Layer two is, then wehave sort of network controls.
So, you know, enterprise can setup their own like, these are the

(42:50):
egress rules around this remotedevelopment environment. This is
what we're comfortable with.Then there's also sort of the
controls around MTP.
So every time before our agentcalls an MTP server, it will
check policy to ensure it'sallowed to do so. And if it
doesn't, it's just blocked. Soit's not allowed to do so.

Cool. Are you using, like, some Go ish, policy
language like Rego or orsomething like that? Like, it's
obviously an internal,implementation implementation
detail. I'm sure just the Golisteners will love to nerd out
on, oh, why did you it's Ricoh.You should have chosen JSONata.
No. You should have chosen thispolicy language. No. You should
have chosen this policylanguage.

Matt Boyle (43:27):
Yeah. So we we did look at Open Policy Agent,
generally as a as an approach.I've actually used that
previously at Cloudflare verysuccessfully. It's it's it's
really cool. I think one of thechallenges with AI agents is
they have a very heavy rewardfunction to achieve the goal,
and they will find ways aroundpolicy.
They'll do really creativethings to to be able to to to

(43:49):
break out of it, which is whichis really interesting actually.
So what we're actually investingin at the moment, is like going
down to the kernel level. Like,we're gonna be looking at eBPF
and looking at ways to actuallyreally stop at the kernel level
agents doing things that we'renot comfortable with us doing.
So you can expect to see us dosome pretty pretty interesting
things very, very soon in thisspace.

Super interesting. Like, they you have an
adversary, an insider risk,like, you're you're you're not
allowed to use this MCP, so itwill use, like, the back the the
backup MCP or something?

Matt Boyle (44:22):
Yeah. The so the one example we saw play out in
practice was, we have this wealready have this concept of
command denialists. So these arethings the agent is not allowed
to do. So let's imagine youhave, an AWS CLI in your
environment, and we block it. Wesay you're not allowed to use
this.
So the agent will discover it'sgot the the AWS CLI. It will try

(44:43):
and use it. It will get blockedby our platform, but it won't
say, oh, I got blocked. Fairenough. It will try and find a
way to interact with the AWS CLIinstead.
So maybe I'll make a postrequest instead, you know, that
looks like the command thatwe're in CLI. Yeah. So you you
always have this balance of AIagents there to serve you, but
also it's it's they're soheavily, invested in the success

(45:04):
of the task that you've giventhem. You've got to really make
sure you have these controls inplace. So we're investing very,
very heavily in guardrails.
And I think over the next yearor so, especially in large
companies, you'll see theconversation move to this. And
so I like to think we're alittle bit ahead of the curve
here. We, we are investing to,to make this so when, the FOMO
wears out a little bit about,you know, we need AI and we need

(45:26):
it now. The next conversation isgonna be, okay, we've let
Pandora out of the box. Like,what do we what do we do here to
make sure that we can keep allthe controls we want to have in
place?
And I think this is this is theway we're gonna start
positioning too.

I was, I just moved here to the Bay Area and here
the AI is like, you know, you goto San Francisco and every
billboard is like about AI. It'slike, it's something in the
water. It's crazy. It's like avery different vibe. I went to a
meetup with, you know, peoplefrom the like Frontier Labs,
OpenAI and Amazon Bedrock andand Anthropic, people from the

(46:00):
safety teams.
And they sit and they're like,they're pondering questions
like, if you train because ofthe reward function, if you
train the model to not havelike, if you look at its chain
of thought and you train it tonot have bad thoughts, will it
stop having bad behaviors? Andeven they, like, their answers
are kind of vague and scary.Like, no. Not really. It just

(46:21):
finds ways around us, so we needlike, it it literally feels like
they let Pandora out of the box.

Matt Boyle (46:27):
It really does get there's there's more than ever,
there's a really fine linebetween engineering and
philosophy, isn't there? Like,I've found we we have to spend a
lot more time in the in thephilosophy, area than we ever
did before, which I I think isso interesting. But, yeah,
there's a lot of hard questionsand things to figure out here.

So coming back to, the ONA platform. So it seems
like something that developerswould use. Seems like developers
are your users, but that, like,big companies would buy. So you
you sell it to like bigcompanies, so you need to talk
about security, but, the theactual people using it are
developers like myself. What'slike my experience?

(47:07):
Am I like writing code in an IDor am I like just talking to, my
agents like over Slack? What'swhat's my experience like? You
mentioned mission control andthat to me reads, you're not
gonna write code yourself.

Matt Boyle (47:20):
Yeah. So the we we work it works in a couple of
ways. So the first one is, youknow, imagine you're a large
company. Our biggest customerstend to be some of the largest
banks in the world. And let'ssay you've got 10,000 repos.
Right? They're all slightlydifferent. They all start a
little bit different. Like, areally common metric that I'm
sure every company measures islike time to first commit. And

(47:41):
it's not really unusual for anengineer to join a company and
it take three, four weeks thento be able to be productive as
they figure out how to navigatethis new landscape.
So one of the first things we wedo and why developers often love
us, even though we talk aboutsecurity lot, is our goal is to
get that time to first commit toone less than one day. And we
and the way we do this is, bystandardizing development
environments around, Linux VMs,which is ultimately what we

(48:03):
deploy into their VPC. It meanswe get to make a lot of
assumptions, which is reallypowerful actually. Everybody's
working no matter if they've gota Mac, they're on the phone or
on an iPad. You're ultimatelydeveloping on the same box,
which is remote in inside, yourVPC.
We use dev container standardfor that. So you define your dev
container, and then ultimately,we can we can spin that up into
your environment. The way I likeframe this is we get your best

(48:24):
engineer to configure it once,and then everybody gets to
benefit from that. So that's thefirst way is is so you spin up
your remote environment, andthen you can connect to it
however you like. So we've gotnative support for, Versus Code,
including Versus Code browser,cursor, Windsurf, all the
JetBrains stack, you'reeffectively inside owner.
There's one button click to jumpinto your ID of choice, and you

(48:44):
can work directly in there.Especially for the Versus Code
forked repo forked browser IDslike the Versus code itself and
cursor and WinSurf, theexperience is very, very similar
to developing locally. It'slightning fast and you can still
use exactly the samefunctionality. So you don't
really feel any difference, fromfrom working locally except the
actual code and everythingthat's happening is is on a very

(49:05):
beefy box inside AWS.

But but if I want, I can spend, like, a ton of
background agents and try tohave them, like, solve the the
like, stuff in the backgroundand then just review it.

Matt Boyle (49:17):
Absolutely. So that's the that's the one way we
show up is like this sort of, wecall it craft mode, like this
manual mode where you'redeveloping exactly the same way
you always have, but with aremote environment.

Boutique mode, organic, farm to table code.
It's becoming more and morerare.

Matt Boyle (49:32):
I actually like boutique mode. I love that. But
then

Feel free to take it.

Matt Boyle (49:37):
I I might do. But then we also have, you know,
what what's kind of commonlybecome known as as as background
agents is you can also open theowner platform. You can pick a
project, and then you can say,here's the task I wanna get
done, And it will spin off anenvironment and use the dev
container standard. Everythingyou configured for humans
actually makes it great foragents too. It will spin up that
environment and then you it'llstart doing the work for you,

(49:58):
and let you know when it's done.
You can then one click jump intoan IDE if you want to. So, you
know, a lot of tasks we see thatagents generally get to 90%
completion, especially formoderate to hard tasks. So once
it's done, you can then jumpinto Versus Code browser, which
we have alongside the agentconversation. You can review the
code there, which I do often,and you can also make minor
changes yourself and then pushit together. So it's kind of

(50:21):
this agent human collaborationwe're we're going for.
But often we're seeing ourengineers spin off five, six,
seven tasks at once. They workin the background, and then I'll
start a new environment, open anID, and do the more interesting
pieces of work. Right? Thefeature I'm excited about, I
work on that. And the agents canwork on the CV updates, the
updates to Readmes, the largescale refactors, what whatever

(50:41):
it might be that, you know, youdon't want jump out of bed in
the morning to do.
Let let Yeah. An agent handlethat, and you work on the new
feature.

I set up a a, like, hacky thing for myself, like,
quite a few months ago with thegit work trees. I was like, ah,
I'm gonna do git work trees andthen I'm gonna have multiple
instances of, like, cursorrunning and I'm gonna jump in
between those windows, etcetera,etcetera. And then everybody
came out came out with likebackground engines. I was like,
ah, all that wasted time.

Matt Boyle (51:08):
Yeah. This is this is what you really wanted. Yeah.

Yeah. Very cool. One thing we talked about in the
beginning of the show, which waslike in the pre roll, which was
interesting and I wanted tomention here was, like you're
hiring I don't know if to callit like woes or or problems or
opportunities. I don't know. Youyou decide how to frame it.
But let's say our listeners arelike, oh, Onus sounds cool and I

(51:32):
like working on dev tooling andit seems relevant to me. Where
are you based by the way? It'sprobably a good question.

Matt Boyle (51:37):
Yeah. So I'm based in London, but, we're
international as a company. I amhiring a couple of roles at the
moment actually. One is Goengineers. So if you are I would
say trending towards senior Goengineer just because of the the
problem space we're operating inis is pretty challenging.
So if you have networkingexperience, you have some Go
experience, we'd love to chat toyou, especially if you live in

(52:01):
London or Europe, which istypically where we're kind of
concentrating on hiring at themoment. But we're also, hiring
what we call forward deployedengineers, and these are people
who would join our biggestcustomers on-site to build
custom solutions for them, helpthem kind of roll out ONA across
their enterprise. We're hiringfor those in New York and
London. So if you have Goexperience and you love the idea

(52:24):
of actually spending more timein a customer facing role, but
also particularly love to speakto you.

Yeah. And of course, don't forget that Cup of
Go gets of course a 20%, cut ofyour No. I'm just kidding. Free
99, that's the price of areference, from the show. We
would love to hear Like if Iactually I actually in the I
think last year, someone waslike, man, I I got into contact
from someone from the show andgot a a job through the Slack.

(52:51):
I was like, oh my god, that's sogood. It's like a totally
unintended side effect ofcreating this community. But
interestingly, these, jobs arenot listed on your careers page.
And I I thought it it was aninteresting discussion. Why?
Like, why not? Because why notshare this role?

Matt Boyle (53:09):
Yeah. I mean, we we may put them back on there. I
think one of the challenges Ipersonally found is, you know,
we're a relatively small companyand we do hiring very, very,
intensely. Like, I will spendhours with someone to like, when
when interviewing them, we wannahave people who are great great
fits. And so we invest quite alot into into hiring.
So when we had job roles open onthe website, we kind of got

(53:30):
quite a lot of applications. Alot of them maybe didn't even
have any go experience. Maybewe're in countries that we
couldn't support, unfortunately.We're quite a small company. We
can't really support visasponsorship.
We have that all listed on theon the job spec. But think
because it's so hard to find jobroles, people just apply anyway
and and hope for the best, whichI, you know, what it's worth,
like, totally get. I, I admirethe hustle, but it's just not

(53:51):
something at the size we're atwe can support right now. So,
kind of doing it the oppositeway, doing, are always welcome
to reach out to me on Twitter.And I would love to either help
you find a role or, you know,see if you're a fit for owner,
if it makes sense.
So you can find me on Matt JamesBoyle. Do my best to respond to
every DM I get. But ultimatelywe do a lot of sort of hiring

(54:11):
where I'll reach out to peopledirectly on LinkedIn, on an X.
And also we'll do work withrecruiters who go and find us
like three to five people tointerview instead, just because
we find that to be a lot moresuccessful than than having
inbound opportunities at themoment. One thing you could take
away from this, which is maybemore optimistic, is I think a

(54:32):
lot of people feel a little bitlike we feel.
So if a company isn't hiring atthe moment, somebody really
excited about working, is youknow don't be afraid to reach
out to someone you know whoworks there and just say, hey
I'd love to work here, here'swhy and you know just give a
little bit of insight into whyparticularly you're interested
in potentially working thatrole. You'd be surprised if that
they may may be looking or aboutto be looking and how much that,

(54:53):
warm outreach and, you know, thefact you've kind of targeted
your your search really countsto to people who are spending
time hiring. But it wouldencourage you to do that.

Yeah. So, you can find ONA on on Twitter, on X,
whatever. I'm gonna call itTwitter, whatever. At Ona
underscore HQ and they can findyou at at matt jamesboyle.

Matt Boyle (55:17):
Yeah. That's cool.

Which is way more official than mattboyle. It's
even kind of intimidating. Yep.Sounds like it sounds like a
bible version.

Matt Boyle (55:26):
Yeah. So I don't I don't know much about, American
politics, but there's a I thinkthere's a semi famous, maybe
he's a politician or maybe he'sa journalist, I don't know,
called Matt Boyle in The US. Andso I think he's got all the good
handles. And so I have to I haveto, I I have to put my middle
name in there to be able toclaim anything. I once tried to
set up a Google alert to see ifanyone was talking about my goal

(55:47):
courses, and I got a bunch of,political charged notifications
instead.
So I had to I had to turn itoff.

Luckily, with a name like Shainich Mahd, you're
not gonna that's that's notgonna happen

Matt Boyle (55:57):
a lot. You don't have the same problem, no?

No. No. Nobody's talking about me. Don't worry
about it.

Matt Boyle (56:03):
Well, that's not true.

Well, we're running up on time here. I actually wish
we could talk more, but I I'mactually literally on a on a
meeting. We do have this, onequestion, which I think for
someone like you would be extrainteresting. So you mentioned
you worked at Cloudflare and youhave these Go courses when you
talk to these like illustriousGo people from like other

(56:27):
companies, Netflix, whatever. Wehave this question we ask all
our interviews, which is who wasthe person who sort of
influenced your Go journey themost?

Matt Boyle (56:36):
So I I have a I have a really nice story here,
actually. I'll try and keep itsuccinct because I know you
gotta go. But one of my, one ofthe first blog posts I ever
remember reading was DaveChaney's on how to do functional
options. And we shared thisaround our company and we
started using it, and I thoughtit was really cool. And I spent
a lot of time reading aroundDave Chaney's blogs and his
writing.
I always enjoyed the depth hewrote wrote at. And then a

(56:57):
couple of years ago, I I wasasked to do the keynote at
GopherCon in Singapore. And so II went there, really good
opportunity. I spoke at it, andone of the speakers was Dave
Cheney. So I got to spend sometime with Dave Cheney, got to
know him.
We became really good friends.We now text often. I went to
Australia in December, which iswhere he lives, for Christmas
and we met for met for dinnerand we still text back and forth

(57:19):
now. I mentioned to him aboutthe functional options blog post
and like how it influenced meand he was like, oh yeah, I need
to update that. So I alwaysenjoy that.
But yeah, like he's he's beenincredibly influential to a lot
of people. Annoyingly, he's alsoan incredibly nice person. So
he's one of those people who'svery into it influential, very,
very smart, and also reallynice. So he's just one of those
people who are just unflawed asfar as I'm concerned, and I I

(57:41):
feel, a great privilege to theysay never meet your heroes, but
I met mine, and it turned outhe's a good guy.

Wow. That's awesome. I think you're not the
first mention first person tomention Dave, by the way, as the
as an influential character intheir GO journey. So so good on
Dave for for getting people onthat on their wave. You know
what I mean?

Matt Boyle (58:00):
Absolutely. I didn't know he

lived in Australia.

Matt Boyle (58:02):
Yeah. Yeah. He lives in lives in Sydney. But they've
been, I think he was writingsome of the earliest Go content.
He's, you know, you can go backto sort of 2012, 2013, like him
and Matt Raya, like, were thethe people who I used to come
across often when I was, youknow, researching Go content
back then, they were using itbefore it was even Go one, you
know.
So they really were Yeah. Forsure. The bleeding edge.

Well, the the main thing, to take from that is what
what you mentioned about, Ithink you mentioned it twice
already, go to conferences, toeither to meet, candidates or
now to meet Dave Cheney. I thinkgoing to conferences is is a is
a good, goal in general.

Matt Boyle (58:42):
I have another funny story on that actually. So the
at GopherCon London last year, Iwas stood talking to Dave
Chaney, and he doesn't actuallyhave his picture very much on
the Internet, so people don'treally know what he looks And
someone was talking to me andDave Chaney about one of the
blogs that Dave Chaney wrotesaying how good it was. And Dave
was just at their nodding, andthey hadn't really clocked that
it was him. And, he didn't tellthem, and, I I thought that was

(59:05):
very funny, and I I admire thatabout him too. But, yeah, at a
lot of conferences, he's likewandering around and people
don't really sort of recognizehim, I don't think.

I have no idea how the how the guy looks. You know,
there are some people who I Iimagine as as gopher emoji just
because instead of a or a gopherlike cartoon instead of the
person. Even if I've met them,like, Bill Kennedy from Ardan
Labs, like I've I've talked tohim on the show. I know how he
looks like. But like, I can'timagine him as anything else

(59:33):
than a cartoon gopher.

Matt Boyle (59:35):
Yeah. With the hat. Like, that famous hat of his.
No. Bill's great.
He's he's another person I firstmet at at GovCon in Singapore,
actually. So that was it was agood good GovCon. He's has so
much energy, and he's such agreat, educator, especially, you
know, like he he did this talkthen where he talked about
something that typically wouldbe pretty dry. It was about like
limits in Kubernetes. But likehe came on stage, he's like

(59:56):
dancing to like, music fromMiami, which is where he's from
and he loves and he makes areally dry topic, really, like,
energizing.
And I've always admired thatabout him. He's a he's he's a
really great educator.

Shay Neymar (01:00:07):
Awesome. Well, Matt, thanks a lot for coming on
the show. I've put, we'll putthe links to, Ona and your
Twitter and, Bite Size Go in theshow's description. You
mentioned you you have, like,DMs open. People are, like,
welcome to talk to you afterthis.
Right?

Matt Boyle (01:00:24):
Absolutely. Yeah. I'll I'll always try and respond
to to everybody. If I don't getback to you, it's probably
because I've missed it. And alsoTwitter has this, they attempt
to classify your inbox for youvery unhelpfully, where they're
like, no, this person's spam, soI'm not gonna show you it.
And when a lot of time it isn't,so I apologize in advance if I
if I miss it.

But, yeah, DM's open, as, the youngsters tends
to say. And thanks a lot forcoming on the show.

It's a pleasure. I hope to come back soon and tell
you how successful I was aboutdriving, AI adoption in Go.

We'll see you next year. Actually, we the survey,
just, came out this week. So,we'll know what's the adoption
now, and then we'll we'll beable to compare it, you know,
exactly in one year.

And I'll take full credit for any uptick.

But not if not for any downticks.

No. No. No. That wasn't me.

Yeah. For sure. Alright. Thanks a lot, man.

No worries. Thank you.

All Episodes

Episode Transcript

Popular Podcasts

On Purpose with Jay Shetty

Stuff You Should Know

The Joe Rogan Experience

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}🪖 Gab's Notes on this week's Go news and an interview with Matt Boyle of Ona and ByteSizeGo

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}On Purpose with Jay Shetty

Stuff You Should Know

The Joe Rogan Experience

All Episodes

🪖 Gab's Notes on this week's Go news and an interview with Matt Boyle of Ona and ByteSizeGo

On Purpose with Jay Shetty