How many times have you seen someone build an MVP with all the cloud bells and whistles, only to watch it go nowhere?
The product had Lambda funct...
For further actions, you may consider blocking this person and/or reporting abuse
Love the no-nonsense vibe here, honestly I'm always wasting way too much time setting up cloud stuff instead of actually building. you ever find yourself stuck on tech choices instead of just shipping?
Hahahaha all the dam time and then finding yourself back to your first choice.
Yep, ReactJS has morphed into an over-engineered framework for syncing what was once just a simple redundant text file. Back in the day, a lean PHP, Bash, and Python stack was more than enough for a solo developer—and it’s ironic considering many of today’s tech giants began as one-person shows.
What started as straightforward file loops now demands JSX, npm packages, and complex build pipelines, mirroring the very complexity it was meant to eliminate.
Meanwhile, AWS flaunts over 200 flagship services (117 with built-in encryption) and more than 1.3 million active certifications—each peddled as “mission-critical” for secure, scalable infrastructure.
That relentless marketing blitz sidelines nimble startups before they ever get a shot at challenging Facebook or Amazon; small teams simply can’t chase every rebrand, documentation overhaul, and new certification.
And if you doubt big-tech ethics, just watch how they leverage massive media muscle to shape narratives, pad profits, and lock in users—pure corporate prestidigitation.
Could any tech giant really go that far to protect its bottom line? No, of course not…
Hahahahahahaha you hit the nail on the head, I can read the pain in your words. Dam the suffering we endured. I can't call myself a full developer but recently while building projects. PROJECT 1 - LE ME 0 🤣😂
Hhuh
💯 When working on net-new product or capability, greatest risk is often market- or user-fit. Testing and rapid iteration is most important. Often we see the value most in the higher-order services. Need to transcribe real-time audio from websockets? Cloud service is often easier and more reliable to experiment with.
But trying to build an event driven architecture practice as a pre-req for your rapid experimentation? You'll be arguing over fat or thin schemas-- and even the philosophical meaning of what events are and will never get to actual innovation (the knowledge of what combination of features, UI, UX and target persona needs the improvement).
I remember setting SSO in AWS, where Cognito was involved. It was a mess. It was frustrating. Many bugs. Many parts didn’t even work. It was about 3 years ago. In general all aws works with very slow web admin, everything is complicated with their iam/ roles/permissions and they actually let you program their service instead of providing a clear, intuitive and simple admin and SDK.
Instead, today I look for new services from young and brilliant startups. These services are done by gurus so their “beta” is already stable by far from any AWS years of production. I’m not a security expert but comparing e buggy AWS to an innovative startup of gurus, I’m sure it’s the same about the service security.
Thousand armature AWS programmers can’t beat a single innovative startup guru.
What many of these so called "gurus" don't tell you is that they are using AWS under the hood and adding a premium on top of their shell, that you pay for.
Yes, since Ai joined the game.. one has to be really careful with how derailing all those extra npm, pips, aptgets, docker extravaganza can lead to.. it’s easy to go with a ”-y” knowing our ”vibe” tools can fix the bugs; and woha.. can be hard to cut the looses of a tech stack realizing down the line that such introduced complexities always bares future costs too; focus & ”archetypes” still rules.. easy to forget how big the incentives are for a new shiny tool echoing on social media, even via our dev friends; it’s what fuels the whole circus; affiliate funnels & more viewers for the short term; the more lost - the more valuable in a milking ”influencer” economy.
A few API calls put in the right places in a chmoded textfile can replace gigabytes of introduced risks from a day of Ai assisted repo explorations.
Sliplane offers no geographically appropriate options and I love it when you say "deploy anything that fits in a container" and then list a stack of JavaScript frameworks which could all easily be hosted on Vercel. Possible poorly chosen stacks the marketing bits.
I struggle to read posts from founders about how not to use X but do Y when there is a clear vested interest in bashing X.
I had the same opinion before I ever had vested interest :D
There are 2 sides to that. Yes, if I need a quick, simple instances, there are always cheaper provides than the big cloud services. Even secure enough. Like OVH, Hetzner, .. (even from Hetzner I had bad experience and the impression stays).
Big providers (AWS, Azure, GPC, IBM, ..) have different advantage (aside scaling). There are all the services matched together , from network, storage, backup, ... and most important - security is managed from a single source with easy delegation and fine-grained control. When you are a single dev, you may not care. Where there's a team/enterprise and when there are security concerns, that's a price for assurances.
Vendor-lock-in really is a rabbit-in-a-hat trick: non-technical execs are dazzled by “security” promises, while the real magic lies in simply moving data around within the same S3 bucket. Meanwhile, AWS keeps pulling new services out of its top hat—each rebrand, bells-and-whistles feature, and AI-powered dashboard demo the latest flourish in its buzzworthy show.
It’s fun to watch, but for freelancers and lean startups on a shoestring budget, the endless renames and tutorial tours torch runway faster than you can say “magician.” Suddenly you’re chasing hype over a text file lost in a labyrinth of permissions, all while juggling family life, mortgage stress, and looming deadlines.
Innovation should be a spotlight, not a smoke screen. Let’s keep applauding big-tech wizardry—but not at the expense of the little guys still learning how to pull their own rabbit out of the hat.
Solid post. The point about overengineering early is spot on. It’s common to see MVPs built with a full suite of AWS tools that never make it in front of real users. There’s definitely a temptation to mimic big-company stacks because it feels like “professional” work, but for most small teams or solo devs, it’s more of a distraction than a benefit.
AWS is powerful, no question - but complexity adds overhead. Most apps don’t need global scale or auto-everything on day one. They need clarity, feedback, and iteration.
If you’re just trying to build and learn, keeping it simple makes a lot more sense. A VPS and Docker gets you 90% of the way without the cognitive load. And if you grow past that, great - migrate later with more confidence.
Start with the simplest infrastructure that supports your current needs, not hypothetical ones., and revisit complexity only when there’s a real bottleneck or operational need.
thank you for breaking it down! i needed to read this
Cargo cult is a phenomenon when external actions or attributes are copied without understanding their internal essence and cause-and-effect relationships, in the hope of obtaining the same result as the object of imitation
Wow.. That's great
Thank you for one of the most logical posts I've read, I've been trying to explain this to people and friends for months or even years. The unnecessary complexity of AWS already makes things more difficult and its pricing is a bit of salt and pepper, on the other hand, as you mentioned, alternatives such as digitalocean and even koyeb not only make things easier, but also relieve the project and us with their fair pricing.
I always tend to go to towards Firebase first. It's dead simple and deploys what I need in seconds to build my MVP. It's also tied into GCP so should I really need something more complex than a database and some cloud functions it's there for me to use.
I totally agree with the author. In many cases AWS is overkill, especially for smaller projects. Their billing is tricky too. If you are not careful, you can get a large bill, even if you use their "free" services.
To be fair, never had that issue with AWS. The billing is tricky, but luckily never got a surprise bill!
Realistically, you don't even need docker. If you already have a template project with it cool; otherwise deploying an app directly on a server with an SQLite database somewhere in /var/www is good enough for your first several thousand users.
There's a reason why so many successful apps run on rails, and it's 0% about how good rails is for large apps and 100% abobut how quickly you can build and deploy something that works.
I've noticed a few posts on Reddit AWS sub that have had a similar take. There's a lot of "colleague showed me how to host X and my mind was blown". I think there's a tendency to think the answer is always AWS or GCP. It's much like the on-prem world used to always go for IBM.
It is overkill for the average Joe. It isn't overkill for those who need multi-regional infra under one brand/platform though with certified and accredited in-house or outsourced staff.
You can definitely do things smaller and I'm all for people learning but definitely use AWS when it makes sense, which is rarely for your side project .
I’ve spent over 25 years developing all kinds of projects and products, and I completely agree with this article. Personally, I’ve been living and breathing VPS environments since 2008—and I wouldn’t trade it for anything.
Over the years, with small tweaks and smart optimizations, I’ve made solid projects run like Swiss watches—without ever needing to migrate to expensive corporate services or overengineer the infrastructure.
More than once, I’ve watched competitors collapse under the financial weight of platforms like AWS or GCP, all while their projects hadn’t even validated a real business model. All that just to look bigger than they really were.
The key is to start simple, scale wisely… not with ego.
Do you recommend to your customers a VPS hosted solution and then provide general admin support services ? I would have thought most companies are hesitant to go with a simple VPS based solution? It too basic or limitations.
Do you think the VPS is a strong enterprise solution? Would love to hear more specifics about your use cases. Would you recommend to host custom LLMs via a VPS for small companies to integrate with their in-house apps and databases (for example)?
Absolutely. I work exclusively with long-term clients — some have been with me for over 14 years, and the “newest” has been with me for 5.
My standard approach is to develop the client’s entire project on my own VPS, managed with WHM/cPanel, and assign a dedicated account under the domain we’ll be working with. Once the client approves the project and is happy with the results, I offer them the option to either stay hosted on my VPS while they gain traction, or to migrate everything to their own private VPS.
In my experience, a properly configured VPS is not only more than enough — it’s a strategic choice. It gives me full control, security, and performance without unnecessary overengineering. It also allows for lean infrastructure and full transparency with the client.
Would I recommend using VPS to host custom LLMs for small companies integrating with their internal tools and databases?
Definitely — with proper resource allocation, smart architecture and strong security practices, it’s a flexible and scalable solution.
When someone trusts you with the performance and stability of their business, that’s a responsibility I don’t take lightly. That’s why my clients don’t let me go, and I never take my eyes off their systems.
That’s my take on that 🙂
Thank you @auvitijuana, very helpful and reassuring to know the VPS can be a solid enterprise ready solution 🏗️
I will have to slightly disagree with how this post is saying that AWS is the equivalent to overengineering. I'm pretty sure this wasnt the intended message from the author, but there are multiple statements in the post that implies so.
With some level of familiarity and bit of expertise, AWS services can be utilized to simplify your application, not complicate it.
Of course if none of your new hires are familiar with AWS, the initial learnings, mistakes, and potential complexity may be very expensive and everything this post explains applies.
But if you hired a senior level engineer who has a strong AWS background, there is a very high possibility that they would be able to bring up an entire prod grade infrastructure in a few days, or even a few hours, based on the requirements of the application. And for such engineer, it would be "just another application," not an "intimidating overengineered overkill application".
I do fully agree that AWS or any other infrastructure design does not correlate with success. But I really did want to point out that aws == overengineered statement is sometimes very clickbaity!
One of the problems mentioned here is the high cost. Hiring AWS senior engineers and using AWS facilities are both very expensive.
Hey folks! This mix-and-match serverless approach really resonates with what I've seen in my own projects! 😄
I've been burned by both extremes - overengineering with AWS's full stack and also dealing with the maintenance headaches of pure VPS setups. The sweet spot for my team has been cherry-picking serverless services based on what each does best.
We use Cloudflare Workers + R2 for most of our compute and storage (the no egress fees on R2 alone saved us thousands!), Neon for our database (the autoscaling to zero is game-changing for dev environments), and Supabase Auth for authentication.
That said, I still rely on AWS Lambda for certain workloads where Cloudflare Workers just won't cut it (like longer-running processes or with specific runtime requirements). And I'm definitely not anti-Docker - we use containers for some specialized services where it makes sense.
The beauty is that each piece can scale independently, we only pay when users are active, and we're not locked into any single ecosystem. When one service had an outage, it didn't bring down our entire stack.
For anyone starting a new project, I'd definitely recommend this "serverless buffet" approach over committing to either the full AWS suite or managing everything yourself.
Just my 2¢ from someone who's tried all three paths! 👨💻
Well said. Those like $8 a month VPS with 1 vcpu and 2GB RAM can get a surprising amount done (no GUI taking up RAM). I've set up several projects for people on one of them and they've run quite well. (In one case the RAM was tight so I turned on a bit of swap space -- it swaps.a few 100MB out of apparently 'dead code' (since it never needs to swap it back in), and away we go.)
This is exactly what I did (kinda?) with my latest side project. I bought a VPS, set up a Coolify service on it, and deployed microservices using Docker images pushed to GHCR. The way I see it, it’s scalable-ready. Right now, all the containers run on a single VPS, but I can move them to something like ECS anytime without needing much effort on the project side.
Coolify is great, good choice :)
There is a much simpler reason to not use AWS. Bezos is evil to his core.
Came here for this comment :D
My job locks me into using AWS. I accept for some smaller apps you don't need it as stated in the article, but does anyone know is there a compelling alternative (other than Azure)?
I think that likely many apps plan for scaling way too early and don't need that kind of immediately scalable cloud infra where they are and if they experience enough growth where it is needed they can cross that bridge then. If you start your project out in a Docker container or similar and follow generally good architecture patterns, it should not be too hard to refactor it for a new style of hosting. Of course that requires good ENGINEERING, not just coding. Which is something I think has become far too muddied as a distinction in our field.
I think also many backends aren't that useful. I have my personal site and another one of my sites that is just pure frontend and the backend is json files that I manually edit in GH codespaces and use it to write articles in pure HTML as well for it instead of using a static site generator, which while cool, add an unneeded layer if you know how to do frontend. I combine this with Cloudflare Pages for better serving and some other tools. GH pages is much better than 5 years ago but just still lacking a few things, which makes sense, it isn't meant to be a robust host. You get your CMS and project mapping built in there though(I wonder if there is a good plugin for making this even better), your backend if mostly static or needs easy adding can be easily supported by using GH as a CMS in many cases.
Edit: CRM -> CMS
100% agree with you on the complexity of cloud deployments. If you already use Docker Compose to define your multi-service application, you can deploy to the cloud of your choice (AWS / GCP / DigitalOcean) with a single command with Defang. Networking, compute (with auto-scale), managed storage (Postgres, Redis), GPUs, even managed LLMs (such as Bedrock or Vertex AI) fully supported. Full disclosure, I'm one its founders and CEO.
At a company I worked for, they provided services to a business that hired freelancers for translation work. The system was conceived and written at the peak of the microservices and NoSQL trend. It had 8 microservices, used MQTT, and indexing with Elasticsearch — all to mimic the Netflix stack.
Was it a project for around 500–1000 users? It was a nightmare to develop locally. Even with my computer having 32 GB of RAM, it couldn't handle initializing everything.
In my opinion, we only needed at most one microservice for the frontend, one for the backend, and that’s it.
Sliplane sounds cool, but you are just reselling Hetzner boxes at a premium, aren't you?
Can we license the actual software to run in our own servers? or why do we have to pay you to provision somebody else's servers for us?
If we are just reselling Hetzner, Hetzner is just reselling AMD cpus. I’d give it a try, we have a 2 day free trial!
Hm, I don't like the business model where you charge a premium for somebody else's services.
We've seen this go wrong a gazillion times - if your idea is successful, eventually some venture capitalist or US corporation will take over or get involved in some other way. Once the clients are locked-in, you can essentially charge whatever you want, since you would effectively own our infrastructure.
I know this is a common business model and you're not doing anything different from most services like this, but I don't like it.
You're asking clients to pay a premium on infrastructure just to get a subscription on a piece of software.
It doesn't feel fair or honest.
And there are alternatives with comparable features, such as Portainer, that only charges for the software, and lets you deploy on Hetzner or any cloud service of your choosing.
Maybe I just don't "get it", but what are you offering, specifically, that justifies the need to resell Hetzner?
Thank you so much sir You save me from a burden. i am working on a digital product and i am the only developer that create that product. I am in my recent days to build my first digital product to lunch it in market. I think about the AWS to host my product on that and searching courses to learn that. But i like your suggestion. thank you for the advice sir
Eh man funny enough I recently had this same scenario with my set up. I have to admit it was more of a curiosity thing, even though it delayed my project alot I learnt about cloud computing and I must say it's a humbling experience. But def spot on also I think sometimes as developers we tend to go overboard and we need others to tell us when to stop.
Looks like advertising to me... Sure, it might be a little more effort to get started but once you get into the flow, k3d + k3s provides an amazing flexible and easy to manage kubernetes Dev laptop to prod pipeline. Make yourself a template and spining up a new project is very easy and quick. Add kilo and you can access your prod services directly over wireguard, direct to their kube internal IP addresses. From you laptop. All nice and secure, all free.
And if you don't care about intercontinental network traffic, you can easily run it on Contabo and save yourself even more than Hetzner!
PS, I have no relationship with the k3d/k3s folks except being a very happy longtime user.
Heroku.com is a great place to start before going to AWS. I'm not affiliated but I've been using them for over a decade. Just build your project in a git repo and then git push it to a heroku app (server). You can do a lot with their bottom tier shared server and never have to upgrade. I strongly agree that app development should be about building what users want to use and you can optimalize and scale the infrastructure after figuring out the magic that brings in the users.
It's one thing to over engineer early and a different one to chose a cloud provider.over engineering early is a bad choice you make regardless of where
Choosing a cloud provider is a strategic choice for the future expansion that you normally hope for. You can build simple in AWS.
Also the title has little to do with the content and they are actually contradicting each other...
sure you can, most dont though once they see all the fancy tools and "best practices"
I made this mistake for my small business, I over engineered the crap out of everything. Granted everything is running so smooth, I've got external databases on RDS, I'm backing up to S3, I've got custom Amis... But my bill is around $1,000 a month and I know if I just hosted it on DO or vultr it would be 70% of that
While the sentiment is admirable, the justification is hand-wavy and ignores reality.
I agree that architecture creep is almost a forgone conclusion when working on AWS. However, you can't just do away with necessary components or just run them yourself.
Each layer and component you need is it's own domain of expertise. Getting them to play nice together is an even larger multidisciplinary domain.
Don't get me wrong. The point of this comment is not to promote AWS, but to highlight the fact that this article is not justifying it's title.
Fair enough! I do agree that the title implies a stronger opinion than the actual post:)
SlipKid your baited articles are great fun and keep up the good work. You're helpIng to educate real developers 😎
☝️ 100% this.
Unfortunately a lot of times its a project requirement, because companies are scared that platform providers that are not at least in the world top 3 are not compliant with all kinds regulations.
I'm convinced that 70% of the problems with AWS are purely naming and dashboard driven. What do I need to use? Where is it? How is it that IDAM is blocking me... again? Oh well, guess I'm using Azure and/or DigitalOcean.
You'll probaly end up don't needing it at all. In my experience it sometimes make you quicker-to-market because you don't need to setup the VPN and keep stuff updates as much and stuff. But when you use all kind of services and start over-engeneering (based on those 'cloud "architects"') you'll end up with way to much services and when unlucky even vendor-locked into the cloud provider where it doesnt matter if its Google, AWS, Azure or anything else.
Actually all you need to start is Firebase! 🔥 In some cases you can do the job even with WordPress.
I know, you are a software engineer and you love Consoles, VIM and bullshit but guess what... User just wants the service!
interesting take, but I think there's some nuance missing here. You're lumping together overly complex architectures (like Kubernetes) and then somehow placing blame on AWS for providing options. That's a bit like blaming a hardware store for selling too many tools!
For startups with actual investor backing (not just side projects), there are legitimate compliance requirements like SOC2 that make AWS a much more practical choice from day one. Try explaining to your investors that you're running on a single €5 VPS instead of infrastructure with built-in compliance controls, audit trails, and security features.
How do I know this? I've helped dozens upon dozens of startups migrate to AWS and set up proper security frameworks, especially for investor-backed companies. The pattern is consistent - those early infrastructure decisions have huge implications once you start dealing with enterprise customers or compliance requirements.
i agree that many MVPs don't need the whole AWS catalog, but the platform gives you room to grow without massive migration headaches later. Plus, services like Lightsail offer simple starting points without diving into the deep end.
Your core point about focusing on product is spot on though.
100% agree, and on purpose:D
AWS is not the problem. It sounds like you’re bogged down by traditional infrastructure hassles, ever tried SST.dev? Serverless skips all that, getting your project live way faster. Plus, it lightens the operational load, so you can iterate and build quicker.
SST framework is a breeze. You can write your infrastructure as code using languages you might already know, like TypeScript or JavaScript, which feels a lot less intimidating. With SST, you can develop locally, test your app, and deploy it to AWS in under an hour—it’s like a fast pass for the whole process. For instance, you could set up a full-stack app with API Gateway, Lambda, and DynamoDB in just a handful of lines. It ties into other AWS services smoothly and comes with pre-built components, so whether it’s a small API or a bigger project, you’re not starting from square one.
Yeah, managing servers sucks—I feel you on that—but AWS’s serverless setup with tools like SST.dev turns it into a non-problem. You can focus on coding instead of fighting infrastructure. That’s why I’d say AWS is still a killer choice—it’s become this “batteries included” platform that’s honestly great to use.
Cost-wise, it’s a no-brainer. I’m spending maybe $6-10 a month on prototyping projects. If one takes off, scaling’s easy—I just keep building the app and tweak the config in code with SST.dev to match the growth. Compare that to old-school server management, where you’re stuck paying for idle resources and slogging through manual scaling. Serverless isn’t just faster—it’s smarter.
The homologation and production servers only need the executable.
We, as devs, need a programming environment that partially reflects what we are dealing with as a problem.
Connections are facilitated by web workers, REST, SOAP and everything else... Do we really need the alphabet soup in which the entire ecosystem is assembled? I don't need a database on my machine, just a Docker database appliance is enough, since the database administrator is more concerned with other environments than the development environment.
Creativity is part of the process and also in terms of improving the process.
The aws free tier is fine for what you're describing.
Strongly agree.
You can get this with AWS too, using Lightsail.
Also, even a cheaper $30/year VPS from a provider on LowEndTalk (like GreenCloudVPS) is sufficient for a lot of use cases.
true!
Disclaimer, I work for Heroku. $5 a month will get you an amazing developer experience, with the power of AWS but without all of the fuss/complexity. Get out of the IaaS layer and into a PaaS layer.
BTW I love your comment
No, no they weren't.
I started with digitalocean app platform. Autoscale and super easy to start and scale. Than the government forced me to go with Aws. Now we spend more than 400 euro at month instead of 90 with digitalocean. Plus the DevOps team has got some very bad nights understanding all the small details of Aws to have exactly the same results. Now I'm pretty good with Aws elasticbean stalk, ECS, secret manager, S3, .... to understand that in the future if I can I will avoid it al all costs
Agree with this take. Our SaaS MyBazaar is also defined using just docker compose. The only reason we run it on AWS is due to the free credits we get. With it, I don't have to worry about my cloud bill for the next 2-3 years if all else fails. Otherwise, our system is very easy to deploy anywhere we want
Move to AWS later as you scale. Start off on a smaller scale like VPS, Vercel , Replit. I have over 15 years experience with scaling large scale applications like Starwars, Hulu and Photoshop, trust me
I mostly use AWS when I am done testing locally and need to see how it works in the real world.
The other part about AWS, they have free tier and depending how wild your App is, this may not fit. Also those free tier EC2'S can be awful and freeze horribly.
But also been a fan of a VPS self managed or even Managed w/cPanel/WHM or Webmin if you had to.
Your suggestion is an infinite amount more expensive as the AWS setup costs 0 dollars per month and requires no management once set up. This article exists only because of bias for your product.
Your take is equally as biased and I don’t think you read what I’m saying. AWS is fine. Unnecessary complexity is not. And again, please read. I explicitly say this does not apply to people who live and breath AWS like you.
The article is just way to one sided and its comparing apples to oranges. Your example mentions Lambda which implies compute, DynamoDB so there is need for a database, Cognito for some kind of authentication. The time required to learn how to do that (the right way, with persistent volumes and proper security) with containers also costs a lot of time.
You forget to mention that the AWS setup runs for free unless you actually will get a lot of users while your service I pay from day 1, even when no one is using it. Also you need to keep your containers up to date as they wont patch itself.
I understand that you are trying to promote your service and that is fine, just make sure the comparison is fair.
All of this yes, same with Azure. Too many large companies have these requirements for MVPs for day one. Ridiculous! MVP turns out to be really expensive. Honestly, I see the security departments mostly push for this over architected solutions because they want everything integrated into their VENDOR native AWS security hub or az security center gateway++++. Unreal.
CLOUD native if you want to get anything accomplished. Then add on as maturity grows.
This is pretty bad advice tbh. There are hundreds of reasons to leave AWS (or not even start there). However, the initial cost and serverless because they are "overengineering", are not some of them.
You could run entire projects at AWS (or other cloud providers) for free if you have few to none users. So, that's cheaper than a $5-$20 fix bill.
And, Lambda functions, API Gateways, Cognito, S3, CloudFront, DynamoDB, CloudWatch and IAM policies, are not overengineering, in fact they are very simple to setup in the first place.
I'm not saying you need AWS, just that those are not valid reasons to run away from it.
What you really need to ship faster is to master an ecosystem. It could be:
1- AWS.
2- GCP.
3- Working with a single server on DigitalOcean with the database flavor of your choice, plus your auth system (hand-made or external), plus whatever else you need.
4- Or even your local computer with a cloudflare tunnel and everything from 3 (this option is also free).
Then, for that ecosystem, create your own templates. Have a repo (or repos) with all the basic and common conf, plus all the basic boilerplate code. It will be an empty shell where you only have to take care of filling the core. And because it the same ecosystem, you will get better and better with time.
Those who think that they absolutely need AWS or GCP or Azure to build any MVP right from the start then they are not sure why we need a cloud service provider in the first place.
But yes if there's a vision, such services do enable us to have a better view and plan accordingly.
The real problem is between "we already know we WANT to scale" and "we would HAVE to scale".
Yes, you’re right. The truth is, you can use everything in AWS, but for initial stages, it's perfectly fine to start with just a server. You can maintain your website without using every AWS service from the beginning. It’s not mandatory to use all of them. You should choose the services based on your requirements and build accordingly.
Agree with the author, don't overcomplicate things unless there is a need to.
Developers always connect cloud services with scalability, but microservices are designed for scaling your team, not your app. There are plenty of companies out there with small teams serving lots of customers with monolithic apps spread across multiple servers.
Do yourself a favour, and only split your app up when you're teams are tripping over each other with their deployment. The likelihood is, 90%+ apps will never get to this.
But the tech leadership justifies their existence with complexity, the executive leadership justifies their existence with buzzwords and spending money. Convince them that a simple, consistent pattern/stack is better than complex overkill.
Not to mention, engineers support these choices because they see it as an opportunity to learn these technologies for the next job.
There is a lot of over engineering these days, the cloud has taken away the simplicity of putting something online.
Great take! It really resonates with what we’ve seen: early-stage teams fall into the same trap—spending weeks wiring up Lambda, API Gateway, Cognito, and DynamoDB before validating their product. Many get hit with surprise AWS bills that nearly wipe out their initial revenue, all for infrastructure they didn’t need yet.
Simpler setups—single containers, managed Postgres, automated deploy pipelines—get teams to production faster, so they can focus on what really matters at this stage: user feedback, product-market fit, onboarding, and iterating the core experience.
For early-stage apps, a lightweight approach—like running on a $5–$10 server with Docker Compose—can go a long way in minimizing cloud configuration while keeping deployment and scaling straightforward.
AWS is powerful at scale, but for an MVP, complexity and unpredictable costs often slow teams down.
This is a thought-provoking article.
I have the impression that the rise of the cloud has led to a lack of understanding at the OS level.
Good point we always go for the minimal approach. With AWS startup credits you can get an EC2 instance and use cloudfront and S3 that will give you all you need to get started and get your MVP to the client.
I’m a bit of a Newbie/late to this discussion but doesn’t something like Elastic Beanstalk take away much of that setup pain? Most of the AWS bits you mention sound like a mouthful but are just key basic ingredients (e.g. IAM.)
Docker Compose sounds interesting.
Last year I thought I work on a startup with a DevOps guy. I learned coding while he demanded he's not a developer, but a DevOps engineer.
All the ideas we were trying to validate required a simple VM running flask and and SQLite. It's that simple.
Then this guy demanded he's already having a k8s cluster, CI/CD pipelines with Argo CD, and that's the right way to do things. Everything else is a tech debt, he said.
We never validated anything, as the guy was always busy fixing his broken Argo CD, and I moved on other ideas solo.
Couldn't agree more and I say this all the time -- AWS is for products that have found PMF and are starting to see scaling issues, and even then it depends on the use case since and size of the team since Heroku and Render have a lower ops overhead.
Very good article 👍.
I appreciate this article for calling out project bloat. I think it's always easier to scale up a project rather than expecting to create everything all at once. I feel like most of the time when people use AWS for the personal projects, they are hoping to learn and get a job with their cloud skills. But always a good reminder to start small and slow.
haafa
Absolutely right. AWS is not necessary, its an option. You just need to know when to use it and when it's not needed.
Leave my CloudFormation config for S3 alone, it has feelings too
Yes! I never used AWS for exactly these reasons!
I had this exactly same ideia for a project, a deployer service to vps based on containers. I hope to learn a lot with your product, thanks for sharing
That's the tldr, and you're absolutely correct. This is akin to the sage advice against early optimization.
Sometimes a Small VPS is more than enough and way simpler
I think the point of using AWS is so that you have projects that can back up you've got AWS skills. Who's building on AWS just for the sake of it?
thanks for sharing knowledge
Very Important lookout explained Very Simply. Thanks!! 😊
Interesting information provided
You got this, from gpt right ? It sounds similar..
No offense, only matching experience..
Happy to see so many people responding positively to this. The infrastructure we sometimes think is normal to display an h1 tag is ludicrous.
Great stuff. Mate. I have no idea why anyone would want it. To deploy on AWS. The best use case is when you are required by a job to use the cloud resource.
AWS IS THE BEST CloudPlatForm!
No doubt in that.
Same vipe here ... the hype around the term cloud native for any project where it's not really necessary gives me cringe and how they demonize monoliths.
I've always used a VPS for my projects. I have used an S3 bucket in the past, but I didn't go through AWS. Even if I go the route of serverless in the future, I still don't plan to use them :p
just ship to a cheap VPS
learn how to deploy your app to a vps is quite hard, but interesting and fun to go.
you will be learned a lot of things.
Elest.io is also a great option for orchestrating open source tools without server management. It's fantastic and cheap as f*ck.
Hahahaha
I prefer my containers on a VM. Maybe a Portainer or even Yacht for a bit easier/quicker look and feel. Why AWS/Azure? Costs are astronomical.
As someone who was a systems engineer for 13 years, and specifically worked in AWS and Azure the last 5 years.. I can vouch for how accurate this is. Well written!
"You want to learn AWS because you're job hunting or building cloud career skills" is the only thing I'd use AWS for
Yet another advertising post.
Surprising how people don't understand this.
real place real people, look amazing, Thank to everyone, real appreciated
Mostly true. You don't even need Docker or container management. That's another two wasted layers.
AWS IS THE BEST CloudPlatForm!
It doesn't provide the services for only VPS or Other Services.
You probably don't even need Docker either.
I really needed this, thank you🙏
This is the same pitch cpanel gave
Aws is still essential for budding developers looking to enter cloud development(which you have mentioned so great work and research on your part)
Totally nailed it on the head here. I’ve stopped with the countless hours of set up, been using railway - almost too easy now
Jmpt
Use EC2. Leave these services to people who are using it.
That’s a cheap copy of heroku
Simple, directly to the point, easy to read. Journalists must learn from the author.
Stop plagiarizing Studio Ghibli.
Tell that to Mr Altman maybe? I never prompted ChatGPT to use Studio Ghibli Style. This is the default. Youre shouting at the wrong person here lol
This is one of the drawbacks of AI – it's already regurgitating itself.
Here, this was my entire chat. chatgpt.com/share/68177a59-6b7c-80...
Nothing is simpler and cheaper than a static host from S3 and cloud front for SSL. Then write lambdas for async calls to a postgres ec2.