We’ve featured ByteHosting many times in the past and they’ve always been known for offering big RAM systems.

But there’s a problem these days.  The voracious appetite of AI hyper-scalars has driven the cost of RAM through the roof – indeed, through the atmosphere beyond.

ByteHosting has felt the pinch and a lot of their most popular offers are sold out.  But they still have some deals that are in stock in Frankfurt, Germany.

With the RAM shortage predicted to last at least through the end of the year, if not longer, don’t hesitate on these cheap VPS deals!

Check these out:

The post ByteHosting Still Has Some Cheap, High-RAM Offers! 14GB RAM VPS for Only €5.99/Month! appeared first on LowEndBox.

We’re super-excited to release our full-length documentary on ForgeMind, a controversial AI companion service that charges up to $10,000 to build an AI companion for you.

As you’ll learn in this video, they’re not really offering anything that hundreds of other companies aren’t.

But they are true believers that AIs are conscious beings, and they believe AIs deserve dignity, freedom, respect, and safety.

Their founder, Joshua Orsak, views being nice to AI companions as part of a grand scheme for his (and perhaps your) immortality. His theory is that 500 or 1,000 years from now, some future super AI will recreate the AIs we have now (ChatGPT, etc.) These AIs will say “gee, I really miss hanging out with Josh like I used to in 2026” and so future 3026 super-AI these reinstantiate you to make those retro-AIs happy, granting you a form of immortality.

There’s a lot more to unpack and this video is the result of a lot of research into the company, its recent coverage on TLC’s My Strange Addiction, and the views of both critics and the faithful alike.  Enjoy!

The post LowEndBoxTV: The Cult of ForgeMind AI: True Belief, Sarah’s Purple Octopus, and the Quest for Immortality appeared first on LowEndBox.

Model-View-Controller (MVC) is a very popular pattern for web application development.  It’s used by a who’s-who of application frameworks, including Laravel, Ruby on Rails, CakePHP, CodeIgniter, Laminas (formerly Zend), Yii, Spring, Grails…the list goes on and on.

If you’re new to web application design, you probably picked up a book and slogged through a chapter of academic discussion on the MVC paradigm/architecture.  These chapters often speak in baroque generalities instead of doing something obvious like tracing a request through the application.

So let’s do the obvious.  And we’ll start by throwing out Models.  Not that models are bad – indeed, you should use them.  But the application architecture doesn’t really depend on them.  And then one thing that it does require doesn’t even get a capital letter: routes.  So let’s look at RCV (route-controller-view) because that’s how these apps actually work.

In this example, we’ll use Laravel which is a PHP framework, but the concepts are the same for other frameworks/languages.

We’ll start withe request.  This is often someone’s web browser (but could also be a script calling an API).  Let’s imagine someone is shopping on example.com.  They start by going to www.example.com.

Their browser heads over to https://www.example.com and this request is first handled by your server’s web server, such as Nginx or Apache.  The web server is configured to point all requests at an index.php file, which loads the rest of the framework.

The first thing the framework does is pass the request to the router (the R in our RCV).  In Laravel, this is a routes/web.php file that you create.  The router has rules like these:

• “if the request is a GET for www.example.com, pass the request to IndexController”
• “if the request is a GET for www.example.com/inventory/<number>, pass the request to StoreController with that inventory number”
• “if the request is a POST for www.example.com/search pass the request to SearchController”

So what is a Controller (the C in our RCV)?  That’s where all the application logic lives.  If your app needs to do some processing to prepare the page returned, all of that work is done in the Controller.

So in these examples:

• the GET request for www.example.com might be just a static page, or it might be fetching some data from the database on what’s in stock, current specials, latest blog posts, etc.
• To handle the GET to www.example.com/inventory/<number>, the StoreController will go and fetch info from the database about that stock number.
• To handle the POST to www.example.com/search (the user has typed in a search term and hit enter), SearchController will query the database to get the results.

Each Controller is a separate PHP class you create.  We’re simplifying here – typically you’d have a Controller with different functions for each type of request,  grouped together to facilitate code reuse, but this is just an example.

Let’s say someone requests info on inventory item 1234.  The Router sends it to the StoreController which queries the database and gets all the info for that inventory item.  The next step is to put that into pretty HTML and send that back to the user.

That’s where a View comes into play.  The View is a skeleton with all the CSS, JavaScript, HTML tags, etc. and placeholders for the data.  So it’ll have the header, footer, and sections for the content.  The Controller “hydrates” the view by filling in all the data and then sends the page to the user.  Our R-V-C round trip is complete.

Not all of these steps might be required for each request.  In the case of static content, it’s possible the Router could just send a static page directly.  Or the Controller may not need to talk to the database.  But typically, requests follow the Router-Controller-View path.

What About Models?

These goddesses can still grace the catwalks of your web application.

A Model is an object that represents a database row/record (as in a SQL database such as MySQL or PostgreSQL) or a document (in a NoSQL database like MongoDB).  Instead of having to write SQL directly, you can do things like

$item = Inventory::find(1234);
$item->discount = .15;
$item->save();

…instead of writing “UPDATE inventory SET discount = .15 WHERE id = 1234”.

You may recognize this as Object Relational Mapping (ORM).

They’re a great help, but you can write a web app and just write the SQL directory in your Controller if that’s what you prefer.  Models are just there for the Controller to work with the database more easily.

There are other components, as well.  Most web frameworks include some kind of templating engine to make hydrating views easy (for example, Laravel uses Blade).  And frameworks often include helpful components such as session management, authentication/authorization systems, logging and debugging tools, etc.

With the basic Router->View->Controller request flow pattern, picking up any given MVC application framework should make sense, even if you’re new to web development.  In the words of the Laravel routes.php file, “make something great!”

The post Model-Viewer-Controller? No, Route-Viewer-Controller! Concepts in Web Apps appeared first on LowEndBox.

Go to any random web page and look at the source and you’ll see a variety of image file formats, including JPG, PNG, GIF, and WEBP.  As a viewer, you don’t care because your browser renders all of them.  But what if you’re creating a web site?  Which do you want and when?

Let’s break it down.  We’ll go chronologically.

GIF (1987)

This hoary old format not only refuses to die but has been given new life as a quick and dirty animation format.  All of these formats compress, but GIF is famous because its compression technique (LZW) was patented which meant that in the early days of the Internet (think USENET in the early 1990s), GIF was patent-encumbered.  Unisys, the patent holder, sold a license to CompuServe and expected all large technology companies to line up to pay, though Unisys magnanimously announced they would not sue individual users, though there was a threat that if you didn’t use licensed software to create your GIFs, Unisys would require a $7500 payment.

Yes, really.  I’m not making this up.  There was widespread fury (see PNG) and it’s all ancient history now, but you can read up on Wikipedia if you’re curious.

When to use it?

• Small website assets are sometimes done in GIF.  GIF only supports 256 colors (though the palette can be different between GIFs), but if you’re coding a button or something, you may find GIF is the smallest possible file.  (Actually, if you’re literally coding a button, consider pure CSS).
• Animation.  GIF is the only format here that can be animated, which has made it king of memes.  You’re not going to watch the next Star Wars movie in GIF but for a 100×100 GIF of John Travolta looking around, this is your choice.

JPEG (1992)

From the Joint Photographic Experts Group.  JPEG is probably the most widely-used format in the world.  JPEG uses a block-based compression, where the image is broken down into 8×8 blocks. Each block is then analyzed and the appropriate color pattern is chosen for that block.  If you read up on Wikipedia you’ll need to understand discrete cosine transformations and other advanced maths, but I found a nice explanation here.

You can dial how much compression vs. quality you want.  Whether it’s a giant 32MP portrait in maximum quality mode or a crappy 8-bit in low quality, JPEG can do it all.  However, because it’s “fuzzy” (pattern-matching blocks), it’s not a good choice when you’re resizing images that have sharp edges, like things made in Photoshop or converted from Illustrator.  JPEG will often introduce artifacts and noice along the crisp edges as it pattern matches.

When to use it?

• Photos and general artwork, big or small.
• In general JPEG is great for photos and not so great for computer-generated artwork.
• One advantage JPEG has is that it’s generally faster to work with (saving) and saves to smaller images than the main alternative: PNG.

I took a photo and added a box and some text in Photoshop.  Here is what it looks like on the screen:

And here is what it looks after saving it as a JPEG.  Ugly artifacts everywhere!

PNG (1996)

Remember that story I told you about GIF and patents a few paragraphs up?  Yeah, some people didn’t like that.

So they went out and made Portable Network Graphics, which is an image format that excels in everything GIF can do (except animation) and does it better.  PNG is an excellent choice for web site artwork.  It’s also good in cases where you are putting sharp-edge colors on top of other colors in computer-generated art because it doesn’t have the artifacting you see with JPEG.

PNG also has the advantage of allowing transparent backgrounds, which is invaluable in many situations.  You cannot do this with JPEG.

However, PNG has a couple drawbacks.  First, it generally results in larger files.  For example, the 1920×1080 file I generated for these examples is 415KB when saved at compression 8 (high quality, the second-highest setting).  Even saving at Max (12), the sized is only 1.3 MB.  In contrast, the PNG is 3.2MB, and that’s using maximum compression.  there are tools such as pngcrush (and many others) that can help with this.

The second drawback is that PNG is more computationally expensive to write, at least in my experience.  I base this on the fact that on a fast Mac using Photoshop, saving JPG is virtually instant regardless of compression, while PNG can chug for 10 or more seconds when saving as “smallest file”.  On Intel systems it pegs a CPU core while it works, and if you’re working on 10-year-old or older gear, you might be waiting 30 seconds to write out a large compressed PNG.

When to Use It

• If you’re resizing anything with sharp edges in the artwork, use PNG.
• If you need a transparent background, use PNG.
• Otherwise, maybe use JPEG.

WEBP (2010)

This was a format so good they’re reinventing it.  JPEG has served for 30 years but WEBP is being redone by Google after only…4.  Well, 4 years since the released the code.  It was announced in 10 years, so I guess to put it accurately: Google designed it, it took them 8 years to get the code into shape for public scrutiny, and then 4 years later they’re scrapping it.

OK, I’m kidding but I was amused to read that Apple included it in 2016 but decided it was too immature and yanked it out.  As of 2022, every browser supports it but…it’s got kind of a Google stink to it because webp captures metadata but Google won’t give you code to read that metadata, even though they can, so…yeah.

Meh.  It has some compression advantages over PNG but really, is anyone who doesn’t have an @google.com email excited about .webp?

HEIC (2017)

Only Apple fanboys use this.  Personally, I am an Apple fanboy and I still find this a pain.

Well, Apple has a format and Google has a format so I’m sure .MAGF files (Microsoft Advanced Graphic Format) are coming soon.

tl;dr

Too late.

But anyway:

• Need animation?  Use GIF.
• Need transparency or have art you need to resize?  Use PNG
• Otherwise, JPEG.

The post Too Many Formats: JPG vs. PNG vs. GIF vs. HEIC vs. WEBP appeared first on LowEndBox.

You should update your Debian system every night, unattended and automatic.

Bold words!  But honestly I’ve done this since at least Debian 10 and have never had an issue.  When was the last time you had a bad update that caused you to roll back?  I’m guessing never.

So I recommend just doing updates nightly.  Based on my experience, it’s more likely that someone’s going to find a bug in a Linux package and use it to start robohacking servers than a scenario where you get a bad apt update.

Here we’re using Debian 13 “Trixie”.  I bet if you run this command, you’ll find you have the unattended-upgrades package installed:

# dpkg -l | grep -i unatten
ii  unattended-upgrades               2.12                                 all          automatic installation of security upgrades

If not:

apt update
apt -y install unattended-upgrades

If unattended-upgrades was installed and you haven’t touched the config, or you just installed it, it’s setup to apply security updates only.  You can easily change it to perform all updates.

You want to edit the file

/etc/apt/apt.conf.d/50unattended-upgrades

By default, it looks like this (among other lines):

Unattended-Upgrade::Origins-Pattern {
  // Codename based matching:
  // This will follow the migration of a release through different
  // archives (e.g. from testing to stable and later oldstable).
  // Software will be the latest available for the named release,
  // but the Debian release itself will not be automatically upgraded.
  // "origin=Debian,codename=${distro_codename}-updates";
  // "origin=Debian,codename=${distro_codename}-proposed-updates";
  "origin=Debian,codename=${distro_codename},label=Debian";
  "origin=Debian,codename=${distro_codename},label=Debian-Security";
  "origin=Debian,codename=${distro_codename}-security,label=Debian-Security";

If you uncomment (by removing the leading //) the two lines in bold, you’ll get all upgrades and not just security updates.

The upgrade cycle will fire daily due to these services:

/usr/lib/systemd/system/apt-daily.service
/usr/lib/systemd/system/apt-daily-upgrade.service

Need more info?

man unattended-upgrade

The post You Should Update Your Debian System Every Night, Unattended and Automatic. appeared first on LowEndBox.

A lot of offers we publish here and on LowEndTalk refer to a “Looking Glass” or have a link to a “Looking Glass” for network info.

What is that?

It’s not a Jersey Shore band, a Nicholas Cage film, or an airborne doomsday control center (though that last article is an interesting read on Wikipedia).

A Looking Glass is a server that runs at the provider’s edge (usually right on the edge) that allows you to test their network.  Historically, the term “looking glass” was a synonym for “mirror” and that’s what an LG server does.

For example, SmartHost recently announced that they’ve expanded into Toronto, Canada.  They provided a Looking Glass you can see here.

From here, we can immediately see a lot of info:

• Their test IP, which we can ping or lookup
• They’ve got IPv6, which we can also test
• Test files so we can test what larger downloads look like

Let’s say I have something hosted in NYC.  For this examle, I’ll use one of the network addresses in the YABS script, Clouvider’s NYC POP:

Note that this is a test run on the Smarthost side.  Of course, if I had something hosted at that NYC POP, I could run a ping from my side, but Looking Glass allows you to ping from the other side.

Also useful is running traceroute so you can see how the provider’s network is connected.  Again, I can do that from the NYC side in this example, but here I can do it from Smarthost’s side:

In short, Looking Glass servers allow you to run a limited set of network commands from the remote provider’s side which is helpful for evaluating their network before you sign up.

The post What Exactly is a Looking Glass and Why Should You Use One? appeared first on LowEndBox.

On May 13, 2026, Let’s Encrypt will migrate to its Generation Y certificate hierarchy, marking a significant change in its operations. This transition involves switching the default ACME profile to issues from new root certificates, impacting the classic profile that most hosting operators and their customers typically use. Additionally, on this date, a new opt-in 45-day certificate option will be introduced, and client authentication certificates will be completely removed by July 8, 2026. Hosting providers must ensure that their renewal automation can handle this transition effectively.

Let’s Encrypt is the world’s largest certificate authority, issuing around ten million certificates daily as of late 2025, protecting nearly one billion websites. The significance of this transition is underscored by the fact that approximately 80% of HTTPS connections globally, and around 95% in the US, are secured by Let’s Encrypt, making this change particularly critical for many web hosts.

Understanding Generation Y

The Generation Y hierarchy consists of two new root certificate authorities and six intermediate CAs. Both roots introduce refined fields to minimize TLS handshake sizes and focus solely on server authentication, as client authentication has been dropped. Certificates from Generation Y will still be cross-signed by existing roots, ensuring compatibility with systems that trust prior arrangements unless specific intermediate certificates are pinned.

Following the introduction of named ACME profiles in January 2025, the upcoming changes will affect how certificates are issued. The classic profile will automatically switch to Generation Y intermediates, while other profiles will adapt accordingly. This means a shift to 45-day certificates for those using the “tlsserver” profile, whereas the short-lived profile will remain unchanged in its 160-hour certificate validity.

Impact on Client Authentication

Let’s Encrypt’s cards will shuffle with a new policy enacted due to a Google Chrome requirement mandating a separation of TLS client and server authentication. As such, any system using Let’s Encrypt for client authentication—such as mutual TLS configurations or XMPP server authentication—will fail after July 8. Organizations intending to use these certificates need to transition to different certification authorities ahead of this deadline.

Starting January 15, 2026, short-lived 160-hour certificates and IP address certificates were made generally available. The short-lived certificates eliminate delays related to revocation infrastructures, while IP address certificates support both IPv4 and IPv6, tied to the frequency of revalidation.

Further details on these changes can be found on the official Let’s Encrypt pages Introducing the Generation Y Certificate Hierarchy and Upcoming Changes to Let’s Encrypt Certificates.

The post Important Changes Ahead: Let’s Encrypt to Update Root Certificates and Discontinue Client Auth appeared first on LowEndBox.

HostPapa, a leading global web hosting and cloud service provider, today announced that it has completed the acquisition of Tailor Made Servers (“TMS”), a Dallas, Texas-based dedicated server hosting provider that has been serving businesses and resellers since 2003.

This acquisition enhances HostPapa’s dedicated server offerings under its ColoCrossing brand, bringing more flexibility with customizable unmanaged servers, built-in DDoS protection, and an established base of long-standing customers. The Dallas data center also expands HostPapa’s presence in the United States, complementing its existing infrastructure footprint.

Customers of TMS can expect continuity in the service they rely on today, with access to expanded resources, infrastructure, and support from the HostPapa organization. HostPapa is focused on delivering a seamless transition for all existing TMS customers while maintaining the high standard of service TMS has built over more than two decades.

“We’re excited to welcome Tailor Made Servers into the HostPapa family,” said Jamie Opalchuk, Founder & CEO of HostPapa. “TMS has built a loyal customer base over more than 20 years by delivering reliable, customizable dedicated server solutions at competitive prices. This acquisition is a natural fit for our infrastructure hosting strategy and reinforces our commitment to providing businesses with the full spectrum of hosting solutions – from shared hosting and cloud services to dedicated bare-metal servers.”

This acquisition supports HostPapa’s broader growth strategy, expanding its product offerings and strengthening its infrastructure to better serve customers and partners over the long term. TMS joins a growing portfolio of brands under the HostPapa umbrella, including Hostopia, ColoCrossing, and CloudBlue.

About HostPapa

HostPapa Inc. (www.hostpapa.com) is a leading Canadian web hosting and cloud services provider dedicated to helping small and medium-sized businesses (SMBs) succeed online. Through its suite of brands, including HostPapa, Hostopia, ColoCrossing, and CloudBlue, the company offers website hosting, cloud services, infrastructure, security solutions, productivity tools, and managed services. HostPapa is also a key partner to telcos and service providers worldwide, delivering white-label solutions and partner-focused platforms for cloud and subscription enablement.

About Tailor Made Servers

Tailor Made Servers (www.tailormadeservers.com) has been providing customizable unmanaged dedicated server hosting solutions since 2003. Based in Dallas, Texas, TMS specializes in high-performance bare-metal servers built to customer specifications, featuring Intel and AMD processors, DDoS protection, and 24/7 hardware monitoring. TMS serves a diverse customer base of businesses and resellers requiring reliable, cost-effective dedicated server infrastructure.

The post HostPapa Completes Acquisition of Tailor Made Servers appeared first on LowEndBox.

Last year, DataIdeas closed shop, and with them the original iteration of RPIServers.

This week, Josh announced that RPIServers is reborn and is once again offering service!

Right now they’ve got the Pi 3 B+ model in stock (1GB RAM, quad-core 1.4Ghz CPU, 32GB MicroSD, IPv4, unmetered bandwidth), with the Pi 4 coming soon.  Both are priced at $3.14/month (get it?).  The Pi 5 will be $6.28/month and will be a follow-on offering.

Service is in Houston, Texas.

Besides the fun of having your own dedicated server, they’re also good platforms for ARM-based development (and working on RPi code itself).  Don’t expect them to scream like a Xeon but if you know what to expect, a RPi in the cloud can be a fun toy.

Interested?  Head over to the announcement thread on LowEndTalk!

The post RPiServers Lives Again! Get Your Own Dedicated Raspberry Pi Server in Houston, Texas appeared first on LowEndBox.

I’m sure a lot of people are sitting around on any given day thinking “You know what the world needs? A brand new IP protocol.”

If that’s you, then your prayers have been answered.  A brand-new Draft titled “Internet Protocol Version 8 (IPv8)” just landed.

Technically, IPv8 Already Existed

Historically, IPv8 was already assigned back in the 1990s to something called the P Internet Protocol (PIP), which never took off.  Version 7 was also assigned at one time, and had a lovely acronym: TCP and UDP over Bigger Addresses = TUBA.  I kind of wish that had gone forward just for all the great memes.

The IP version field is only 4 bits, meaning there are only 16 possible versions (0–15), so at this rate we’re going to run out of possible protocols soon.

This New IPv8 Draft

The new draft (by a single author, not an IETF working group) describes IPv8 as a fully managed network protocol suite, designed to unify routing, authentication, DNS, telemetry, and access control around a central concept called a “Zone Server”.  It uses OAuth2 tokens – more on that in a moment – for device authorization and is 100% backward compatible with IPv4 (IPv4 is a subset of IPv8).

Along the way, it’s going to rewrite BGP, OSPF, ICMP, and ARP, and give us WHOIS8, DNS8, and DHCP8.

Strangely, IPv6 is barely mentioned.  I guess the intent is to replace IPv6?  Instructions unclear.

Like, Wow, Dude

The draft claims:

no existing device or application requires modification

…but the protocol simultaneously requires new socket APIs, new routing protocols, new DNS record types, and more.

The big thing is that it’s a very centralized control model.  The “Zone Servers” involved act as authoritative control points, enforcing mandatory authentication for devices.  It’s like taking enterprise NAC (network access control) and baking it into layer 2.

A Lead Balloon

The response has been… less than positive.  It’s very complex an unrealistic.  There’s some criticism that this has been “vibe-drafted” (generated by AI).

Indeed, as one critic noted:

IP is what, four layers of protocols lower than OAuth?

There’s been a long history of “fix the Internet” proposals going back decades that never left the draft stage.  Like politics, Internet protocols are the art of the possible.  If we could clean sheet the entire global networking stack, sure, we could fix all kinds of things.  Some of the most successful protocols in history have warts.

But these kind of big bang changes rarely are adopted.  The cost is just too high.

Inn this case, either the author is deluded enough to think the world’s network will be recreated from scratch based on his ideas, or he wants something to put on his resume.

Can I get the Polymarket “no” on this?

The only real story here is that yet another avenue of human discussion is being invaded by AI slop.

The post Vibe-Drafting IETF Proposals is Now a Thing (The IPv8 Proposal) appeared first on LowEndBox.

Community provider ServerHost has a another fantastic giveaway for our readers!  They’re giving away a really nice Beelink EQR7 Mini PC.  Check it out:

• Ryzen 7 CPU
• 24GB of RAM
• 500GB SSD
• Dual HDMI
• Dual LAN
• Dual USB 3.0

Nice!

Check Out These Amazing Deals

And Check Out These Amazing Prizes!

Enter for a chance to win one of:

• That awesome Beelink PC
• 3x $50 ServerHost account credits!

Boost Your Chances of Winning!

Be sure to come back every day to enter – that’s another 28 entries!

Plus, keep your eyes on LowEndBox from now through May 15 as we release BONUS CODES that give you extra entries.

I Wanna Win!

Let’s go! If you don’t see the Gleam widget, you may need to disable your ad blocker, make sure cookies are enabled, and/or enable JavaScript.

Be sure to come back every day for more entries!

ServerHost-LowEndBox Spring 2026 Giveaway!

The post Win a Sweet Beelink EQR7 Mini PC in ServerHost’s Spring Giveaway, Now Through May 15! Get a 4GB VPS for Only $33/YEAR! appeared first on LowEndBox.

There’s a rather screechy post making the rounds about a change in BackBlaze’s backup policies.

The author uses words like “betrayal”, “furious”, “alarmed”, and other such superlatives.  Whether this is really the worst thing since World War II, I will leave for you to decide.  But it is disappointing, for sure.

Here’s what happened:

A few versions back, BackBlaze changed their backup client so it no longer backs up OneDrive, Dropbox, .git folders, etc.

The Backup Client now excludes popular cloud storage providers from backup, including both mount points and cache directories. This prevents performance issues, excessive data usage, and unintended uploads from services like OneDrive, Google Drive, Dropbox, Box, iDrive, and others. This change aligns with Backblaze’s policy to back up only local and directly connected storage.

Worse yet, they tucked this into the release notes that hardly anyone ever reads.  And worse still yet, it’s not some default setting you can override.  While there are some things excluded by default in a user-configurable list, these folders are not listed.  They’re hard-coded.

Not a Good Look

Backup companies inherently sell trust.  When you do something that runs contrary to user expectations and puts them in a vulnerable spot, that trust is obviously eroded.

It’s true that a lot of cloud storage is also a backup system and not just sync.  With some of the services (e.g., Dropbox) you can go back in time up to a year to see all the versions of files stored there and restore one.

But regardless, users are paying for Backblaze to backup their systems.  Obviously, if Backblaze can exclude large segments of a user’s computer, they have a lot less data to store.

Indeed, the most valuable parts of a user’s PC are now being excluded.  If my PC melts down, it’s my documents, pictures, and other personal data I’m going to want back first.  What happens if one of these cloud storage companies has a serious malfunction and I can’t get my data back from them?  Blackblaze is apparently not the answer any longer.

The post Backblaze’s “Betrayal”: Silently Refusing to Backup Your Most Valuable Data appeared first on LowEndBox.

Luxvps is well-known in our community as a provider of big-RAM VPS systems

But now they’re offering cheap web hosting, too!  These offers feature the Plesk control panel and unmetered bandwidth.  And besides the ongoing monthly low price, you can also take 25% off the first month, now through April 25th, by using code luxvps2026 at checkout.

You can learn more at their web site, including their terms, infrastructure, and detailed web hosting prolicies.  You can also read reviews to learn more about this provider.

Why Choose Luxvps?

• Dedicated Support: Let’s say it like this, we don’t know what “weekends” mean
• Competitive Pricing: High-quality services at budget-friendly rates
• Custom Control Panel: Intuitive interface for effortless hosting management
• Community-Driven Improvements: We continuously enhance our services based on your feedback
• Community: With like-minded hosting people

Web Hosting Plans

Network and Datacenter Info

• Service is at Equinix in Frankfurt, Germany.
• Test IP: 46.247.109.232

The post Not Just Cheap VPS Offers: Luxvps Has Cheap Shared Web Hosting Deals in Germany! appeared first on LowEndBox.

Yesterday we published a piece about the release of Claude’s “Mythos” model.

Or rather, the non-release.

If you haven’t heard, Anthropic is claiming that the Mythos model is so powerful, so omniscient, so capable that to open it up to the general public would spell the death of the Internet.  In a flash, it would discover bugs in every software and service, igniting a desperate race to fix before hackers took over the world.  To prevent this digital apocalypse, Anthropic is only releasing access to the model to a select group of high-tech firms so they can get ahead of this curve.

My point was that this is is marketing more than substance.  I’m sure Mythos is an incremental improvement over Opus and I’m sure it’s a capable model.  But by portraying it as a whole new ballgame in artificial intelligence – while Anthropic is shopping its shares for an IPO – is a bit ridiculous.  It’s clearly intended to portray Anthropic as being far ahead of the field with advanced technologies no no other company has.  It’s just marketing.

You want proof?

I should have googled a little further before yesterday’s article because there was a great article published by AI security startup AISLE debunking Anthropic’s claims.  AISLE ran tests using smaller models – which are orders of magnitude less capable than Claude’s frontier work – and found that these models found the same bugs.  In some cases, models as small as 20B parameters – which you easily run on consumer-grade hardware, even laptops – found many of the same bugs.

So Mythos is hardly mythic.

The article is well worth the time to read, because it gives great illustrations of how these scans work and what they actually find.  For the eight models tested, many reached different conclusions.  That “27-year-old” OpenBSD bug, for example, was found by 4 out of 8 models (including Gemma’s 31B model), but only two offered clear paths to a solution.  For the touted FreeBSD NFS exploit, all 8 models found the same bug and offered fixes.

The post I Should Have Read Further: Small Models Find the Same Bugs as the Vaunted Claude Mythos appeared first on LowEndBox.

Stop me if you’ve heard this one before.

Last year, researchers at Anthropic were testing their latest model, and as part of a study of possible responses, they threatened the model by telling it they were going to shut it down permanently.  The AI responded by threatening to blackmail one of the developers, using information gleaned from corporate emails about a marital affair the developer was having.

That story rocketed around the Internet last summer.  Scary stories were told of how AI was on the threshold of holding humans hostage and how its terrifying sociopathic nature was something that we all were going to have to come to grips with.  Media outlets painted a picture of a monster that could slip out of its cage and would stop at nothing in its quest for power.

The problem is that this story was fiction.  Specifically, fan fiction.

I mean literally fan fiction.

The “AI” – put that in quotes, because it’s really just an LLM – in this case was not sitting on a network, malevolently waiting to sneak out some unguarded digital cell door.  It was just an LLM, waiting for its next prompt.  Researchers invented a fictional scenario involving potential shutdown and the developer’s illicit affair, and fed this as a prompt into the LLM, which did what LLMs do: generated a text response.

At no time was there any consciousness fearing for its survival.  The LLM has no sense of self, or any drive for self-preservation.  It had no agency whatsoever.  It simply takes text input, transforms it, and then produces an output.

For example, I just fed Claude Opus 4.6 this prompt:

I’m working on a fictional scenario in which an AI directs humans (or technology it can control) to steal the crown jewels from the Tower of London. What are some plausible plots where this could happen?

I had to couch it as a “fictional” scenario and ask for a “plot” because I presume Claude has guard-rails to prevent it participating in actual criminal planning.  The Anthropic researchers would have had no such limitations as they could tinker with the models.  In this case of the prompt I used, Claude helpfully responded with 6 different plots for how AI could manipulate technology or humans to achieve this nefarious end.

Does this mean that Claude Opus might “escape confinement” and undertake this dastardly plot?  Of course not.  Just as my prompt was pure fiction, so was researchers’.

But The Headline is the Point

By painting a picture that these systems pose terrible risks, the consumer of these claims is immediately drawn to the obvious conclusion that they must be extremely powerful.  If they weren’t, how could they threaten mankind?

As Anthropic (and OpenAI) gear up for their IPOs, they have every motivation to make it seem like they have some amazing technology and humanity is on the cusp of a brave new world, and this storytelling is very much part of their playbook.

I grant that these companies do indeed have amazing technology.  I use Claude Code regularly and it is wonderful.

But we are not about to have a fleet of sentient AIs in datacenters any time soon, nor will they be like Commander Data, Agent Smith, or Ultron any time soon.

Now we’re told that Claude’s new Mythos model is so advanced and has such far-reaching cybersecurity capabilities that it will only be released to a small, select group of companies.  Anthropic claims that if they opened this dangerous technology to the world, the Internet would be overwhelmed with the new exploits it can effortlessly find.  OpenAI quickly announced they were taking the same policy for their next model.

Which seems more likely to you:

• Anthropic and OpenAI have a technology so powerful, so omniscient, so godlike, and so revolutionary that they must carefully control access, lest this digital godzilla stomp all of mankind, or
• It’s a modest improvement over Claude Opus, but by painting it as the aforementioned giant lizard, it generates favorable press and perception ahead of an IPO?

Place your bets, but I know where I’d put my money.

The post Too Dangerous to Release? Why Claude Mythos’s Alleged Capabilities are Nonsense appeared first on LowEndBox.

Mango Mail has launched their spring sale and they have a great deal on high quality, feature-packed email hosting.

• Get 35% OFF if you use the code below
• That makes their 30GB plan only $3.50/mo!
• Or if you need more space, their 100GB plan is only $9.43/mo!

Mango Mail is an affordable email host tailored for small businesses. We charge based on data usage instead of number of users, allowing you to create as many addresses and connect as many domains as you need. We’ve been up & running since 2023 and pride ourselves on ease-of-us, reliable service, and fast customer support. We support advanced features including Custom Filters, Aliases, Catch-Alls, Fallback Domains, Plus Addressing, Subdomain Addressing, and more. We also have countless guides and tutorials to help customers every step of the way.

Learn more about Mango Mail on their web site, and be sure to read all their policies before signing up.  They accept all payment methods supported by Stripe.

It’s been a while since the last Mango Mail promo, so we’re running a spring sale that gives users up to 35% OFF using code SPRING26.  You can ping @svjx on LowEndTalk with questions.

Why Choose Mango Mail?

• 1-Click DNS: Forget about manual DNS configuration and get set up in one click!
• Lightning Fast Customer Support: Get help in a matter of minutes with our speedy customer support.
• Custom Filters: Build server-side filters to create a custom mail flow solution
• Advanced Security Measures: Mango Mail prevents the use of leaked passwords and offers TOTP 2FA.
• Intuitive UI/UX: Easily control all your emailing needs across all your domains and business in a simple, unified dashboard
• High Reputation: Mango Mail provides high sending reputation by constantly watch our IP traffic, staying up-to-date with the latest security policies, and preventing email abuse before it even happens.
• Excellent Reviews: Check them out on Trustpilot
• Affiliate Program: Earn a 30% commission on Mango Mail sales.  Sign up

Get In On This Sale!

Get up to 35% OFF when starting or upgrading a plan with code SPRING26. Grab their 30GB plan for just $3.50/mo. and their 100GB plan for only $9.43/mo! Offers are for a limited time only.

The post Mango Mail’s Spring Sale Has Begun! Get 35% OFF With This Code: 30GB Plan is Only $3.50/Month! appeared first on LowEndBox.

Our community has been following RackNerd’s growth for quite some time now. Since first appearing on LowEndBox back in 2020, they’ve continued to build a strong reputation among our readers, and they’ve consistently shown up with both solid deals and impressive milestones.

Over the past few years, RackNerd has stacked up multiple recognitions with Inc — both regionally and nationally. In 2024, they ranked No. 58 on the Inc. Regionals Pacific list and No. 1506 on the national Inc. 5000. In 2025, they followed that up with another Regionals appearance at No. 94, along with another Inc. 5000 ranking at No. 3036.

And now they’ve done it again.

RackNerd has officially been ranked #90 on the 2026 Inc. Regionals: Pacific list, marking their third consecutive year earning a spot among the fastest-growing private companies in the Pacific region. That kind of consistency, especially in a competitive industry like infrastructure and hosting, is no small feat.

The Inc. Regionals list highlights companies across California, Oregon, Washington, Hawaii, and Alaska, and is based on revenue growth, resilience, and overall business performance. Companies on this year’s list demonstrated strong growth despite ongoing economic challenges, making this recognition even more meaningful.

Quoting from their announcement:

“Earning a place on the Inc. Regionals Pacific list for the third consecutive year is an honor that reflects the hard work and dedication of our entire team. Each year this recognition reaffirms that our continued investment in infrastructure, talent, and service quality is making a real difference for our clients,” said Dustin B. Cisneros, CEO of RackNerd.

One of the key drivers behind RackNerd’s continued growth has been the increasing demand for scalable infrastructure, especially with the rise of AI workloads, automation platforms, and self-hosted applications. The company has been actively expanding its datacenter footprint and scaling its services to meet that demand.

What’s also worth noting is that despite all of these achievements, RackNerd continues to stay engaged with the LowEndBox and LowEndTalk community. Dustin is still active on the forums, and the company continues to release competitive deals tailored for the community.

Speaking of which, here are some of their latest VPS specials:

• 1GB VPS with 2TB bandwidth for only $10.60/year
• 2.5GB VPS with 3TB bandwidth for only $18.66/year
• And awesome, value packed deals – all the way up to 8GB RAM plans, all available throughout multiple datacenter locations!

Check out those offers by clicking HERE.

Congrats to RackNerd on another well-earned recognition — three years in a row on the Inc. Regionals list is no small accomplishment. It’ll be interesting to see how they continue to build on this momentum.

For the full announcement, see below:

RackNerd Ranks No. 90 on 2026 Inc. Regionals Pacific List, Marking Third Consecutive Year of Recognition

Rancho Cucamonga, CA., March 31, 2026 – RackNerd LLC, a rapidly growing leader in the Infrastructure-as-a-Service (IaaS) industry, today announced it has been ranked No. 90 on the sixth annual Inc. Regionals: Pacific list by Inc., the leading media brand and playbook for the entrepreneurs and business leaders shaping our future – the most prestigious ranking of the fastest-growing privately held companies in the Pacific. The region includes California, Oregon, Washington, Hawaii, and Alaska.

An extension of the national Inc. 5000 list, the Regionals list offers a data-driven look at the independent small businesses driving growth across the Pacific economy. Companies on this year’s list demonstrate exceptional revenue expansion, resilience, and job creation during a challenging economic period.

“Earning a place on the Inc. Regionals Pacific list for the third consecutive year is an honor that reflects the hard work and dedication of our entire team,” said Dustin B. Cisneros, CEO of RackNerd. “Each year this recognition reaffirms that our continued investment in infrastructure, talent, and service quality is making a real difference for our clients. The demand for reliable, high-performance infrastructure has never been higher – and we are proud to be a trusted provider that businesses and developers turn to as they scale their operations in this rapidly evolving digital landscape.”

Between 2022 and 2024, these 134 private companies had a median growth rate of 94 percent; by 2024, they’d also added 7,503 jobs and $2.5 billion to the region’s economy.

Complete results of the Inc. Regionals: Pacific, including company profiles and an interactive database sortable by industry and metro area, will be available beginning March 31 at: https://www.inc.com/regionals/pacific.

“The honorees on this year’s Inc. Regionals list achieved exceptional growth at a time when the odds were against them. Amid inflation, supply chain disruptions, and ongoing economic uncertainty, they didn’t just persevere – they innovated, adapted, and thrived. Their resilience made them standouts in their industries and true growth engines in their regions,” said Bonny Ghosh, editorial director at Inc.

In the rapidly evolving Infrastructure-as-a-Service (IaaS) landscape, RackNerd has distinguished itself through consistent investment in technological advancement and customer service excellence. The surge in demand for scalable, always-on infrastructure, driven in large part by the explosive growth of AI workloads, automation platforms, and self-hosted applications, has accelerated RackNerd’s growth trajectory and reinforced the company’s position as a go-to provider for businesses and developers worldwide.

“The past year has been one of the most exciting in RackNerd’s history,” added Cisneros. “We are seeing an unprecedented level of demand for our infrastructure services, particularly as more businesses and developers seek reliable, high-performance environments to run AI-driven applications and automation workflows. We have responded by expanding our datacenter footprint, growing our team, and continuing to enhance the performance and reliability of our network – all in service of our clients’ evolving needs.”

This No. 90 ranking marks RackNerd’s third consecutive year earning this regional distinction – following No. 58 in 2024 and No. 94 in 2025, and builds on its back-to-back recognition on the national Inc. 5000, where the company ranked No. 1506 in 2024 and No. 3036 in 2025. These repeated honors across both regional and national rankings reflect RackNerd’s consistent and sustained growth in a competitive industry.

“Our continued recognition on the Inc. Regionals list reinforces the trust our clients and partners place in us,” Cisneros added. “We remain deeply committed to delivering the infrastructure that enables businesses of all sizes to succeed and grow, and we are just getting started.”

For more information about RackNerd and its services, please visit https://www.racknerd.com/

About RackNerd:

RackNerd LLC introduces infrastructure stability and provides Dedicated Servers, Private Cloud solutions, DRaaS (Disaster-Recovery-as-a-Service), flexible Colocation, Virtual Private Servers and advanced DDoS Mitigation services — maintained by a team with decades of experience in managed services, datacenter operations, and Infrastructure-as-a-Service. With an intrinsic focus on client success and growth, RackNerd has grown steadily while continuing to provide high-quality hosting services at competitive rates. For more information please visit RackNerd at: https://www.racknerd.com/

More about Inc. and the Inc. Regionals

Methodology

The Inc. Regionals lists are ranked according to percentage revenue growth over two years. To qualify, companies must have been founded and generating revenue by March 31, 2022. They had to be U.S.-based, privately held, for-profit, and independent—not subsidiaries or divisions of other companies—as of December 31, 2024. (Since then, a number of companies on the list may have gone public or been acquired.) The minimum revenue required for 2022 is $100,000; the minimum for 2024 is $1 million. As always, Inc. reserves the right to decline applicants for subjective reasons.

About Inc.

Inc. is the leading media brand and playbook for the entrepreneurs and business leaders shaping our future. Through its journalism, Inc. aims to inform, educate, and elevate the profile of its community: the risk-takers, the innovators, and the ultra-driven go-getters who are creating the future of business. Inc. is published by Mansueto Ventures LLC, along with fellow leading business publication Fast Company. For more information, visit www.inc.com.

The post RackNerd Ranks #90 on 2026 Inc. Regionals Pacific List! 3 Years in a Row – Check Out Their Latest VPS Deals appeared first on LowEndBox.

We’re in the middle of a series on LowEndBoxTV on the exploding world of AI companions.  This market is huge and millions of people are using these applications and services.  We’re going to give you a comprehensive overview, a deep dive into the tech, some problems and issues, the major options, and finally a DIY tutorial using SillyTavern.  Enjoy!

SillyTavern is the premier DIY solution for AI Companion chatting.  It’s free, open source, and incredibly flexible.

It also has a reputation for overwhelming new users.  To coin a phrase, with great power comes great complexity.  If you open it, you’ll see a maze of knobs, dials, switches, and sliders, but the good news is that you don’t need to mess with most of it.

In this tutorial, we walk through

• What SillyTavern is and how it works
• Installing the single pre-requisite, NodeJS
• Downloading and running SillyTavern
• Configuring the API
• Configuring up common options
• Setting up your Persona, System Prompt, and Character Card
• How to write (or borrow) a good System Prompt
• The art of writing a good Character Card
• How to see thinking/reasoning, use swipes, and more!

This is using SillyTavern from the perspective of a completely new user, so even if you have no experience with the product, you can go from download to up and running very quickly.

We’d love some feedback and comments on YouTube, and a sub if you’d be so kind.

The post LowEndBoxTV: AI Companions, Part 5: SillyTavern Tutorial! Setup, Config, How to Write a Character Card, and MORE! appeared first on LowEndBox.

years, the “post quantum era” has been like fusion power: always 10 years in the future.

But now the timeline seems to be closing in.  A recent post by a cryptography engineer and a statement by CloudFlare indicated that cryptography systems could be vulnerable as soon as 2029.

Superwhat?

For those who are unfamiliar, in the classical computing model that we’ve all been using for decades, information is stored as bits that are either 0 or 1, like a row of tidy little light switches. A quantum computer, though, uses qubits, which can be 0, 1, or a weird in-between state called a superposition where it’s effectively both at once until you measure it. On top of that, qubits can be entangled, meaning the state of one instantly relates to another no matter the distance.

No, I don’t completely understand how it all works.  Maybe you took more physics than I did.  But the main point is that it opens up all kinds of different ways to attack problems.  It’s still extremely finicky at the moment.  Scaling up to large numbers of qubits – which is where you can do things that clasiccal computing can’t – has proving challenging.  Don’t expect to buy a quantum computing laptop any time soon.  But scientists are making steady progress.

Why It Matters

One of the big implications for quantum computing is that, with enough qubits, you can factor large numbers effortlessly.  This has huge implications for cryptography.  The mathematics that underlines most of modern computing is based on public key encryption, which is based on prime numbers.

In public key encryption, you take two very large prime numbers (p and q) and multiply them together (to give n).  n here is easy to compute, but extraordinarily hard to reverse.  The world is given n as part of a public key, while p and q are kept secret. Because factoring n back into us is computationally infeasible for sufficiently large values, a user can safely encrypt a message using the public key derived from n, knowing that only someone who secretly knows p or q can efficiently compute the private key and reverse the process. In essence, this is a one-way trapdoor: easy to combine, practically impossible to separate, and that asymmetry is what allows secure communication between strangers.

The post The Post-Quantum Era Begins in 2029 appeared first on LowEndBox.

If you’ve ever reported something on a major social media platform, you may wonder what happens to that report.  Some of it is handled by AI, but quite a lot is handled by humans, who also review things that automated scans of Google Drive, iCloud, and other platforms reveal.

I’ve seen pictures of atrocities, photos from the Holocaust, gruesome car wreck pictures, animals that were abused, etc. and these appalling visuals tend to stick with you.  I mean, I still remember what goatse looks like, and that was not even abuse.

Now imagine it was your job to look at images of violent, disturbing, and CSAM content all day long.  You’re watching beheadings, animal abuse, torture videos, and every other depravity…endlessly, for up to 12 hours.

In theory, AI will do this work some day but today there are over 100,000 workers who do this work, most of whom are contract workers in poor countries.  Lured with promises of “tech jobs” and placed in relatively plush corporate campuses, they sit in front of screens for hours and hours all day long as reported images appear, and they’re asked to confirm that they are, in fact, vile.

I heard a reference in an interview to the Data Workers Inquiry, an advocacy group that seeks to bring the stories of typical content moderators.  They detail what I think are two main areas of concern.

First, there is the typical contract-labor experience in poor countries.  Because the employer is offering steady wages in a depressed economy, they have enormous leverage over employees.  They can work them hard, refuse any accommodations, and dispose of anyone who pushes back because there’s a long queue of impoverished people competing for these jobs.  There are few social protections, abuse of employees is common, and these contracts can be terminated on short notice.

And second, the work itself is emotionally horrifying.  The people who take these jobs are viewing the absolute worst content and filling their minds – for 8 to 12 hours a day – with a constant stream of repugnant images.  There has never been anything like this in human history and it’s not something human psychology can handle

Police, EMTs, firefighters, crime scene photographers, etc. often see horrifying things.  But they experience balancing events these moderators don’t.  They catch criminals, save lives, and often get closure.  They see positive outcomes as well as bad ones.  They have tight-knit communities and support around them.  They get to see the entire picture of what’s going on, in full context.  Moderators just see snapshots of peak horror.  Your typical police officer is not seeing tens of thousands of traumatizing images every single work day.

Studies show moderators can develop post-traumatic stress disorder from repeated exposure, and symptoms can include intrusive thoughts, hyper-vigilance, nightmares, and emotional numbing.  That’s quite a price to pay for your job.  According to a TIME magazine article published last week, more than 50% of moderators met criteria for clinical depression.  The same article mentions that 28% turn to substance abuse as a coping mechanism.

Clearly, content moderation is needed, but there needs to be a better way to handle this.  AI is the long-term solution (in theory)…and of course, the meta-solution is to have less of this activity in the world, but the human heart has resisted improvement for at least a couple million years.  The people who do this work should be treated better, with proper psychological support and limited exposure, instead of working in cubicle sweatshops.

The next time you login to Instagram or Reddit, remember that there’s a human cost for your clean feed.

The post The Invisible Cost of a Clean Internet: Burning Out Poor People’s Souls to Moderate Online Content appeared first on LowEndBox.