New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Oh but you would use our data and update the privacy policy without notifying a single person, using AI with clearly no guard rails or any safety nets... Or even asked if the users themselves wanted AI to look at their posts.
Still waiting for my questions to be answered, which you still continuously dodge unless I directly reply to you.
Hi @MaxTakeba,
I do try to make an earnest effort to reply to most tags, nearly all PMs and a lot of general conversations where I’m mentioned. That said, it’s easy to miss things here and there, and on top of that I don’t reply to everything that is directed to me. I hope you can appreciate that.
On the terms of service point, the prior version of the TOS that was in place specifically allowed for changes without notice, which is a pretty standard clause. The exact wording from the old policy was: “Please note that the information within this Privacy Policy & Terms of Service may change at anytime without notice. We recommend checking back often to stay current with the latest policy.” (https://web.archive.org/web/20250901010704/https://lowendbox.com/privacy-policy).
Every effort is made on LET to allow for free expression, fair and light touch moderation, and a place that feels comfortable and interesting for as many people as possible. That balance is not always easy but it’s something we take seriously.
On ServerVerify, I am really bullish about the concept. I think it’s going to be a big help to the buyer audience by surfacing useful data that wasn’t previously easy to find. We are literally working on it every single day and we continue to take in and react to all feedback, positive, negative and neutral.
Lastly, I put myself out in the public eye clearly. My real name, real picture, and real information about me are on my profile here and on my company site (wnyitservices.com). I’ve been part of this industry for over 20 years which is the majority of my life. I always try to lead with kindness and a high effort. That’s true in all things I do, including things like ServerVerify which are still a work in progress but which I’m working hard to make better.
We've released v1 of the new Advanced Search feature to ServerVerify (https://serververify.com/advanced-search):
You can get to this new feature by hitting the "enter" button while typing in the search box or clicking "Advanced Search" in the drop down as you enter text in the search box.
I think @MannDude has a point there. After all, all three sites are owned by the same company so it seems reasonable to expect that all three sites have the same stance on AI. Also and additionally I seem to have understood that you recognize that using an AI for ServerVerify was a mistake.
Plus, of bloody course people are pissed off to see an AI in relation to their work!
In other words: wouldn't it be the right thing to do to clearly state that AI is not acceptable on both LET and ServerVerify?!
Side note: I get it, you may not be able to just rip out that AI from ServerVerify, so maybe some patience from us is required. But at least make a clear statement, albeit with an "asterisk" for the time being.
Nice to see tangible progress! Thanks also to the team.
Just realized we probably need the ability to search by minimum and maximum CPU cores. Adding that now.
I don't even understand why you would connect the 2 situations. None of us want to have a conversation with generated text, but how does that apply to using it as a tool to massage data?
Uhm, do we really want to get served massaged data? My take is that benchmark data should be honestly shown (as in "what was gathered is what is shown") and that reviews should come from humans.
If by "massaging" you mean to compute relations like some VPS offering x% more processor performance than VPS B or "look how average performance has improved over [x] years, based on benchmarks from [x] well-known providers" that's OK with me if it's clear how it was calculated, like e.g. "as calculated by [name, version of AI] ".
As requested, the score of your review has been updated to a 5/5.
Additionally we've updated your review @emgh as you requested.
On September 30th, I updated this thread with a list of things we were developing and planned to release soon (https://lowendtalk.com/discussion/comment/4527533/#Comment_4527533)
Since then we have released the following new features to ServerVerify:
Things we are still working on right now:
Yea massage was a bad choice of word but I couldn't think of a better one really. There's going to have to be some degree of interpretation to fit "reviews" into a specific format no matter what though. I'm not sure there's a better solution than what's being done, taking reports of outlandish interpretations and "fixing" them.
We can argue all day about how it would have been better to start the site blank and entice people/providers to fill it in but that ship has sailed and I estimate approximately 0% chance @jbiloh is going to zero it out. I absolutely would not be starting from scratch. I probably would also have told several posters to eat dicks already too though, so.
Why do users need to point out this and then review ratings modified? First by AI and next manually back to what was set originally?
How many ratings are still tainted?
YES! I've begged you multiple times to run your posts through AI to clean up the grammar issues and make it readable.
You can have AI polish your message to be more clear, you can proofread the results to confirm your message is still being conveyed.
Edit: I took massaging to be for formatting and presenting the data better, not creating or modifying data from out of thin air.
It seems that you are not succeeding.
Submitted a benchmark of one of our H8 instances:
https://serververify.com/benchmarks/6d660b92-8d51-419c-9f6e-afea98920b25
I guess it'll also be #1 in VPS once it's approved
Impressive benchmark.
Single-core CPU performance should have at least similar importance as multi-core performance. Many real-world applications and hosting workloads still depend heavily on single-threaded performance, so multi-core scores alone don’t always tell the full story.
Also, 4K disk performance tends to reflect real-world usage much better than 64K or 1M tests. It’s usually a stronger indicator of how responsive a system will actually feel in daily use
We are looking at the cpu weighting for the algorithm (to weight single core performance slightly more than multi core).
We got a "C" grade on a 128 core hypervisor (Dual 7742's) with 12x NVMEs in RAID-10 and 2TB of RAM because Geekbench failed or didn't complete or something, so automatically the CPU grade is an "F".
https://serververify.com/benchmarks/77f06255-9555-43e6-b247-e864606ef3fc
A $7/year oversold VM on some ewaste Xeon will be rated higher on CPU if the test completes (even if it scores low) versus a beast of a machine who doesn't complete the test for whatever reason. A mechanism for applying an untested "known average score" would be better than a "0" in situations like that, I believe.
I probably stopped the test early anyway since I was posting it to show off the hypervisor in the Yabs thread, not knowing it'd be scraped and published and ranked some random 3rd party site later on haha.
Yes, I noticed this as well. Honestly, incomplete tests should just be excluded from getting a score.
This is changing. Over the weekend we pushed an update that no longer rates incomplete tests as an "F". Now it shows "N/A".
We are working to adjust the algorithm as well on how to handle those benchmarks (if at all). More thinking has to go into it.
i only see ads from colocrossing hosted companies
Ads are available to any and all hosting providers. We just signed up @alexhost -- woot woot.
Isn't that just a visual update and not an algorithmic update? Just changing an "F" to an "N/A" in the absence of an incomplete geekbench?
Had I known the the Yabs was going to be scraped and used for a 3rd party site to place a score that relates to my business I would have either not posted it on LET or simply did whatever needed to be done to see why the geekbench failed on a brand new hypervisor.
Good news is, you've made it possible to delete (manually, one by one) content that was scraped from LET, so props for that.
The N/A is visual and in the database of course, too. It's a first step towards improving how we handle benchmarks without GB scores.
We haven't updated the algorithm and how it handles 0/non-scores for CPU benchmarks yet. We are working on the best approach for that before we roll out something to production.
The optimizations are completed, and you should see faster page loading times on ServerVerify now plus lower CPU load because we optimized the moving top banner.
Before: Google Page Speed Performance Score: 42
After: Google Page Speed Performance Score: 88
We have dramatically improved page loading speeds and and reduced the CPU load of the moving banner. Let us know if it works better for you.
This should now be fixed.