Fifth Element. Resident Evil. And now — a free memory system for LLMs that most engineers haven’t figured out yet.

She spent months organizing thousands of docs, hit a wall, and asked the right question: why does AI still search like it’s 1998?

Her answer: give it a Memory Palace. The same technique Ancient Greeks used to memorize 70,000 digits of Pi.

She built the architecture, released the code. For free.

It’s called MemPalace: https://github.com/milla-jovovich/mempalace

Here’s what she said on her Instagram:

What inspired the idea of MemPalace is that I realized after months of meticulous filing, AI is just not great at finding things, even if you keep the best files.

Every system out there still has it using keywords to search through and that’s really clunky and messy in my opinion. In the end, your folder is like a huge warehouse and all those files are akin to a pile of junk with dates and names on it, so you’re still having to do these archaic word searches across thousands of documents and in the end, probably not even finding exactly what you were looking for.

Which gave me an idea. I wanted to create a new way to make filing and retrieving more intuitive. I had read a lot about how Ancient Greeks memorized long speeches and how the most famous memory masters are able to remember up to 70k decimal places of the number Pie.

I felt like we should be able to give AI the ability to remember information in a more “organic” way. And how this was achieved was through what experts call a “Memory Palace”.

Ben and I discussed it and thought “Why not create a virtual Memory Palace?”I mean at that point anything would be better than a warehouse full of junk.I came up with the architecture which is fairly logical: splitting the big, open space into different rooms.

This took me months of making many mistakes and starting from scratch (and with almost 1k docs, I’m sure anyone can relate how frustrating it got), but in the end, it was worth it. Because out of all that research and various blueprints that didn’t work, I finally managed to create the architecture and Ben engineered and “fine tuned”, that brought our “MemPalace” to life. Check out the MemPalace GitHub for free today!

She found a problem and built a solution for it with her friend. Now she is gathering feedback. How deeply AI changes our world every minute!

First: to launch the instance from Google console click SSH button in the VM line.

And you are in!

By the way, via Google console SSH window you can easily upload some files to your machine (works well with small ones at least). For that use UPLOAD FILE button on top of the window above. You also can download files using DOWNLOAD FILE button (I know, undexpected 🙂)

Now we already can set up the remote machine for work.

However, let’s make sure we can connect from our local console, it will be the second and most preferable way.

For that go to your local console and type

ssh-keygen -t rsa -f ~/.ssh/*keyfilename* -C *username*

where the username on our remote machine machine by default is your email part before @ symbol; keyfilename is up to you, but better make it recognizable as we will use it every time we need to connect to our remote machine.

Also, note that GCP doesn’t accept ED25519 keys, but RSA only (at least in November 2024 it was like that)

Command above generates a public and secret key.

Now, we have to put the public one to GCP to be able to connect. We’ll do it once and you will be able to use it for any machines you may later create within our project (the one chosen near Google Cloud).

For that, in the GCP console on the left panel scroll down to Metadata.

In an opened page you will see SSH Keys tab. Click ADD SSH KEY.

In our local console view and Ctrl+C the public key

cat ~/.ssh/keyfilename.pub, copy it and paste to the Google Cloud metadata key field. Click SAVE.

Now we can connect from our machine using this command:

ssh -i ~/.ssh/keyfilename username@machine_public_ip

where machine_public_ip is the IP you see near your instance on GCP and username is the same machine username (by default — part before email used to sign up on GCP)

You should log in without problem, however sometimes you may need to manually add public key to ~/.ssh/authorized_keys file on Google remote machine.

For that, connect via browser as shown at the beginning of this chapter and type in console

sudo nano .ssh/authorized_keys

If your public key is not there, paste it (with mouse right click) and Ctrl+X to save the file.

If the key is there — double check your connection line from local console 🙂

If now you want to deploy your project, set up a webserver, attach domain — check out my hostfast.me guides to do it all saving time and money. Basic version is completely free, you only need to provide email and GitHub username to get access. Instructions are given in form of convenient markdown in GitHub repos.

AWS vs GCP: what cloud provider to choose

And if you chose GCP, let’s go!

Go through simple signup via free trial registration link: https://console.cloud.google.com/freetrial. It will need your bank card to verify your payment ability, however you will still get notifications

Once we are ready, from the main page of our account we scroll to Products section and click Create a VM.

As it is a new account, it will at first redirect us to Compute Engine API, where we have to click enable (hereI have already clicked and waiting for it to be activated).

After that, you will be redirected to VM instances page

You should click Create instance (blue button on top or on bottom)

There you will see the basic characteristics for your new machine.

You can rename it here, change region, zone, machine type and other parameters.

If you have created your new account, you may have Google credits to test services. There is also a free tier, but make sure you follow instructions to use it. For free (even for no credits) you may use only e2 micro (0.25 cpu) machines in three regions: us-east, us-central and us-west.

Other machine types will cost you money (or credits).

However, you might be confused seeing the price on the right anyway. That’s a basic price you would pay after finishing your trial.

Usually for computations and server E2 is a good choice.

But do not rush into creating machine as it is, because default settings are not what you might expect.

To change machine type, click on the e2-medium combobox

We will see the free e2-micro instance there in Shared-core category, its usual price would be about 7.11$, but if it is the only machine, you won’t even pay for it.

Next few settings we’ll leave like this

And will scroll to Boot disk

Here you can choose preferred OS for your machine. Default one is Debian, but I usually go with Ubuntu, so let’s click “CHANGE”

In the opened popup you can also set the disk size and boot disk type. These parameters change leads to price updates. I go with defaults for now.

You can also load custom images or snapshots, archived snapshots and other existing disks here — write in comments if you want me to create a post on it. For now, assuming you have a fresh account, we won’t touch it and go with a public new image.

Under advanced configuration you can also set up extra parameters, such as deletion rule, encryption and other

Next we will make sure we can access server via HTTP and HTTPS and allow Health Checks.

Under Advanced options you can also modify several things, like allow access to VM only from some users, 2-step verification (security recommended, can be also done later).

You can manually add SSH-keys to be able to log into the VM not only from GCloud Console, but also from your local machine console, VS Code. You may also do it later.

Under the Management section you may also add a Startup script, the code which will be launched automatically when the machine boots. You can do it now or adjust later.

After setting these things up, let’s click CREATE on the bottom.

You will next see the VM instances page with your newly created machine preparing, it may take several minutes, so you can do quick exercise or make a cup of tea.

You will see the machine is ready when green tick is next to it and IPs set.

Details on how to connect to the machine using SSH you can find here:

How to connect to remote GCP machine using SSH key

If you want not only to rent a machine, but also deploy your project, set up a webserver, attach domain and set up health check notifications — check out my hostfast.me guides to do it all saving time and money. Basic version is totally free, you only need to provide email and GitHub username to get access. Instructions are given in form of convenient markdown in GitHub repos.

To use the deal, go to https://aws.amazon.com/free/

Fill in all data. You’ll need a bank card and mobile phone number to verify account

Then go to AWS Management Console.

From here we go to EC2 Dashboard:

Then go to Launch instance

It will tell that instance will be launched in a chosen zone. Zone has not much of a difference, especially on free tier, but you can choose it on top right.

Name your instance and choose OS, we’ll go with Ubuntu here.

Under it, near Key pair click “Create new key pair” — you’ll use it to connect to machine. I choose ED25519 algorithm. Save it on your machine, preferably in .ssh folder (in linux /home/username/.ssh) with some name/subfolder you will recognize later.

In network settings ideally we have to change SSH traffic from anywhere to our IP (DON’T FORGET TO TURN OFF VPN), but it will require additional setup when configuring deployment, so we’ll skip it for now)

Also, as we are setting up a webservice, tick https & http traffic options.

Now to storage

Free tier claims to work with up to 30 gb, let’s do this!

After this step (and making sure nothing is missed) click “Launch instance”.

It won’t take long.

You’ll see green success indicator and after this you may go to see your instance via “View all instances” on the bottom right

After this you are able to connect to it via your key, saved during configuration steps, congrats!

Next, let’s connect to our machine:

How to connect to AWS remote machine using SSH key

If you want to get full instruction on how to choose cloud provider, rent a machine, deploy your project, set up a webserver, attach domain and set up notifications — check out my hostfast.me guides to do it all saving time and money. Basic version is totally free, you only need to provide email and GitHub username to get an access. Instructions are given in form of convenient markdown in GitHub repos.

However, I would recommend using local console and SSH key as more flexible and convenient solution.

To do that, from your local unix console change the permissions for downloaded key with this command:

chmod 400 keyname

400 here means that only owner has rights to read the file.

After this, use ssh connect command using our key:

ssh -i path-to-key ubuntu@your-machine-public-ip where ubuntu is the default user on AWS machine with Ubuntu OS, and your-machine-public-ip can be found in AWS Console Instances page near our newly created one.

And you are in!

If something is not working, verify your key placement, double check IP and instruction steps.

You can change security permissions in Security groups from the left menu, under the Network & Security:

You may also let your machine some time to initialize properly.

If you want to get full instruction on how to choose cloud provider, rent a machine, deploy your project, set up a webserver, attach domain and set up notifications — check out my hostfast.me guides to do it all saving time and money. Basic version is totally free, you only need to provide email and GitHub username to get an access. Instructions are given in form of convenient markdown in GitHub repos.

AWS vs GCP

They are quite different yet both possess all the necessary stuff we may need.

For me, these are the main differences that play a big role. If you want access to the table with links, go to hostfast.me and get a basic guide for free!

Both AWS and Google Cloud allow to register several accounts to continue using free version longer. Both of them have nearly the same pricing after the free version stops: some services are slightly cheaper in one, some in another.

Decide for yourself, check both out and figure what suits your needs best!

What points would you add to the table? Which platform do you prefer?

Key advantages of cloud providers

1. More control over infrastructure

Cloud providers allow you to customize your infrastructure to meet the specific needs of your application. For example:

Non-Static Frontends

If your frontend dynamically fetches data, connects to APIs, or requires server-side rendering (SSR), platforms like Vercel might suffice for light usage but can struggle under high loads or complex dependencies. On AWS or GCP, you can tailor resources like virtual machines, load balancers, and storage to handle growing demands.

Constantly Running Backends

Hosting a backend service 24/7 on platforms like Fly.io or Railway utilizing serverless functions might incur additional charges or encounter downtime. With cloud providers, you can spin up persistent instances optimized for constant uptime.

2. Pricing and free tier comparison

Most platforms offer free tiers, but the costs grow as your application scales.

Cloud Providers

Both AWS and GCP provide substantial free tier options for new users. For example, GCP offers $300 in credits along with free usage of e2-micro VMs in specific regions, which is sufficient for small starter or pet projects. AWS offers similar flexibility through its EC2 free tier and S3 storage, however longer free term. These tiers provide the backbone for testing and early-stage deployment without incurring costs. And of course, free tier is eligible for commercial purposes.

Specialized Platforms

While platforms like Vercel, Render or Fly.io simplify deployment, their pricing models often charge by usage (e.g., compute hours, data transfer) and they have less generous free term. And, as I already said, most platforms specifically prohibit using free tier for business purposes.
For instance, Vercel has only one month free tier for hobby projects and charges for bandwidth beyond it, making it expensive for unexpected traffic loads, surprising website owners with huge fees.

3. Efficiency and long-term costs

For production-ready applications that require sustained performance and optimization cloud providers may be far better option.

Efficiency

Cloud providers let you configure resources (e.g., upgrading from a micro instance to a compute-optimized one).

Cost-Effectiveness

If your app grows, cloud providers become more cost-efficient since you only pay for what you use, with extensive options for optimization.
Vercel and other similar platforms often use the same cloud providers, but additionally charge you for convenience of their service or non-optimized flow. So going straight to the cloud can be a good saving option.

When to use specialized platforms

That said, platforms like Vercel or Fly.io shine when speed of deployment or simplicity is the primary goal and you can spend some money from the start for your project:

• static websites: for simple one-person blogs, documentation, or marketing pages, Vercel or GitHub Pages can be cost-effective and easy to set up
• prototyping: quick proof-of-concept applications can benefit from the managed nature of these platforms

For larger projects or those aiming for long-term growth, cloud providers offer a better foundation, flexible configurations and more control over the budget.

Getting Started With Hosting on the Cloud

If you’re ready to explore cloud providers, my HostFastMe guides are here to help you:

• free hosting: learn how to deploy your website on GCP or AWS without spending a dime (up to a year)
• attach a domain and get free SSL certificate: simplified steps to link a custom domain and connect https to your cloud-hosted site
• bonuses: tips and tools to optimize your hosting setup and maximize free-tier benefits

Want to give it a try totally for free? Contact me via X or LinkedIn or just go to my website and obtain a basic guide for 0$ (for first 20). I’ll provide you access to my repository with all the materials to get started and will be waiting for your feedback.

Conclusion

While proxy platforms are excellent for certain use cases, cloud providers give you unbeatable flexibility, scalability and more control over services and budget. They’re ideal for projects that demand custom configurations, constant uptime, or dynamic functionality.

Explore both options to find what suits your needs best, and when you’re ready to dive into the world of cloud hosting, my guides at HostFastMe will walk you through every step.