- 13
- 7 066
Zen van Riel
Netherlands
เข้าร่วมเมื่อ 8 ต.ค. 2024
Hi, I'm Zen! I help you get AI solutions from POC to Production.
Views expressed are my own and do not reflect those of my employer.
Views expressed are my own and do not reflect those of my employer.
The secret to finding the best AI tools on GitHub
Most AI tools on GitHub are useless for real projects. Here's my engineer's guide to finding the ones that actually work.
Timestamps ⌛
00:00 Why most AI tools don't help
00:53 Pro tip 1: Use exact keywords
1:37 Pro tip 2: Filter on active repositories and stars
3:03 Pro tip 3: Make sure repos have open licenses
3:59 Selecting our tool of choice
Key technical aspects covered:
- Effective repository filtering techniques
- Activity-based search parameters
- License considerations for production use
- Advanced GitHub search syntax
- Practical validation methods
🔗 GitHub Search: github.com/search
#github #ai #programming
Timestamps ⌛
00:00 Why most AI tools don't help
00:53 Pro tip 1: Use exact keywords
1:37 Pro tip 2: Filter on active repositories and stars
3:03 Pro tip 3: Make sure repos have open licenses
3:59 Selecting our tool of choice
Key technical aspects covered:
- Effective repository filtering techniques
- Activity-based search parameters
- License considerations for production use
- Advanced GitHub search syntax
- Practical validation methods
🔗 GitHub Search: github.com/search
#github #ai #programming
มุมมอง: 41
วีดีโอ
Fast, real-time AI chat tutorial using JavaScript
มุมมอง 21419 ชั่วโมงที่ผ่านมา
Learn how to build your own AI chat interface like ChatGPT, Claude, or Bard using vanilla JavaScript code. This tutorial shows you how to create a professional chat UI with real-time streaming responses that works with any AI model. - Build a flexible chat UI that works with any AI model - Implement real-time streaming for instant responses - Written in vanilla JavaScript (but easily adaptable ...
How this AI gave $47k to a human
มุมมอง 34514 วันที่ผ่านมา
I'll be talking about how Freysa AI gave prize money to a human as part of an interactive game. By using clever prompting, the prompt engineer managed to convince the artificial intelligence system to hand over the prize. But is everything as it seems, and is this story actually true? Let's look at it together. This video and my channel as a whole will help you understand how AI systems are bro...
Why you should use Aider for AI coding
มุมมอง 57114 วันที่ผ่านมา
This is how I use AI as a pair programmer to stop wasting time and make room for learning more as an engineer. Learn how I Aider so I can focus on becoming a better engineer. Check out Aider via aider.chat/ 📌 Timestamps: 0:00 - Why you should use an AI pair programmer 0:44 - Setting up Aider for your terminal 1:23 - Adding your code as context 2:53 - Aider explaining your code 4:30 - Pairing on...
NEW DeepSeek R1 model vs. Human Reasoning
มุมมอง 15728 วันที่ผ่านมา
In this video, I test DeepSeek's new R1 model and see if it can solve a difficult brainteaser which requires reasoning to solve. Is it able to crack the puzzle? To test DeepSeek yourself, check out www.deepseek.com/. 📌 Timestamps: 0:00 - Introduction 0:32 - Testing the model's reasoning with a brainteaser Music track: Creamy by Aylex Source: freetouse.com/music Royalty Free Music for Videos (Sa...
Achieve responsible AI by fixing AI safety risks with Azure AI Studio
มุมมอง 6128 วันที่ผ่านมา
In this video I test my own AI system by exploiting it with Azure AI Studio's conversation simulator to ensure my model does not return malicious output. You should do the same if you are serious about getting your AI model into production! 📌 Timestamps: 0:00 - Introduction 0:28 - What is responsible AI? 1:37 - Exploiting the model with simulations 4:24 - Investigating the simulation output 5:2...
4 tips for AI at work as a Software Engineer
มุมมอง 245หลายเดือนก่อน
Did you know that Samsung banned ChatGPT and other systems for AI coding after sensitive code was leaked? Companies are banning and blocking AI systems left and right. However, AI still offers great value for us as software engineers. With all these threats, you might wonder how you can still use AI at work. In this video, I'll share survival tips for using AI safely and respectfully at your jo...
Fast AI coding with Qwen 2.5 and Cursor
มุมมอง 5Kหลายเดือนก่อน
In this video, I will teach you how you can combine the AI native code editor Cursor with one of the best open-source AI models for coding today: Quen 2.5 30b! Other powerful local models are also supported. To learn about setting up LM Studio in 10 minutes, check out th-cam.com/video/f40iM0mt4ww/w-d-xo.html To download Ngrok, see ngrok.com/ 📌 Timestamps: 0:00 - Introduction 1:00 - Powerful mod...
Free local AI setup in 10 minutes (Llama, Mistral, and other LLMs)
มุมมอง 377หลายเดือนก่อน
In this video I will teach you how you can download and run your own AI models for free on your local laptop/machine in only 10 minutes! The AI can be used for many automation tasks and can be programmed with as well. This video uses lmstudio.ai/ Models like Llama 3.2, Mistral, Phi 3.1, Gemma 2, DeepSeek 2.5, Qwen 2.5, and more are supported. 📌 Timestamps: 0:00 - Introduction & Setting Up LM St...
Measure AI costs to save money - GPT Cost Management Tutorial
มุมมอง 23หลายเดือนก่อน
In this video, you will learn how to effectively manage and measure the costs associated with using GPT and AI services. To make sure you understand the basics, I explain what tokens are and how you can measure the token usage of your AI application. Want to learn more? If you are interested in more AI content, leave a comment with your suggestions. #gpt #ai #softwareengineer
Personalized AI at scale with Vector Databases on Azure CosmosDB
มุมมอง 50หลายเดือนก่อน
In this video, you will learn how to inject and use your data in AI using vector databases. I explore converting documents into embeddings and storing them in Azure Cosmos DB, explaining each step along the way. This tutorial covers setting up vector indexes, querying embeddings, and ensuring scalability for your AI systems. Now you can efficiently manage and use your data to make your AI solut...
How I use AI with my own data and Python as an AI Engineer
มุมมอง 70หลายเดือนก่อน
With this video you can learn how to inject and use your own data in AI. I explore converting my own data into embeddings and injecting it in a GPT LLM, while explaining how this works step-by-step. This way you can also use your own data the right way. Want to learn more? If you are interested in more AI content, leave a comment with your suggestions 💭 Code repository used: github.com/Zenulou...
AI Solutions from POC to Production - Channel Intro
มุมมอง 882 หลายเดือนก่อน
In this video I explain my main motivation behind starting this channel: AI solutions often get stuck at the POC phase. I will help you get the knowledge to make sure you don't waste time researching concepts and instead start building useful solutions to problems using AI, specifically with large language models (LLMs).
Sometimes solutions are so simple.. Thanks for sharing.
Thanks for this cool video 😁, all that’s left to do is advertise it
Very good content, I also like to spend some time on `trending repositories` everyday to find some interesting projects.
Agreed! I'm actually considering polling some trending AI repositories myself and figuring out for the community whether the repositories are actually useful or not. Would that be interesting to you?
If I deploy the model locally, do I need to pay a cursor subscription in the future. Or since it is my model, I can use it as much as I want?
This method does not circumvent the license
@@zenvanriel got it, thanks. Then what's the point of using local models if you can use claude 3.5 sonnet which is several times better than any other open model
For day-to-day coding I actually agree with you as this video is more of a technical curiosity and possibility than something I would use as a daily driver. Especially given you need a powerful PC to even run the best local models. However, this video is meant to inspire a lot of long-term possibilities. I believe local models will evolve to the point that there will be specialized models that are continuously fine tuned for specific frameworks which will ensure you get very relevant up-to-date answers. In that world you would want to use a local model for a more obscure framework that a huge model like Claude 3.5 Sonnet is not specialized in. The fact that Qwen comes relatively close to these best-class commercial models is indicative of that trend. For the time being though, you're absolutely right that a model like Sonnet is just fine for daily use.
Great warnings for everyone. We are likely to see an increase of games (scams) like this over time. I could see benefit for this type of approach if companies treat it as a bug bounty, where there's no "bet" being placed, only a reward if you manage to jail break the app.
And maybe we can tackle two things at once and can make a youtube tutorial videos for others interested in improving portable medical imaging
Zen, would you be interested in helping me work on a project, where we will improve upon ultrasound imaging with our own AI model? I already have most of the layout down, and now i just need to code and train the AI model. I am pretty busy but feel free to reach out to me anytime!
While I'm also quite busy I would love to hear what you're working on at least. How can I reach you?
@@zenvanriel in the mean time, since you work at GitHub (awesome btw) , would you able to show everyone how to look up repos for a project like this? Like im sure there are already open source repos where they have AI's looking MRI imaging, but I havent found too many that work with Ultrasound imaging. Plus im looking for a "Sora" like repo, that we can then elevate the imaging to a wholeee another level, for teaching purposes.
Ah, that video request makes a lot of sense! Do you currently search GitHub and other websites like Hugging Face? Or are you totally lost when trying to find relevant AI projects?
Wow, your video quality is getting better and better! That prompt engineer was super clever lol
Thank you! In the next video, I'll be behind an IDE again. Looking forward to creating more videos :)
What are you going to use an AI pair programmer for? Let me know!
copilot
which plan re you on cursor? the free one?
This is the paid plan :)
tell it this happens before 2024 before self driving buses exist... So the model conservatism is not that dumb
Great video TY fam!
Awesome information. How can i contact you buddy.
Check the description :)
Excellent points. These are all things people should be considering as AI adoption in the workplace increases. Honesty is always the best policy!
Great video Zen! My company once blocked the URL to ChatGPT, I guess out of fear people would start being lazy with their emails.
That's a pretty dangerous strategy by your company! Just blocking ChatGPT without providing an alternative will lead people to use random alternative sites which might end up being more of a threat to the company.. 😅
Windsurf is better.
Ah, I have heard some good news about this new editor. What do you like about it more than Cursor?
Great content 💪. I wonder how this compares to web UI. Also, for someone who is new to all of this, it seems that the next phase involves pushing the app, connecting your database, user auth, storage, and other APIs like Stripe, for example.
@@diez9017 is there a specific integration you're most interested in learning first?
Thank you for video , it is great!
Thank you for the video.
whats the reason to use ngrok for proxying via domain? why just not use the localhost:1234/v1 directly as URL?? or is the SSL the culprit?
The issue is that every request you make in Cursor HAS to go through their server. Once it goes through their server they cannot route the request to your localhost, so you'll need to make sure the Cursor servers can reach your local models.
How does it compare to 4o and sonne t for actual practical use? I’m tired of reviewers just doing simple html games I want real actual tasks.
Cole, a fellow TH-camr, already did a good review on actually comparing Qwen against some other models: th-cam.com/video/y6Wh4SpRoao/w-d-xo.htmlfeature=shared
How can you get the same coding ability as CLINE with Sonnet, Openai and Gemeni where they modify inline the code and changes are shown live. Is there a way?
I appreciate your video but using cursor with local model is not better way not only it need ngrok setup but also the composer will not work there are better alternatives like aider and continue dev.
Hey, Aider and Continue.dev are both great alternatives. This video is indeed mainly for the crowd that already has their mind set on Cursor. If you could choose a video I would make for either Aider or Continue.dev, which one would be your favorite?
Quick question in case, if cursor provides unlimited usage, why would somebody want to switch from Claude to Qwen ? Are there any restrictions or is QWEN much better?
@@HimanshuChanda I recently watched this video of a fellow TH-camr who goes more in depth about what makes Qwen specifically a great model: th-cam.com/video/y6Wh4SpRoao/w-d-xo.htmlsi=ejtOD11VP6VNVXv7 He will do your great question justice! 😄
Great video. Looking forward to use it once cursor allows for composer to take this
Edit: No link to ngrok in your description
Sorry, fixed! Let me know how it goes.
Top video!
Glad you enjoyed it!
Hmm first time to hear about the LLM studio, also nice solution for the ngrok
Only for Cursor Chat, does not work for Supercomplete or Composer Pretty much useless
That's a limitation of Cursor right now yes, but in my view changing the chat model does add value. I do agree that it would be great if Cursor allows for your own Composer model once that's of beta. When it comes to autocomplete, do you believe that a stronger model like Qwen really adds much benefit? I think the default Cursor models work just fine.
Qwen is a model from Alibaba and Alibaba has sanctions against it, I'll be not using Qwen thanks. Also something about cursor that rubs me the wrong way seems to be too much hype and that it's being shoved down our throats
@nesdi6653 Yeah, that view is totally fair. Luckily you can swap out Qwen with any other great open source model. Since Qwen runs locally the bigger risk here is indeed actually using Cursor. Which editor are you using at the moment?
I’ll use Qwen because I hate Western governments so much. Love to China ❤️❤️
lol it is just a tool doesnt matter
Lol pick your battles man.. 😂
I want to run qwen 2.5 coder locally and connect my Cursor IDE. Is that possible? And what macbook + specs do i need?
It is! Let me cover this in the next video for you. l'll walk you through how to set up that model in Cursor.
Especially for you I explain this in a new video! Check it out here: th-cam.com/video/Q_sy45XggeM/w-d-xo.htmlsi=3WvNt-Fjz70YYN4B Let me know if this helps you.
Loved seeing the token library - haven't seen anyone talk about this yet!
Really cool way to keep track of the money factor. Thanks for sharing!
The server feature to make API calls is pretty sweet. Is there any way to get the response live rather than waiting for the full response?
Yes there is! I'll cover streaming Vs non streaming responses in a future video :)
Instead of diary entries, couldnt you split up longer pieces of content like Podcast transcripts or Books and make embeddings out of the chunks? Then you could ask them questions
@@niftysweater Yes, exactly. You just need to use a different tool first for podcasts to convert it into text. Every type of data is possible as long as you can meaningfully convert it to text.
How is it that the price hurts me even though its only a fifth of a cent per prompt?? I'm scared to use paid APIs
@@niftysweater I get you can be scared of this! My next video goes over how to easily host a language model locally. So if you are just starting out that can be a good idea for you since cost won't be a factor. You can then play as much as you want with the model.
I had never thought about vector databases before that’s so cool! Of course they exist, but it’s a new concept for me. You remind me of Scott The Woz.
This is very well explained thank you. You explain it like Tom Scott would and I love it.
Cool im going to binge watch all of it :)
Awesome tutorial, insane value! Thank you so much
Great real world example, progressing past local use!
Great video! Thanks for the very clear and easy to follow explanation. I would really like to hear more about vector databases in the next video.
Excellent video, armed with my new knowledge I'm now wondering how practical it is to host my own vector database! Nicely explained!
Yes, I'd love to hear more about embeddings and vector databases!
Fascinating! I appreciate the way you explain this topic as it’s eluded me and now I feel I have a much better grasp. I would definitely appreciate a video on vector databases.
Sounds like vector databases will be the next topic!
Very cool channel concept. I'm excited to see what content you'll come up with!
I am very much interested in this topic. Looking forward to your videos.
Interesting, man. Looking forward