How I Deploy ALL My Apps - My 2023 Stack

My 2023 Stack's infra is pretty dope. Be it hosting, compute, servers, serverless, databases, crons, events, logging, and pretty much everything else, there's a solution in here for you 🙏
Blog post with EVERYTHING: t3.gg/blog/post/2023-infra
Vercel - vercel.com/?ref=theo
PlanetScale - planetscale.com/?ref=theo
Axiom - axiom.co/?ref=theo
Upstash - upstash.com/?...
Clerk - clerk.com/?
2023 Stack Tech - • I Ship This Tech EVERY...
2023 Stack Tools - • Tools I Use Every Day ...
ALL MY VIDEOS ARE POSTED EARLY ON PATREON / t3dotgg
Everything else (Twitch, Twitter, Discord & my blog): t3.gg/links
S/O Ph4seOne for the awesome edit 🙏

[@t3dotgg]

CHECK OUT THE FULL LIST ON MY BLOG t3.gg/blog/post/2023-infra

[@karanmungra5630]

You can see on theo's face, how happy he is using this tools and talk about them with us.

[@TheSaintsVEVO]

Huh he looks like he’s at gunpoint

[@protoxpire0]

Not being a doubter or unnecessarily skeptical. I appreciate the positives you brought up for all these techs. Could you also talk about scenarios you wouldnt use them for and some potential downsides of these new companies? Also while using these tech to build a side project may be cool and a good learning experience, would you say they are also ideal for a larger scale project? Especially when we consider how fast tech moves, would you question longevity or replaceability of these companies? Hope its not a dumb question. Thanks

[@bluengreen71]

If you have to do anything requiring a BAA for HIPAA compliance you're most likely gonna still have to be in AWS. Vercel says they are working on it. It would be a major game changer for the health care industry to enable devs to be compliant if they did.

[@thomblank2915]

For anything on a larger scale or if you’re working a job, you’ll probably still need to be on AWS/GCP/MS for one reason or another. The companies and tools he talks about are for small scale side projects and/or MVPs.

[@victormendoza3295]

@@thomblank2915 I agree with you, but what would you say the user count / load is to leave those products?

[@roeelcohen]

I can think of few down sides on top of my head but they are just my opinion (and shouldn't be taken as a professional opinion) :) First - scalability and resiliency : IDK all of the mentioned services but on AWS for example those goes without saying ($$) Second - Latency: when you're using a specific cloud provider (AWS/GCP/etc) the latency between your services should potentially be a lot smaller, especially if you are deploying your services on a specific region. Third - I would say that once you tweaked enough some of the services (like Kafka K8s etc) adding other services becomes "easy" and they usually play nice together. Lastly - business perspective: If you are a company (and a big one) you would usually struck a deal with a cloud provider for all of their services and not pay for many 3rd party companies and services you use god knows what, also companies that charges money for their services usually don't want to "f around" and would rather to rely on the big boys and not new companies that might fail at some point. have that said, usually big companies will have a team of dev-ops to take care of most of the hard work.

[@noccer]

Always great content but for the love of Jeebus turn down your gain, don't really wanna hear your mouth sloshing into the mic 😄 ☘️

[@t3dotgg]

This was the 2nd of the 3 I filmed, hadn't realized how much I fucked up gain levels until after 😅

[@xX_dash_Xx]

I like the mouth sloshes, something something asmr

[@SoggyBagelz]

Nah love that tbh

[@albertosilvaus6417]

This content you're sharing is invaluable, thanks!

[@caiolins2495]

Man, the way you sold Vercel in this was just a banger 💥 I'm having the opportunity to propose a new if not THE stack the company I work for will be using in its next projects and I'm proposing a lot of these specially Vercel. Thank you and please keep on making these 🚀

[@ricardoamendoeira5689]

I would love to know more about that 7 minute response time for bug fixing, it sounds nuts to me, would like to know more about what an average bug at Ping looks like and how the process works for notifying an engineer, how they debug so quickly, etc.

[@OryginTech]

Not that hard really… I work on a pretty popular productivity tool, bug fixes take about 5-10mins from reporting

[@berrywarmer11]

Yeah! What if ol' Bobby's on the Porcelain Throne? How do you get your required two approvers then? Does Bobby have a 15 second SLA to review code on a toilet?

[@ricardoamendoeira5689]

@@OryginTech Can you give a summary of how that works? Are you supposed to drop whatever you're doing immediately when a bug report comes in? There isn't even an evalutation of the impact of the bug? And are the reports really that detailed that you can understand them in 1 minute? And if the vast majority are so quick to fix doesn't that mean that many trivial bugs are being pushed to production that? I have so many questions about this, it would be really nice to know more.

[@skdamico13]

This mixed with no testing is just a terrible recipe for burnout

[@MickGardner-vc4us]

yus I'm early!!!! gonna try to venture into frontend this year with your blessings Theo

[@RodrigoMallmann1]

Hey theo, hope you see comments on older videos. As a sole developer for a reasonable scale company (ecommerce) I use GCP and they have most things I need with really good dev experience for me. I dont really need to run vms and stuff. I run most of my things on cloud run containers and functions. Its well integrated with databases and has logging for basically all services. Disregarding the price, which goes up to about 500 USD for our production environment and about 200 for our Dev, its really cheap for a company that makes six digits in revenue monthly. Would you care to give your input in how GCP fairs against T3? I feel like learning it to have it in my options, but I don’t really have an idea of how much would I be gaining, except from beign on top of the hype

[@anasouardini]

[just curious] What about automating AWS and keeping the low price?

[@RapBatLe]

Thnak you Theo, nice video. I would like to know more about Upstash services (Redis, CRONs, event queues, kafka, rate limiting) when and how to use them, maybe you can share some your experience of using this services or some useful examples.

[@colbyberger1881]

Great video, love the new tech. I'm still a k8s fan though.

[@anmolnoor7373]

Love your work Bro, Keep it up... I was looking for this type of video for my Infra and there you have it. Thank you very much.

[@zappist751]

W MANNNNNN I LOVE TTHIS VIDEO

[@stephenbrown-bourne465]

Really like your shirt in this vid (the 1stone)! Where did you get it? Great vid too 😆

[@shaunmodipane1]

Thanks for the content, i dont understand everything but i will soon.

[@lescobrandon2202]

Lol, this dude has the best thumbnails

[@rodrigoacuna5884]

What do you use for file storage instead of S3? Images, videos, etc.

[@bandit8977]

Agree with this 100%

[@Ropbastos]

What do you recommend for sending push notifications?

[@SimonCoulton]

Vercel is currently not HIPAA compliant, really shuts things out for those of us in the healthcare space.

[@BarisPalabiyik]

AWS Amplify is ?

[@paul1]

Fly is a really nice hipaa compliant alternative

[@mctrials23]

@@justin.johnson Thats putting it mildly. A build that was taking 15 minutes on Amplify takes about 2 minutes on Vercel. Astonishingly poor from AWS.

[@hughmungusbungusfungus4618]

What would you recommend for a data warehousing solution that could feed into a BI platform?

[@thundergabriel]

Any startup use the whole bandwith of vercel ? . I know one that pay 5k for bandwidth width no more than 1000 users

[@clockworkop]

Well, this video has made me realise how far behind our school is in this way. I don’t blame them for it, all of this seems quite bleeding edge, so it doesn’t really fit to teach it. But anyway it is still quite a leap to go from taking my old computer and running linux server with php on it, to serverless functions, running who knows where hosting a full stack typescript app, that communicates over trpc. Is there like any book or guide or course or something I could read to catch up with that, because it’s quite a lot to take in.

[@MyPhuckDub]

These stacks are nice, but most companies will use something other than Vercel, Planetscale etc. You have to learn AWS, GCP, or something else; Vercel costs at least 6 times more than any of these platforms.

[@berrywarmer11]

Don't worry about it so much buddy! The gist of the concepts between these wildly different technologies are still more or less the same. Take your current school projects seriously with the techs that aren't so great, and I guarantee you won't have any trouble picking up the new stack when you get a job in the industry. Story time! I was in your boat fiveish years ago when I was doing a web development class learning JQuery & Bootstrap on the frontend and an older version of NodeJS & Express on the backend. Back then, I was just happy to learn how to build a full stack web app. Shortly afterward, I graduated and got an internship where the company had me build a simple replica of one of their apps in React + TS + MaterialUI on the FE and Java + Spring Boot + Postgres on the backend. I was salty at the time that the techs I learned in school were out of date, but now I understand I still was lucky to work with them. I found a lot of the new techs came quickly to me; the other intern I was paired up with was struggling since she didn't have the prior level of experience I did. The little project was a success, and the company decided to hire me full time and work on one of their apps. "Clockwork op": don't be too worried about working with older tech stacks. If anything, they're probably better to learn with since they don't hide as much of the problem from you like modern tools tend to do. I wish the best of the luck as you're breaking in to the industry!

[@BosonCollider]

The PHP stack with a DB on the same machine will likely scale much further (in the sense of customers served for a given expense) than the setup he mentioned, and with a tiny bit of HTMX added it will be much faster to develop in as long as you have a good VM.

[@user-pq8lo1eu2j]

I usually really like your videos. This one just feels like a commercial

[@maquinaparlante5257]

Early gang let's gooooo

[@GumowyKaczor17]

Let’s gooooo!

[@samuelebarbiera1075]

What about a full project tutorial with the t3 stack and this edge stack???

[@vbstreetz]

I still prefer aws especially automated infra using cloud formation.

[@skdamico13]

Yep and with AWS SAM or Serverless it’s dead easy once you learn it. Plus infra diffs are in the repo, then when you inevitably need control, you have it. Also if you get to a moderate scale (traffic, compute, whatever) on some of these PaaS, your burn rate ends up being 2x AWS without the flexibility in networking, etc. This all depends on the business/usage/etc so it’s really hard to recommend something to everyone. It’s a damn luxury that we have all these great options.

[@PausePlayRepeat]

What is the difference between NextAuth and Clerk?

[@agustinaon]

Great recommendations, as always! Love all the apps and have used most of them personally. Unfortunately the pricing is too steep sometimes for a small team that doesn't (yet) produce money. Not complaining about paying for them as they DEFINITELY are worth every penny, but they do stack up when you start adding them. Do you have any alternatives that you've tried that could fit the bill for a small team? ]

[@t3dotgg]

All of the services I recommended have insanely generous free tiers

[@brianfarley2723]

Amazing.

[@zyzlol]

aws amplify provides a similar experience to vercel. lambda is not a good comparison

[@ilovenaturesound5123]

I’ve been using Railway to deploy my full stack project. This includes express server, Postgres, redis for the backend and React for the frontend. The experience is pretty nice but not much customization is possible.

[@lutfiikbalmajid]

I can't pause my db with railway.

[@lutfiikbalmajid]

Sorry i mean, the project

[@ilovenaturesound5123]

@@lutfiikbalmajid I think you can cancel the deployment, no?

[@ouss]

@@ilovenaturesound5123 no you have to delete the db separately from the project hence many people forget and have a big bill at the end of the month

[@Anteater_Taming_Masterclasses]

Audio is low. Love the content :)

[@stephfh]

Woah, for once I'm part of the early crowd.

[@leonardojaques6586]

So will Clerk be on the create.t3 as well?

[@mavenwander2374]

Clerk is great! Any plans on adding it to create-t3?

[@james-perkins]

i have a t3-minimal-clerk that is t3 minus next auth. i don’t expect it will be added to create-t3. it’s also referenced in the Clerk docs

[@user-kt3ub9ce8v]

What do you use for images and videos ?

[@whatplan4335]

I’m so split on nextauth vs clerk. Clerk obviously has some nice QOL stuff but I can’t decide if it’s worth $0.05 a user when I scale. Maybe I’m completely wrong though and $0.05 is nothing compared to the time reimplementating everything they do. Anyone have advice?

[@joepetrillo6185]

Clerk all the way

[@eafadeev]

Isn't it better to colocate everything on the same network and not having to encrypt traffic between the nodes? If all those pieces are hosted on different companies you'll necessarily have to put all traffic on tls, plus have extra communication delays. In addition, with using disparate services multiple points of failure would be introduced, which is hardly a good thing.

[@hadavid2454]

Would planetscale scale better than mongoDB since mongoDB is noSQL?

[@krazymeanie]

You forgot to put clerk in the description 🙌

[@EddyVinck]

Me: literally on my way to my AWS Developer Associate exam KZbin: the infra that saved me from AWS

[@EddyVinck]

I passed 🥳

[@juanmanuelyounghoyos7350]

@@EddyVinck Congrats! 🥳

[@mnik0128]

The Big Big vendor lock. Nothing that you actually own in your projects. Is that a problem?

[@flavourfulski]

You ever try just deploying aws with docker-compose to fargate? insanely easy.

[@Patrickdaawsome]

Very curious to hear more. Is it easy/possible to create multiple staging/QA env apps on a Fargate instance?

[@iamgauravbisht]

Hi sir, i am searching for some good and serverless database but planetScale has depricated its hoby plan and now we can't use it so whats its good alternative pls make a video on it or suggest me database name in this comment reply.

[@Leto2ndAtreides]

Heh. Good stuff all. Why is this video unlisted though?

[@a.perkemantus]

What to use for file storage?

[@weirdo3116]

thinking about the servers is actually fun (for me anyways). frontend code has become too much of a nightmare. also prefer AWS since i've run into too many things that won't like be change stuff the way I want to.

[@stephenzhang3825]

good video

[@BrentMalice]

anyone know if this is updated? trying to piece it together from other videos based on Vercel drama but im dumb too so

[@chiubaca]

thoughts on railway?

[@chrtravels]

I was looking into Clerk but it's pretty expensive. $25 for the hobby plan and 2 cents per additional. $99 for business and 5 cents per addtional. Those prices are fine but only includes 1000 users. I'm not sure how Clerk compares to a service like Supabase but with that service you get the database plus the Auth and 100,000 monthly active users included for $25 monthly. Unless I am misunderstanding Clerks pricing but that's how it looks to me. I have heard a number of KZbinrs say good thinks about it but if this pricing is right, you would be paying thousands of dollars a month for what you get in Supabase for $25. Please correct me if I am wrong here. Of course that is provided you are at max users, which is a big if.

[@riser9644]

Vercel vs netlify, which is best

[@MangoBroLadM8]

is that windows in the background?

[@rtorcato]

Cloudflare pages and their other services will be your 2024 stack.

[@schtauffen-plays]

What do you do for websockets?

[@developerpranav]

Why video unlisted?

[@smarkx1276]

good luck

[@bahaaalhalabi8940]

Vercel is just amazing to use, i absolutely hate aws, i've been usng it for 2 years because of my work needs. When deploying my own projects on vercel, things seem so so much better.

[@reydavid7300]

What apps do you have?

[@DarkzarichV2]

Tfw I'm still just manually deploying and managing everything using my VPS, everything there is in Docker. Am I old 😂

[@butterfly7562]

Railway can do most common things!

[@bob_kazamakis]

I prefer a stack built on sst. Makes life a lot easier to have my project and the IaC in one place and makes AWS much easier to deal with with high level constructs and local debugging of lambdas live.

[@mjerez6029]

Great stack overall! But u can't be talking about edge computing and redis in the same sentence

[@Hypawolf]

What would you recommend as an alternative to AWS S3?

[@JasonAtallah]

I’m so glad Heroku sabotaged itself otherwise I’d probably still be using it for all my side projects

[@martiananomaly]

Your words couldn't be any true. I'm so glad I was forcefully shifted to Vercel from Heroku for my hobby projects because damn it's so much better.

[@spickyoo]

What are you using for file storage? I tried setting up S3 (first time AWS user) and it’s a unbelievably bad experience

[@victormendoza3295]

What error were you getting?

[@abdelhakimkhabir]

I think that we should at least deploy one server from scratch to know how things works then go for other solutions, but not for beginners, i still didn't did that but I'm wanting to gain a little of experience in networking to make things a little easier.

[@gotxe]

Guessed every point, i'd pick the same. Maybe also Inngest

[@coffeeintocode]

Did I miss object storage solutions in this video? 😅

[@t3dotgg]

…okay I still use AWS for ONE thing. Someone please kill S3 😅😅😅

[@coffeeintocode]

@@t3dotgg it is beyond me how nobody has killed this fairly fundamental part yet. I got into a nice conversation with Chronark about it, Upstash are considering something but nothing more than that at this stage. For now, I use Digital Ocean just so I don’t have to keep an AWS account. Not perfect but it works for now….

[@milos018]

Cloudinary?

[@coffeeintocode]

@@milos018 it only supports images doesn’t it? If you have video, json, pdfs etc this isn’t an option?

[@milos018]

Cloudinary FAQ "As a complete media-management solution, Cloudinary can serve all file types as is, including images, videos, as well as audio, text and document files.“

[@DharmaScienceRadio]

This trips me out because I can tell how valuable this could be.... Yet I feel like I can hardly follow along with the sentences said.

[@danvilela]

I swear this is the fifth time I see the notification for this video.. Wtf is happening?

[@HamburgerTSMK]

Thank you for this video! My main pain point with these BaaS solutions is: how do I protect my endpoints? Do I need to protect these BaaS endpoints myself? If yes, how? It's always marketed as "super easy, set up in 10 seconds" (e.g. Supabase), but how good are they really at preventing spam (auth) registrations/DDoS attacks, and other security threats (like denial of wallet)? When I look at the documentation of these services, the topic of "security" seems to be non-existent. I guess a video that actually goes beyond the surface level and discusses the security implications of these BaaS solutions would be unique on KZbin (as everyone only discusses happy day scenarios) ;-)

[@t3dotgg]

- Vercel is provided by Cloudflare, and benefits from a lot of their DDOS protection - I literally recommended a ratelimiter (upstash) in this video - Edge compute is insanely cheap and one of the best ways to handle these types of attacks. NONE of these services are exposed publicly (I also don't recommend Supabase for a reason) Your "endpoints" are hosted on Vercel. They're trivial to protect. Next question?

[@alberjumper]

Is there any solution for billing? Maybe like Clerk is for auth?

[@emilsadek3621]

Stripe

[@muhammadzainabbasbaloch3200]

Stripe yes it is

[@martiananomaly]

I'm certain everyone who's used Vercel loves Vercel lol

[@noext7001]

planetescale is gargabe is free mode, host is always down

[@earthling_parth]

This feels like a weird TS/JS bubble of infrastructure.

[@namesare4fools]

you missing out Blob storage

[@CodeWithDean]

A lot of mouth noise going on…don’t forget to stay hydrated. I also recommend using a pop filter

[@echoptic775]

How can this be legit, cuz he probably only featured companies that payed him. Thats probably why he added clerk later, they might took longer to respond to the offer

[@t3dotgg]

If you've paid attention to the history of myself or my channel, it's pretty clear this is legit. I could have made way more money working with other brands. Feel free to pretend I'm lying if that makes you feel cool, I guess.

[@echoptic775]

​@@t3dotggmy thinking here was what if there were more great tools that didnt get featured cuz thry didnt want to pay you? For example like clerk was almost not featured. Was that the case or not?

[@t3dotgg]

​@@echoptic775 Clerk was always in the video, they got bumped into the T3 Deploy Partner program because I have been using them more. I tweeted about how much I was liking them, they hit me up to do another vid, and I decided to invite them to the program. They agreed within the day. I originally planned to announce them in ANOTHER vid, but since they moved so fast and I was so hyped, I decided to include them in this. I paid my editor for overtime because I was so excited.

[@dmnkb]

While I like your suggestions here, I'm always a little concerned about GDPR compliance. As an EU citizen, the sheer complexity of the data policy laws discourages me from using the services offered by (mostly) US-based companies. It would be great if you could address this issue 🙂

[@t3dotgg]

- Vercel does not store data. - Planetscale is a database. If it automatically deleted data based on ANY policy, it would be a bad database. - Upstash is a cache. If it automatically deleted data, it would also be bad - Axiom is a logging service. Don't log PII. - Clerk is gdpr compliant Your services should not magically delete data for compliance, regardless of where they are located.

[@diretitan9871]

“Amazing, every word of what you just said... was wrong.”

[@riser9644]

Arnt vercel and planet scale very costly

[@rickdg]

Enough about AWS/Cloudflare, let's talk about AWS/Cloudflare.

[@boatxd]

I appreciate this video but it would be nice if you mentioned some of the downsides of each of these services :)

[@niciusb]

I tried Planetscale but it was way too expensive, even with just a tiny load. I don’t know how anyone could afford it at scale

[@t3dotgg]

Planetscale is so cheap I’m making a video about it lol. 1 BILLION reads for $1, is that not “tiny”???

[@Proxus]

The audio track is out of sync with the video. Good video nonetheless.

[@wlockuz4467]

Its great to know what tools you prefer, but you should shed some light on what you actually ship with these otherwise this video comes off as "you can do anything with this stack"

[@ABHISHEK-jc8kn]

Reuploaded?

[@t3dotgg]

No, you just watched the early upload before it was public ;)

[@ThePieas]

what about the cost ??

[@SandraWantsCoke]

I've read opinions that Vercel is expensive af when you have high traffic. I have no experience with it whatsoever, it's what I've read.

[@chockman3833]

Google cloud run :)

[@HemstitchedIrony]

Maybe it's just because I used them first but I like Netlify more than Vercel, they do basically everything vercel does and often better

[@milos018]

I too use Netlify, very happy with what they provide, I do not think they lack anything, but tried Vercel as well and both are really good services, I don't think you can go wrong with either, and switching is pretty easy if need be

[@juliusmarminge]

Hiiiii

[@berrywarmer11]

underrated comment

[@stenhenriksson9306]

Most of these are just aws wrapped and added fees

[@sharannagarajan4089]

Looks like a commercial for a bunch of tools

[@RedStone576]

thats how partnership work

[@cristianarmandofloresalvar5966]

Let's be honest, this is a commercial of 7.5 minutes!!!

[@ThePandaGuitar]

Typical SaaS solution for every problem in JS land. I'm so glad I moved over to Laravel after 10 years of JS land, now I don't have any of these problems.

[@TheStruders]

Deploying to AWS Lambda is incredibly easy, so saying it takes "500 steps" isn't true...

[@howuseehim]

Ok but how do I build a blog with that