Thomas Bouldin: Insights from Staff Tech Lead at Google / Firebase

today I’m really excited to welcome Thomas Balden a true leader in software engineering Thomas has worked at some of

the biggest companies in Tech he’s currently a staff Tech lead at Google leading projects like Firebase app

hosting the original Firebase hosting gen kit and Cloud functions before that

he was at Facebook where he helped scale pars to over a billion users and even earlier he was shaping Microsoft’s

application model Thomas was also recently married and is an IDC staff

instructor Thomas it’s awesome to have you here today yeah it’s great to be here awesome um Thomas you have built

some really cool systems that have handled billions of connections and um so I’m curious what’s been harder what’s

been more challenging scaling this significant Cloud infrastructure to handle all these pedabytes of data or

teaching someone to dive in open water at night uh so honestly teaching scuba is

closer to my now former management responsibilities in my technical ones uh

and night diving isn’t that challenging it’s part of the advanced core so people have already been familiarized with their gear and have some confidence

before they agree to the night dive so after that it’s really just about like Good Vibes share with them that I love

it and then they get excited and they do it too um but there are some parallels so like this the Patty scuba curriculum

Builds on previous knowledge to teach people the next skills and similarly Big Data starts with small data uh you learn

lessons you rearchitecturing 10 to 20x scale um you know at Google

many of these lessons are codified in the tooling and systems we use but parse we definitely learned the hard way and

had to scale ourselves so you started your career at Microsoft and you’ve contributed to lots

of different aspects of innovation at Google and then also with Parson Facebook um can you tell us a little bit about some of the experiences that

shaped your approach to engineering yeah a really early moment for me was when I worked on Windows on a

feature was going to be built that was request by a large volume customer I was fresh out of college and treated

software like an art um I was almost socialist in that regard like you shouldn’t software be designed for all

customers is it unethical to give a particular customer say in our road map

and over time I just accepted that software engineering is engineering you solve problems you solve business

problems um and so I became much more comfortable with that and in fact uh I’m

dealing with it more and more every day with Firebase app hosting where my internal Mantra is Firebase easy

Enterprise ready so one of my current projects is actually making it easier to extend your website with more of Google

Cloud specifically because it’s blocking a Fortune 100 customer uh a more recent

lesson that we shared was the value of documentation and ease of use uh if your future feature isn’t learnable it just

doesn’t exist uh James taught us that at parse with his Hacker News article on good documentation that even made it

into the quarterly newsletter uh it led me to a philosophy I call Demo driven development uh you need to design your

product to minimize the time from Reading docs to just saying wow is these kinds of products that are really easy

to pitch and learn and they’re much more likely to get adopted yeah I I love this

idea that if a feature isn’t learnable it doesn’t exist because there are so many of these incredible features uh

products that like they are hidden and if you don’t know they exist all this work that went into them doesn’t really

help the user get better be more effective um and so I think I I love this statement um let’s let’s talk a

little bit about your time at Google um you helped redesign something called the image search indexing pipeline uh it

sounds pretty complex and I would love to know what were some of the biggest challenges in tackling such a a core big

system that was presumably at huge scale yeah yeah I mean fundamentally the first

challenge is just understanding and wrapping your head around how big that scale is uh like at the at the time 100

megabits per second was the fastest internet we had and during college visits I used to quiz people how long

would it take to download one petabyte of data to process it and the answer is

about 2.8 years uh so when you realize how big that problem is you realize how

bad congestion in that pipeline can be so one of the biggest problems we faced is what we call the fan in problem this

is when you have many rights that are going to go to the same SP part of the database and overload it so in image

search you have many websites linked to the same image URL you have many image URLs linked to the same bytes and the

same bytes turn out to be re-encoding to the same image and all of those could hotspot the database so we had some

Innovations where you know we could drop from synchronous to asynchronous processing when we detected

contention but frankly just that’s not enough like imagine that that one pixel

tracking pixel for marketing it’s everywhere you cannot handle that load and so in these particular edge cases we

learned to just drop it and not worry about it um you you you were part of a small team at parse which was a y

combinator startup in the Bay Area as employee number seven which seems kind of crazy um getting in there so early

especially you know leaving Google and and having been part of such a great business um you helped scale it before

it was later acquired by Facebook what was it like to be at the heart of that team in the early days and then what was

it like transitioning to Facebook you know the vibe was just so great it was so many smart people and we

just got stuff done you know coming from big companies I really loved the lack of

bureaucracy and how well we could just iterate and launch product rapidly uh and it really was just a product of

passion uh it really drove the product forward you I remember I was so addicted I spent my weekdays focusing on my

assignments and then some week begins focusing on the stuff that I thought would be a good idea but wasn’t on a

road map um you know and I had a bit of a hybrid role uh so I started with

things like the file API and then I moved into parse push uh I was the lead there which technically meant I was the

only person on it uh and then we we grew a little bit more um but I also I

touched everything I I medal with our API servers the data layer the compute product I was kind of a a devops

engineer for a while just I couldn’t stay in one place um you know one of the things that

I think is really interesting is that our push Network for Android actually predates Firebase Cloud messaging or

back then called Google Cloud messaging and we learned a lot of lessons the hard way like Telos do some weird stuff with

just like breaking connections and not even telling you they’ve broken the connection because they don’t want you to reconnect and so like we had to learn

to detect and fix that type of problem uh it turns out some viice drivers not

only throw errors but they’re errors that aren’t documented and aren’t declared in the function so you can’t

handle them directly and so we had to find ways to fix that and eventually we we got over all that type of

stuff um you know and and there was a lot of optimizations we had to make as

we grew uh especially when the transition happened to Facebook when we transition to Facebook uh that that

growth curve was was steep uh we spent all of our time fighting fires and and we had to work overtime just to make the

infrastructure scale so that we weren’t spending all of our time fighting fires um part of that included rewriting our

system from Ruby to go uh we had some Advanced software in Ruby like rescue that I had to reverse engineer and

reimplement and go uh but that gave me the opportunity to add a couple of cool new features um you know the ability to kind

of say the heavy work is done go ahead and give me another task but don’t shut down the service yet because I still

have some light work to do um there was also some challenges

because we were built on and it wasn’t as mature back then as it is today uh we were told at the time that

we were the world’s largest installation and so we ran into all the edge cases uh

especially I remember Geo queries were just problematic and so we we had to

really work that was where like my devops side came into play is I had to work with our operations Engineers to

find contingency plans so we could do these MTH mive you know billion device pushes without affecting our API servers

our bread and butter products um I I remember a few times specifically where there would be some

related problem and I would get an email from one of our customers like WTF

what’s going on over there guys and I would basically drag our CTO Kevin onto

the call to have him explain and it was sometimes really painful but you know that’s sort of part of the exciting

aspect of things is pushing things to scale trying to figure it out as you go um so with all these different platforms

you’ve worked on uh I’m sure there’s been a lot of pressure and you know whether it impacted your sleep or you

know it made things hard and challenging can you think of times where like the plan it wasn’t going the way that you

wanted to and how did you handle those kinds of situations you know one of the first

things that comes to mind is the cloud functions for Firebase V2 launch um we

we really have a a timeline we wanted to launch at google.io with the vanilla Cloud counterpart uh but we actually

have more use cases um that we stitch together than just the cloud counterpart

and they weren’t really on people’s road map and supported I literally had a dock called

Crash Landing to to try to make sure that the launch went as well as possible

given that things were kind of being tied up um you know we outlined all the

gaps we had mitigation plans uh sometimes we were able to get a partner to prioritize what we wanted

other times we got Partners to say go ahead and edit our code base for us and so we we did manage to launch with a

compelling product uh it wasn’t until later and we got some of the things like the cloud functions. net URLs which are

like a an implementation detail core to our like really easy to use API uh

product um but we held back for GA for that and it really kind of customers

understood that like the beta product it’s cool you can use it but it may be featuring complete GA product here we go

and you can really see that in our adoption curve so with Firebase you’ve built some really cool products and

features some of them are extremely widely used across um things like Firebase hosting and Cloud functions um

what what’s it been you know what’s your process for building some of these uh super scalable High impactful

products uh so when I on cloud functions for Firebase uh or I currently work on cloud functions for Firebase which was

originally much tighter collaboration between cloud and Firebase like I actually edited their code I did an on

call rotation or two and in fact uh Firebase invented the event pipeline

that is now known as event Arc um but Cloud took over event Arc um and we’ve

really kind of separated responsibilities so right now we just do tooling and sdks it’s more of a people

scaling problem because you know I managed a portfolio of multiple products and anytime you put someone on cloud

functions you’re taking them away from other efforts um with hosting when I I took over the team it was it was running

pretty smoothly um increased scale was mostly a matter of allocating resources I will say that there is one component

that’s starting to explode but I can’t go into detail on what it is and we’re re architecting it and stuff um but uh

Firebase app hosting was really like it’s a brand new new product we had to be careful not to make brand new

mistakes and luckily Google really does have institutional knowledge and tooling

to uh make scaling easier uh there’s actually a joke internally that Google makes the immensely hard moderately

complex at the cost of making the immensely simple moderately complex um so we saved a lot of trouble

by building uh like on top of these building blocks our internal code that

runs inside we call Borg is built on SRE design platforms that really solve

scaling problems for you um and also one of our our selling points is that we’re

built on Google Cloud so so many of the resources are actually in the customer project we manage resource for you and

by leaning on those other products we’re leaning on on teams who have already had their fire drills they’ve already

managed their scaling problems so we really didn’t have those types of fires this time around um can you tell us a

little bit more about one of the products from Firebase called extensions um from what I understand it’s a really popular aspect of your product um and

I’d love to hear more about like some of the craziest um or creative usage usage of

Firebase extensions um just think about parse I remember the one that really stood out it wasn’t a crazy one it

wasn’t particularly exciting but it was called yo it was an app where you literally sent a message to someone else

that said oh yes yo I remember yo and I remember the scram Rambles when they had

uh not used passwords anywhere and like I had to like do editing in our API

server to like invalidate everyone’s blank password because I think like Elon Musk got docks from it or something it

was it was such a Calamity it was oh man you’re bringing me back um yeah so I

mean extensions aren’t apps uh or or I I like to say that they’re apps for your Cloud not for your users they’re like

click I have functionality fully installed in my back end and two really good ones come to mind uh one that we

wrote uh copies data from cloud fire store into big query so Cloud fire store

is a really great horizontally scaled real-time database but it’s no SQL which

isn’t always ideal for like business intelligence style queries meanwhile big

query isn’t really designed for you know end user applications but it’s great at

answering those business intelligence questions so they extension to replicate the data across the two helped people

solve each use case with the appropriate tool um the other one uh was actually

end user driven uh so when we announced the deprecation of Firebase Dynamic links a user basically came up with a

fully functioning system an open source tool that you could just install in your back end and get much of the same

functionality without having to rely on the Firebase product and this this was our dream we wanted to grow Firebase

beyond the Firebase team and extensions was a really good way of doing doing that have there been um parts of the

project that you’ve been working on where you felt like you were totally surprised or like whoa actually how did

we do that or wow we really did that that’s Way Beyond what we were expecting h i I’ve got to say it’s

Firebase app hosting we just released this at Google IO this last year but it was years in the making and it was a

massive undertaking both technologically and politically to organize so many teams to create one TurnKey experience

um I was both a tech lead and a manager at the time and so I had to build it while also balancing the needs of our

other portfolio products like Cloud functions for Firebase the original Firebase hosting cloud storage for

Firebase and yet we launched on schedule although some features were

dropped um and this was really important because there was no full stack web developer experience at Google before

this and it was a pretty obvious Gap in Google’s portfolio uh and unlike the original Firebase hosting this is

vanilla Cloud terms of service so it can be used by Enterprise customers uh and we’re really just getting started I

literally have a document called Firebase app hosting fiveyear strategy

uh there’s just so much more this product will do uh but seeing the beginning where it’s now in customers

hands has just been a dream come true sounds amazing uh so like let’s move a

little bit from you know some of the projects to more leadership and management um seems like it’s it’s been

a an great big part of your your journey to go from Tech lead to manager um and I

think from what I’ve seen in uh personally managing people can be quite a different experience so I’d love to

know a little bit about you know what your experience has been like there what have you

learned yeah what was really different was the number of people in my influence um so directly I managed 11 people I had

five indirect reports and we had countless numbers of Partners um um I I

really had to learn to document my thoughts better I learned how to delegate more uh to grow leaders who

could take on some of the burden as the or grew um you know a tech League grows new products a manager grows new people

who grows new products um and I I was both the manager and Tech lead and as my or grew I had to

just get comfortable giving up some more of that technical leadership uh I just eventually I didn’t even have time to

focus on the technical Road M um so which is why like even though I had a lot of success as a manager I recently

decided I had to choose either between leading technically or people managing uh because I’m a perfectionist and doing

both required way too much overtime and I worried it would endanger my relationship with my now wife which

obviously you know that’s Priority One um so after speaking with my manager

we agreed that my tech skills would be harder the two replace so we expanded my technical scope but handed off

management to a new person uh Julie comes from the Chrome team so I’m really excited to see how her expertise in web

Technologies like helps benefit our portfolio I I love that you mentioned

that you know you’re worried about the level of perfectionism and how it could impact your wife um I’ve I’ve been

married 16 years and there are so many things that could get in the way that you have to actively choose to not let

get there so you you’re recently married congrats it’s incredible it’s amazing uh

what’s it been like you know you’ve had to shift some priorities maybe you’re thinking about life a little differently love to know more about that and then

also sort of balancing things uh in in the the career of tech yeah um you know wedding prep is

hard uh we did a lot of work prepping for the wedding uh enough that I actually took six months off of teaching

scuba and it really reiterated that lesson I just can’t do it all uh um it

was really stressful I was doing a lot of overtime to get Firebase app posting out the door so once it launched I had

that heart tohe heart with my manager and I I think my new role as Uber Tech lead for firebases serverless portfolio

is really going to give me the freedom to chase after my priorities I couldn’t fund before as a manager uh and this

means I’m going to be in a better mood when I’m home with my wife it also means I don’t have the same sort of like

quarterly crunches for performance reviews and check-ins which which gives me a more consistent work life balance

and so my focus is finding meaning in my work but no longer defining Myself by it

that’s beautiful um have you found that your approach here to leadership has

changed you know I I don’t mean for how you approach your own work your own output but how about your team and how

you lead them yeah I mean as I I just mentioned I’m I’m shifting back from people

management to purely technical management which you know as you grow in your career there is a lot of overlap I still coach people I obviously have to

get Buy in from people and get them aligned um you know for anyone who hasn’t read it yet I highly recommend

they read our mutual friend and former colleague charity Major’s essay on what she calls the engineer manager pendulum

good technical skills and people skills help both individual contributors and manager roles you know individual

contributors need soft skills to collaborate and align with partners and managers benefit from technical skills

they can be good coaches to their teammate uh so getting married was a component of what made me pivot but the

whole journey of forming a team by combining others getting them aligned bringing them up to speed on the road

map I had worked out previously as an IC and then frankly letting them just blow

past me with their own their own planning was a huge point of growth for me uh I know delegate a lot more because

it’s both better for my sanity and their growth um but I keep a literal document of my priorities in order uh I also have

a combine on my whiteboard with things like overdue do today tomorrow next seven days next few weeks in progress

blocked delegated and just to feel good completed uh it’s important to decide uh

when you’re doing these things what you’ve completely handed off but need to check in on what you’re advising on what

you’re collaborating on and what you’re driving yeah and just just to go back to charity for a minute charity major is

the CTO of a company called honeycomb and um she and along with another one of

our former co-workers uh Christin Yen lead that company it’s a super cool product so uh Thomas you mentioned

earlier that you took some time off from scuba diving uh excuse me scuba instructing I guess you’re also diving of course which is cool um and you’re a

scuba instructor H you can tell us a little bit more about your expertise and level there but what kind of got you

into scuba diving in the first place and and do you use it as a as a release to

kind of get away from Tech or how do you feel and think about it yeah um you know

this’s a a long path to how I got into it but what I really value about it now

it’s what I call Active meditation now even your breath affects your depth so you have to be completely immersed in

the moment which really is good for disconnecting from all that stress that’s above the water um and I I find

that teaching is kind of like a microcosm of management you see people grow from like being these timid but

eager candidates to certified in two weeks you know you transition from The Challenge or frustration of teaching a

new skill to the joy of Basking in their achievement at a really rapid pace and it reminds you that there’s joy when

people achieve their goals which helps me stay positive when work might be in a rough patch uh also In Scuba people’s

lives are literally in my hands and it’s a very different kind of stress especially when teaching children

uh so it helped me remember that no one is going to die if I slip on my my

commitments uh which really helps me put things into perspective um but in my now defunct blog I talk about some of the leadership

skills that do translate uh so some examples uh is that I learned the trick that rather than

correcting a student when they make a mistake setting up their gear uh I actually critique their buddy for not

catching it and helping them so this is important in scuba because I won’t always be there and they need good buddy

etiquette to help keep each other safe uh but in my professional life coaching the coach means that fewer people block

on me and my scope of influence can grow uh another lesson I learned is that it’s

much more effective to point out what people should do rather than telling them what they’re doing wrong there’s a

thousand things you can do wrong and if you just say this is wrong you’re not helping guide people to

success I remember about six months ago I was scuba diving I think I I mentioned this to you but I was scuba diving in

Hawaii and the uh path that we decided to take this day was to go through like

six or seven caves I had never been through a cave before I was completely freaked out the entire time and I

remember at one point I felt like the shark that came out of the cave was like it had to it had to been 30 feet long

and immed immediately I took the deepest breath and I I I hit the roof of the

cave and I freaked out and my dive instructor was like hey just just chill

he helped me get my breath back helped me kind of navigate it but I love your

idea or your comment there about staying in the moment the breath work that’s involved because those things if you can

stay focused on where you are rather than where you know you could be it really brings you back and kind of keeps

you grounded um yeah I say also uh I I don’t know if

I’ve ever mentioned this but I’ve I’ve picked up scuba photography and so that’s another level of focus you

know you’re in the moment you’re staying in the detail you’re staying in your Zen but you also have to somehow scan the

world and find that little tiny creature that would make a good photograph oh wow

I think there’s too much going on there for me um so you know you’ve you’ve worked

with some incredible companies you you’ve gotten married which is incredible you’re a staff instructor on

on scuba um when you look back at things professionally personally what have been

some of the biggest toughest scariest challenges that you faced we we’ll keep it technical for now

uh so I think some of my my craziest bugs were at Microsoft um so I actually

did uh some innovation in the zip file format uh and for legal reasons I had to

build it on my own not talking to anyone else with only a redacted copy of the RFC on how to zip a file um and one of

the memorable bugs only happened on AMD processors and not Intel processors and

it was because of an edge case on how they handled a particular assembly instruction that was just the weirdest

bug uh another monster was when I was uh helping diagnose an assertion failure in

Internet Explorer uh IE predates a lot of coding

best practices uh so this was like thousands of recursive calls deep in a

function that was thousands of lines long with go-tos in both directions uh luckily we had what we

call a time traveling debugger which was absolutely critical of find the bug uh but nowadays my my problems are much

more organizational uh when we first launched Cloud functions for Firebase I think we had like eight launch Partners

we had to huddle weekly to make sure we got status request or status updates on every single partner figure out who was

having difficulties figure out how to unblock them uh and we managed to launch on time with everyone there um and with

Firebase app hosting um it was really unique because so much of Google is

designed for rpcs not for web hosting and so we had to come to a lot of people and ask them for things that they’re

just not used to doing and to prioritize a product that hadn’t even been built yet um but luckily we were able to to

convince everyone of our vision uh people were either able to give us what we wanted at launch time or they’re

actively build it so I have a lot of faith in our success um you mentioned earlier that

you’re transitioning to this thing called an Uber Tech lead um first you know I’d love to know a little bit more

you know what that title means and then second um what are some of the upcoming projects that you’ve got going on in

your life professionally personally yeah so an Uber Tech lead is basically a mid-level manager of tech

leads so a tech lead of tech lead so there’s people who figure out some of the day-to-day problems I work about

things like um long-term road maps some of the more thorny problems um things

that affect an entire product portfolio making sure that stuff works

better together um and so it really like it’s it’s brand new for me it’s it’s just started this month uh so I’m still

learning a lot um I need to learn what influence means now now that like I’m

not a manager I can’t actually assign the resource to work on what I think needs to be worked on um and it also

means that as my portfolio grew it’s like I’m at a new company I have a lot more to learn I need to learn more about

gen kit I need to learn more about extensions uh and figure out how I can make all of our products work better

together um so my current focus is still on app hosting I want to be the best

place where you can host angular and nextjs apps where possible uh I want to build open source tooling I really

believe in the open web and trying to you know we can compete on the merits of our product we don’t need to lock you in

um you should be able to host Firebase on nextjs or angular wherever you want though on the other hand we do have this

really cool novel infrastructure and we shouldn’t shy away from leveraging it for really cool benefits either um on a

personal level uh my next big project is designing the welcome cards for the wedding uh and then after that my wife

and I have talked about how we just suddenly have this big gap in all of our you know the time that had previously put to wedding planning and we want to

be intentional with how we fill it rather than just like collapsing exhaused from work and watching Netflix

um so uh one of the more responsible things I could do would be to learn farsy so I can speak to my in-laws when

they switch over uh but we’ll see Ah that’s great I I actually have an home automation built to prevent me in my

wife from doing too much of this exhausted from working watching Netflix forever at nine every night my TV

automatically turns off Alexa says it’s time to go to bed and boom we’re out of there um I think I think it you know

thinking about it as a newlywed but also somebody who’s been married for a while just having time that you actually talk

with your wife and do things with them that sounds beautiful so congrats on thinking about it that way um Okay so

we’ve got uh current Tom who I’m not saying is old by any means but let’s

let’s Look Backwards at the the younger Tom you know um if you could look back

and and actually just reach out and talk with him what would you what would you tell him what would be some of your your

advice uh so this is advice I actually got in college so I’m kind of cheating but it’s just never sto learning um the

industry moves so fast and everything you know or do can be out ofd faster than you realize it um this was a major

motivator for me to join parse which has been one of the best moves of my career uh and now I’m studying AIML uh I’m

trying to do both like the highle stuff learning how to use these agents um you

know prompt engineering that sort of thing uh but also understand the underpinnings of it the linear algebra

perspective amazing so let’s let’s do this I want to look a little bit further

down the line and talk about the future I’d love to know what you think the future holds um maybe we could talk a

little bit about AI since it’s sort of right now at the Forefront it’s on everyone’s tongues uh so I mean one of the things

that like drives me nuts AI is not new generative AI is new um but when we when

we approach generative AI I think of it in a few lenses so in Google’s case we

offer a direct AI product we sprinkle it into our SAS products like Google workspaces uh within Firebase in some

cases our customers use AI we need to give them the best tools to leverage it in their own applications or we can use

AI to accelerate our customers development so you think of like GitHub Co co-pilot or Firebase has a new Gemini

plugin that will help you design and code your app uh and then finally I think AI is GNA help operate businesses

so for us backend Engineers uh maybe AI can help identify and correlate anomalies to address or even maybe

mitigate uh outages I look forward to the day when AI can read my architecture

document and tell me I have a dumb idea um and then uh as we have you know

new manager on my team uh they’ve been using AI to summarize you know three

years of architecture diagrams and brainstorming sessions to kind of bootstrap onto our product uh and you

know your world which is very different from mine uh I’m sure AI has really changed the way that you do things like

sales and marketing like tools like Salesforce have helped coordinate sales and customer management for decades but

I assume they and competitors are really upping their game with the ability of AI to give you a bird’s eye view to tell

you where you need to put Focus uh and potentially even using agents or other

automation to expand your reach with fewer headcount yeah you’re absolutely right I I talked pretty recently with

the CEO of a company called gainsight and they do customer Success software and he mentions that he loves the idea

of a customer success manager sitting next to you know sort of sitting next to an AI agent who helps keep track and be

more aware of all the accounts they’re working with see some data that changes something that’s indicative of maybe a churn or a contraction and just just

helping all these people do more be more effective without feeling as much stress about what are we

missing um absolutely so so in the future uh you know there’s all these you

mentioned AI but obviously there’s going to be some emerging Trends technologies that maybe we can start to see now that

you might be already excited about anything that you’re you’re starting to see feel or or look towards that you

think will shape the industry in the next few years I’m not sure if this is a Debbie Downer or hopeful response um but

I do think that AI or generative AI may be approaching Gartner’s plateau of

unrealistic expectations I’m personally bearish on uh the idea of artificial

general intelligence I think that like so many advancements have come from a single white paper all you need is

attention and there isn’t that next White paper yet that will lead to AGI uh

but frankly like the genie is out of the bottle we are going to have a world that will never be the same um I think you

know a lot of times we hear about like well will AI take people’s jobs but you know there’s been so many other

technological advancements in the past from the printing press to agricultural improvements the obsoleted old jobs and

yet Society found new ways to employ those people and expand into a more rich

and cult diverse society and so I I do wonder sometimes uh you know maybe this

is a little off topic but like what will the post AI Society be like you know is it really Doom Gloom and mass

unemployment or will this just be an explosion of new careers that today we can’t even imagine um you know I hope we

get more than just prompt engineering as a new profession yeah that’s that’s kind of where I’m at too I I currently feel

like I’m a prompt engineer writing better doing more but I

think uh there there’s a movie called The Incredibles where there’s a little boy who looks at the main character as

he’s dressed up and I think the little boy says something like or the the man the Mr Incredible he basically says like

what were you expecting and he says something incredible and sometimes I feel that way about AI like I’m hopeful

that we’ll see more and also I I I’m kind of like you I I don’t think that the future is going to be really Doom

and Gloom um I just see that more people people will be able to do more faster I

I I prefer to see as opposed to replacement it’s more of a supplement you become like Iron Man You’re a human

within a mega suit I there’s a great saying I saw online uh AI is not going to take your

job someone using AI is I think that’s the other big part of

of AI that’s actually pretty problem problematic in this goes back to one of the things you said earlier which is you

know never stop learning there’s a lot of people who are unwilling to try new tools new things and AI is just one of

those things you you just have to try it you have to see how it works you have to understand it you do have to become a bit of a prompt engineer and as you do

that obviously you can develop a skill but like you said if if you don’t you can and will be potentially replaced by

those types of people yeah absolutely it’s just it’s such a force multiplier

that you have to adapt or you’re going to be left behind yeah yeah for sure uh so so

you’ve built all these systems with you mentioned earlier pedabytes have data which you know looking at you know how

how much my computer holds 250 gigabytes of of data petabyte is gigantic um and

now the world’s talking about AI you’re talking about Cloud infrastructure Edge Computing all these really amazing things uh what what is the the next big

thing that you’re excited about Beyond AI for example uh so I I I prepped for more of

like a a Nuance in each of these sections uh so you know Edge Computing

has been something as close to my heart as I’ve been focused on app hosting uh a lot of companies told you that the edge

was the future and we internally had a hard time because it’s it’s unprofessional to publicly argue with a

competitor statements even though we thought that the promise was overstated

uh people eventually realized what we knew which is that edge compute without Edge data is a really bad idea um and

frankly Edge data is either partitioned across the edge where it’s eventually consistent and people just aren’t used

to thinking about that yet so yeah let me just ask you one quick clarifying

question for those who may not know or understand Edge Computing talk a little bit more about what that means so this

is uh so depending on what level you want me to go to uh just kind of super

high level typically so CDN content delivery networks uh most of the time your website doesn’t actually uh Reach

the the central server you know we generate a page it’s the login page it’s the Ikea page for some stool whatever

it’s been generated once and it’s it’s stored on the edge and served and that

you know you might have one location in the world in Iowa for example that you

generate new content but then you have hundreds of machines out there that

store you know what Scott visited for When I visit it remembers and it it stores but we’ve started evolving with

what we call Edge Computing where that machine is no longer just dumb and what

do you call it cash it can actually be programmed um and I think there’s a lot of really cool Innovations there you

know some of the areas of my focus are looking at how that caching can be augmented

for um you know experimentation or whatever there’s a lot of things that would break the cash that I think a

smarter cash can handle um and then some some people have put

authorization other things inside that CDN that way you can maybe fail a

response more quickly or without having to go to the more expensive Computing layers but for a while everyone’s saying

there’s going to never be any more Central Computer we’re gonna always program at the edge and it’s like well

you had one computer before and now you have 120 computers that are all talking and they can’t wait on each other

because then it’s not fast anymore and that programming model is just so fundamentally different and there’s

there’s two things that are going to happen here one people are going to realize when they got overzealous and it

wasn’t actually the right solution uh and I think here you’re going to see a couple of different layers of going from

the very edge to the very Middle with some steps in between um so for example

with Firebase hosting the original product uh we kind of have three layers there’s the CDN layer where you can cash

content maybe in Iowa but we also have a

middle layer where if someone goes to Europe to download a page the image is in Europe already even if it’s not in

the CDN and then finally if they need Dynamic content it goes back to Iowa and

so I think we’re going to see topologies emerge where you can program with

different levels of complexity because it is very complex to program at the edge um to benefit with you know the

right amount of benefit for the right amount of cost then separately where the edge really is uh appropriate I think

that we’re going to see people learning new types of Technologies like dealing with eventually consistent data or

designing their data pattern so it can be totally partitioned so that you don’t need you know your Mumbai customers data

in the London CDM uh here I’ve been monitoring a couple of things uh there’s a movement called Cloud native where

people are really rallying behind taking independent projects and consolidating them into new standards and then those

various companies can conform to the standard and this has really defragmented the industry so that you

have the flexibility to program to a standard and try different vendors it’s really given flexibility and Agility to

people um the next is I I think I’m starting to see a pendulum swing away

from microservices back partway towards monoliths you know not not the true big

you know 10 gigabyte monolith of yester year but you know sometimes people got

so overly zealous with microservices that they call them Nano Services you know they’re they’re building sand

castles instead of Lego bricks and I I really do think that um some of the New

Primitives are going to help you find a balance between these two extremes

um so this is one of the reasons why I’m really excited about some of our products that cloud run so because uh

you can run any container uh with event Arc you can have any number of triggers what used to be 20 functions can be one

service you still get all those serverless benefits high level programming um with you know Auto

scaling scale to zero pricing but through consolidation you can mitigate some of the downsides like cold

starts um you know and i’ I’d be a little remiss if I into our own horn Firebase app hosting is built on cloud

run uh which you know because it’s container based feels a little bit like infrastructure as a service not platform

as a service um and so you get to have concurrent executions and you pay for

container seconds not request seconds so this is a bit more of a lower level concept but in exchange we’ve seen from

multiple customers that this leads to significant cost savings um I’ve done some back in the napkin math and I’ve

seen that Firebase app hosting can easily be up to 10 times cheaper than some of the competitors at high volume

workloads thanks to this one competitive advantage and then finally you mentioned

AI here I think we’re just scratching the surface you know right now because

of that all you need is attention white paper everyone is creating chat Bots and image generators but we’re starting to

mature uh rag is helping reduce hallucinations uh image generation is

getting significantly better with the ability to fine on previous queries uh we’re seeing video generation and

recognition as some of the Cutting Edge Innovation uh and then I’ve noticed that Gathering and summarizing information is

providing just as much or even more utility than just merely splitting or

spinning out fluff with an agent um so I I see in our Portfolio

Services like gen kit really being a parallel to what we did together at parse uh by abstracting infrastructure

rure over multiple providers you can develop more quickly and with much more agility because you can swap out

components rather than comp coding directly Against The Primitives um so we hope that our open- source work on

pipelines pluggable extensions tunable parameters you know a local developer story with tracing is going to really

help hand like hold the hands of developers as they Pioneer the evolution of this

industry I certainly am using AI to generate and create lot lots of images

uh and very simple questions so far but I I’m also excited to see it see it mature um one of the things that you

said that I thought was really cool was you know talking about building with sand castles not with Lego bricks I think my my whole floor and my home is

just covered with Lego bricks but you know they they work and they help you build things the right way um and sand

castles they disappear U there’s there’s places in the outer banks in North Carolina where the homes that were built

on the sand the homes now tower 20 to 30 feet over the sand in the water because

the sand is just sort of gone away it’s not built for for time and it’s not built for the scale that you need so I

think that’s a really appropriate um example there uh it’s it’s been awesome having you it’s been a pleasure to have you on our podcast and just wanted to

say thanks for joining us today it was great to speak to you again