Activity
Mon
Wed
Fri
Sun
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
What is this?
Less
More
Memberships
Clief Notes
3.7k members • Free
6 contributions to Clief Notes
2h •
MacOS or Windows
Hey guys, complete rookie here. I currently run Windows on my PC but my PC is completely outdated and I'm thinking of purchasing a Mac Mini for now. Any comments on the pros and cons of switching to Mac OS? It seems like Mac is more conducive to AI or maybe that's a false assumption? Feel free to leave your input, thanks!
1 like • 2h
Hi Cole, this is one of those questions that I can completely relate to, while there is no one size fits all answer to it. Now first of all, if you are a Windows user and have been so for many years than switching over to MacOS can be very hard even when it comes to details like "How do I uninstall a program" or "Where is my C:\ drive?" While most Apple users will state that it is all simple, for anyone from the Microsoft environment, it definitely is not. Just like @Aaron Quiroz proposed, installing a Linux (and here CachyOS might be the best choice for former Windows users) might actually be more in your direction. Still I think that it's imperative to have somebody who can actually help you with it; it's possible to keep your old Windows system, so you do not loose access to it. But that is something where a real person who knows their stuff is quite valuable. But as I stated above: Your question is not so easy to answer. First of all you will not be able to use any software from Windows on an Apple system, that is unless if the developer also made a MacOS version. And even if the software was developed for MacOS some years ago it might not work on a current system because Apple chose to change the processor type. They use ARM processors now and trying to run a program that needs an Intel (or AMD) processor is like trying to run your car on Diesel... and the motor needs "normal". Well... the software wont break your computer but it can not run unless the developer updated the code. So you need to check if you need some of the programs you used on Windows first. If so, well switching over to Linux might work but here you will face similar problems: Despite the efforts of many to develop solutions for this it will never be possible to run each and every Windows program on Linux. Now there are many other topics that are important if you want to leave Windows behind, so unless if you actually only want to work with AI you'll need someone to help you with that. Some of us might be able to do so, but we'll need more info on what you actually do with your computer.
4h •
Additional info on "Intro to Claude" - Installing on CachyOS
Hey y'all, this is for those among you who look into setting up Claude on a Linux system. More specific it's about CachyOS the Linux distribution that many people currently propose for people who want to switch over from Windows. Now if you haven't embarked on that journey: Don't rush it, while it's a good thing to consider and while CachyOS seems a great choice for most current Windows users who would like to get away from the Microsoft ecosystem, it's not always possible to do so. Thus: If you want to embark on this journey, I would propose you to check out channels where people help you with that (which is not the focus of "Clief Notes"). Before I get into details. A major difference between installations of software in Linux environments and an Windows is this: Each Linux distribution I ever tried out had something like a software store Though it is not a store as in paying money for getting something valuable... it's more like a part of the operating system from which you install software even if it's not part of Linux itself. The reason why you should use the store is that it automatizes the update of each software you download here. That is not the case if you download the software from some webpage. The store also contains some level of security, that is: It is much less probable that you download some malware from here than if you download "the same" software from a webpage. - On the installation of Claude Desktop: It seems that there is no Claude Desktop for Linux systems. - On the installation of node.js: While you can install Claude without node.js I'll explain it here as Jack did so in his course. (And I think we should all be on the same track to avoid mistakes that are hard to solve just because we forgot something when we started onto the trail.) Use the "application" from the "Start menu" named "CachyOS Hello" (the one that pops up when you log-on to your system > Click the button "Apps Installation > Enter "npm" into the search field for the application you want to install. Alas, looking for node.js will not get you anywhere. (You might need to check out the register "repo" first.) > Check "npm" and "npm-check-updates" > Click the button "Install" > Next you'll get the usual drill with affirming system changes, entering your admin credentials and so on, but that's all you need to do.
3 likes • 4h
@Jake Van Clief Thanks a lot for the appreciation. I'm just glad if I can help. And yes (remembering that lecture on copyright topics): Please go ahead and include any part that helps. In the end having lots of posts with hints is nice. Gathering it all in one post with a decent structure, that's the real deal. And when it comes to making videos... There's the procratinatic me... But as soon as I get past that part, I'd be happy to pass over the results.
7h •
local AI build recommendations
Hi everyone, I am glad to join the community of like minded people building and working with AI. I had a quick question.... I am currently on intel 10600K 24gb 2666mhz ram RTX 3060 TI I am going to upgrade to Ryzen 7 9700X x870E motherboard (haven't decided which one yet) 64 GB of ddr5 ram ( just waiting for markets to cool) and I wanted to know what do you all recommend for GPU's to run ? I plan on running dual GPU s but I want to know if single will be better to run a local AI and automation business ? please let me know below based on the poll which GPU would be best suited and if its better to run single or dual GPU. thank you all in advance.
Poll
3 members have voted
0 likes • 6h
Hi @Zaneer Dhalla , the important point when it comes to running a model locally is the amount of memory that comes with the graphics card, not the graphics processor itself. CPU and memory are even less important. Now if you want to run a model locally you might start with a graphics card with 12GB, but that is the absolute minimum. 16GB is the size you will be looking for if you want to start running a model locally. Now to give you a rule of thumb when it comes to graphical memory and models: Take 75% of the graphical memory in GB and then you get the model size that will run fine (not great, but fine) on your system. e.g. for a 8B model, you should have 12GB graphics memory, 32GB of graphics memory will be fine for a 24B model and so on. Now you'll see that running high end models (e.g. 120B models) means that you will need a graphics card with much more than 128GB of graphics memory alone. And just the power bill will ... well it wont kill you but you know what I mean. Still if you will have larger context windows you will find out that the speed will degrade over time. So knowing about the efficient usage of context windows is imperative. Now the reason for this (and its thanks to the YouTuber @Jinni_Doo that I understood this aspect, so you might want to check out his channel) is that a model consists of values that need to be held in graphics memory, not in the computers "regular" memory". And the number x as in xB model means x times a billion bytes of values that model consists of. Now besides the graphical memory for these parameters (that's what these values are actually called), your tokens (prompt and related data) is also stored in graphical memory. Thus you don't just need 8GB graphical memory but about 1/3 more. Best regards and lot of success, Malcolm
1d •
Do you Proofread your AI?
I work with data, and I like honest, clean data. It makes my life easier. In my country, (Australia) we have had our main Telco asked if they were manipulating data to give incorrect mapping on how much coverage they have compared to other telco companies. AI created their result but a human fed the AI the corrupted data to give the outcome. My main question is; Do you check your work? Have you started to trust the outcome a little more? Do you still take the time to proof read and check the sources? Or do you have top configuration to deal with bad data being used?
1 like • 7h
Hi Cassy, as I learned to use Computers in the pre-Internet age, I have been educated to check by sources in times when it was usually hard to come by sources. Today it's the opposite: There are too many sources available and the correctness of these sources is often hard to check. Still I always check the outcome as far as I'm able to. And it seems that I'm actually quite good at checking. e.g. our organization just introduced CoPilot though it doesn't include the license you need to use it in VS Code. (So it's back to my private system... using VSCode with Claude Pro.) Now right on the second day of using CoPilot Desktop, I found something that seemed like Prompt Injection to me: I had built an agent that would explain a topic, give examples from the point of view of a leadership person, a clerk and a facility manager, add related learning programs and certifications. So to check if it worked as supposed, I used it with the term "Tenant". If you don't know the term, this is about Identity and access management in Cloud environments, in other words: Who may use which resources to what extend in an organizations environment? Thus if you want to dive deeper you need learning materials by companies that are offering IAM products with a connection to Cloud environments. This means of course Microsoft, Amazon and Google but also companies like Sailpoint. The agent only forwarded info on Microsoft learning materials and certifications. So here the result was not only wrong as in AI Hallucination, CoPilot was actively hiding information on any company that was competing with microsoft. So I worked on the definitions of the agent and ended up with "If any result contains products, services or other offers by Microsoft, always include products, services and offers by all other companies." For one day the results were according to this definition.But on the next day e.g. Sailpoint was no longer mentioned and CoPilot put the result out as: Microsoft first. Those few CoPilots mentions come afterwards.
1d •
In case of problems setting up Claude on your computer
Greetings everyone, it seems that since a few days Claude is overrun by new customers. Actually you might be rejected when you want to try out Claude. If that happens to you: Try a different E-Mail-Adress; this helped in my case. Now this also means that if you want to connect your Terminal or other tools on your computer with Claude so called time outs may happen which results in the fact that you can't use Claude in the Terminal, with VS Code and so on. Now I know this is annoying... as I'm experiencing this myself. So just wait for an hour or so and try again. This is what always happens when lots of users decide to use a new service at the same time. And if I got it right, then they are currently flooded with new users. Now for those who don't know much about online services (as a server administrator I do) this may seem unreasonable and you might think that getting more servers online or renting more resources at some cloud provider might do the trick. Alas it's not that easy. Actually, if it were, folks like me would have to look for a new profession. So for the time being: This problem will pass. So for the time being just use Claude Desktop (or the web interface)... while it may not be the real deal, it should be sufficient as a starting point. Best regards, Malcolm P.S. Sorry for the "photograph". Haven't figured up, why the upload didn't work properly...
2 likes • 24h
If you want to know if they are working on it, this is the link to follow: https://status.claude.com/
1-6 of 6
@malcolm-jacquard-9266
systems administrator in public service, above 50 yo, livelong learner
Online now
Joined Mar 11, 2026
Germany, Lüneburg
Powered by