META's New Code LLaMA 70b BEATS GPT4 At Coding (Open Source)

Join My Newsletter for Regular AI Updates 👇🏼
forwardfuture.ai/
My Links 🔗
👉🏻 Subscribe: / @matthew_berman
👉🏻 Twitter: / matthewberman
👉🏻 Discord: / discord
👉🏻 Patreon: / matthewberman
Media/Sponsorship Inquiries 📈
bit.ly/44TC45V
Links:
MassedCompute - bit.ly/matthew... USE CODE "MatthewBerman" for 50% discount
ollama.ai/libr...
ai.meta.com/bl...
huggingface.co...
huggingface.co...
ai.meta.com/llama
github.com/def...
/ 1752329471867371659
www.facebook.c...
ai.meta.com/re...
/ 1752013879532782075
Disclosures:
I am an investor in LMStudio

[@matthew_berman]

I'm creating a video testing Code LLaMA 70b in full. What tests should I give it?

[@santiagomartinez3417]

Metaprogramming or transfer learning.

[@johnclay7422]

those which are not available on youtube.

[@SteveSimpson]

Please show how to add the downloaded model to LMStudio. I add the downloaded model to a subdir in LMStudio's models directory but LMStudio doesn't see it.

[@so_annoying]

What about to write a kubernetes operator 🤣

[@OwenIngraham]

suggesting optimization on existing code, preferably having context of many code files

[@bradstudio]

Could you make a video on how to train an LLM on a GitHub repo and then be able to ask questions and instruct it to make code, for example, a plug-in?

[@lironharel]

Thanks for actually showing the errors you encountered and keeping it as real as possible! Great and enjoyable content❤

[@jacobnunya808]

True. Keeps expectation realistic.

[@TubelatorAI]

0:00 1. Meta's New CodeLama 70B 👾 Introduction to Meta's latest coding model, CodeLama 70B, known for its power and performance. 0:22 2. Testing CodeLama 70B with Snake Game 🐍 The host plans to test CodeLama 70B's capabilities by building the popular Snake game using the model. 0:25 3. Announcement by AI at Meta 📢 AI at Meta announces the release of CodeLama 70B, a more performant version of their LLM for code generation. 0:56 4. Different Versions of CodeLama 70B 💻 An overview of the three versions of CodeLama 70B: base model, Python-specific model, and Instruct model. 1:21 5. CodeLama 70B License and Commercial Use 💼 Confirmation that CodeLama 70B models are available for both research and commercial use, under the same license as previous versions. 1:40 6. Mark Zuckerberg's Thoughts on CodeLama 70B 💭 Mark Zuckerberg shares his thoughts on the importance of AI models like CodeLama for writing and editing code. 2:37 7. Outperforming GPT-4 with CodeLama 70B 🎯 A comparison between the performance of CodeLama 70B and GPT-4 in SQL code generation, where CodeLama 70B comes out as the clear winner. 3:25 8. Evolution of CodeLama Models ⚡ An overview of the various versions of CodeLama models released, highlighting the capabilities of CodeLama 70B. 4:21 9. Using Olamma with CodeLama 70B 🖥 Integration of CodeLama 70B with Olamma for seamless code generation and execution. 5:18 10. Testing CodeLama 70B with Massive Models 🧪 The host tests the performance of CodeLama 70B using a massive quantized version and shares the requirements for running it. 5:47 11. Selecting GPU Layers Choosing the appropriate number of GPU layers for better performance. 6:08 12. Testing the Model Running a test to ensure the model is functioning correctly. 6:43 13. Running the Test Requesting the model to generate code for a specific task. 7:27 14. Generating Code Observing the model's output and determining its effectiveness. 8:16 15. Code Cleanup Removing unnecessary code and preparing the generated code for execution. 8:40 16. Testing the Generated Code Attempting to run the generated code and troubleshooting any errors. 9:09 17. Further Testing Continuing to experiment with the generated code to improve its functionality. 9:15 18. Verifying CodeLama70b's Capabilities Acknowledging that CodeLama70b has successfully generated working code. 9:20 19. Conclusion and Call to Action Encouraging viewers to like, subscribe, and anticipate the next video. Generated with Tubelator AI Chrome Extension!

[@DanOneOne]

I asked it to write a program to connect bluetooth 3D glasses to a PC. it responded: It's not a good idea, because bluetooth is limited by 10m. Use wi-fi. I said: 10m is good enough for me, please write this program. -Ok, I will. And that was it 😆

[@EdToml]

Mixtral 8x7B was able to build a working snake game in python here...

[@efifragin7455]

can you share exactly which model was it? i also looking for model that can run on my pc i9-11k gtx 3060 16gb that i can code and make programs like snake

[@EdToml]

@@efifragin7455 I have a 7700 cpu with 64G of 5600 memory and rx6600xt (8g) gpu and am using rocm 5.7. The model is Mixtral 8x7G K_M_4bit (thebloke on hugging face). Using llama.cpp about 7G gets loaded onto the gpu with about 26G in cpu memory.

[@charlies4850]

@@efifragin7455Use OpenRouter

[@ordinarygg]

It worked, you just have bad driver

[@marcinkrupinski]

Cool, all the time we get better and better open source models!

[@emmanuelgoldstein3682]

GPT-4 ranks at 86.6 on HumanEval versus CodeLlama's 67.8. Meta used the zero-shot numbers for GPT-4 in their benchmark comparison, which is pretty dishonest.

[@michaeldarling5552]

🙄👆👆👆👆👆👆👆👆👆👆👆👆👆

[@romantroman6270]

They used GPT-4's HumanEval score from all the way back in March.

[@pcdowling]

I have codellama 70b working well on ollama. Rtx 4090 / 7950x / 64gb. The newest version of olama uses about 10-20% gpu utilization and offloads the rest to the cpu, using about 55% of the cpu. Overall it runs reasonably well for my use.

[@freedom_aint_free]

What is his context window ? How big is the code that it can generate ? Is it accurate ?

[@warezit]

🎯 Key Takeaways for quick navigation: 00:00 🚀 *Meta's Code LLaMA 70b Announcement* - Meta announces the most powerful coding model yet, Code LLaMA 70b, which is open-source and designed for coding tasks. - The model comes in three versions: the base model, a Python-specific variant, and an instruct model optimized for instructions. - Code LLaMA 70b is notable for its performance in human evaluation and its applicability for both research and commercial use under its license. 02:31 💾 *SQL Coder 70b Performance Highlights* - SQL Coder 70b, fine-tuned on Code LLaMA 70b, showcases superior performance in PostgreSQL text to SQL generation. - The model outperforms all publicly accessible LLMs, including GP4, with a significant margin in SQL eval benchmarks. - Rishab from Defog Data highlights the model's effectiveness and the open-sourcing of this tuned model on Hugging Face. 03:39 📈 *Code LLaMA 70b Technical and Access Details* - Introduction of Code LLaMA 70b as a powerful tool for software development, emphasizing ease of access and its licensing that allows for both research and commercial use. - Details on the expansion of the Code LLaMA series, including future plans for LLaMA 3 and the model's exceptional benchmark performances. - Mention of mass compute support for testing the model and an overview of the quantized version's requirements for operation. 06:11 🐍 *Testing Code LLaMA 70b with a Snake Game* - Demonstration of Code LLaMA 70b's capabilities by attempting to write a Snake game in Python using a cloud-based virtual machine. - Highlight of the potential and limitations of the model when generating code for complex tasks and the practical aspects of running such a model. - Transparency about the author's investment in LM Studio and the intention to disclose any interests for full transparency. Made with HARPA AI

[@theguildedcage]

I appreciate your disclosure. I intend to check this out.

[@auriocus]

The error comes from libGL failing to load and is clearly NOT in teh code that codellama wrote. It's a problem with your machine's graphics drivers.

[@BOORCHESS]

It would be great to get a price breakdown on how much computer you need to have to get in the door to run these locally and compare those ranges to the VM host options.

[@countofst.germain6417]

I just found this channel, it is great to see an AI channel that actually knows how to code.

[@scottamolinari]

Can I make a request? If you are going to highlight the text you are reading, just highlight the whole sentence with click and drag (which you do later in the video) and get rid of that highlighted cursor.

[@brunoais]

2:14: Not in the near future. AI is still programming worse than a junior programmer. Right now it's almost as good as a code monkey.

[@DoctorMandible]

AI will replace some Jr devs. Never replace coding entirely as you suggest.

[@dominick253]

I think if anything it just will expose more people to programming. At least that's the effect it had on me. Before I I felt like it was such a huge mountain to climb and now I feel like the AI can do the templates and 90% of the work then I can focus on getting everything to work together to actually make the project.

[@JT-Works]

Never say never...

[@starblaiz1986]

"Never" is a long time ;) When AI gets to human-level intelligence (likely this year, or at most by the end of the decade), what will stop it from replacing programmers?

[@EdToml]

Suspect coding will become much more of a collaboration. Less so with poor human coder sand much more so with good & great coders.

[@seriousjan5655]

@@EdToml Actaully, as a !£ years living from programming .... to write actuall code is the last thing. This models do not know what they are doing, that just had huge sets of probabilites. Last week I spent an hour with 6 coleagues discussing 3 options of approach from technical, economical and future advance standpoint. No, no replacement by AI. Sorry.

[@yaelcisneros1089]

This is it guys, the singularity has begun.

[@first-thoughtgiver-of-will2456]

Thank you for investing in LM Studio. I regard you as the most transparent AI Engineer journalist (for lack of a better term). Please keep up the important and quality work you've been doing for AI.

[@dan-cj1rr]

a dude on youtube click baiting everyone about AI isnt an engineer

[@matthew_berman]

❤️

[@ChrisS-oo6fl]

@@matthew_berman I take it that although you invested in LM studio you’ll still discuss other projects like oobabooga, open-llm, 02 LM Studio, hugginfacechat, silly, or the countless others if there’s anything notable to cover right? Or inform the public of the options and tools that are available. I do use LM studio but for some reason I personally don’t trust it especially with any uncensored model. Even as an extremely novice user I find it a little meh..so often stuck with oobabooga for most stuff. I also use other platforms for different use cases like my Home Assistant LLM API. Its human nature to become biased and unintentionally push, showcase or primary feature a resource which we are personally invested with. I personally prefer my creators to remain neutral with diverse content experiences unless o sought them out for their products.

[@MEXICANOGOLD]

META made a coding animal cooler than all the rest.

[@K.F-R]

1. Install ollama 2. Run 'ollama run codellama:70b-instruct' No forms or fees. Two or theee clicks and you're running.

[@MH-sl4kv]

I'm surprised it didn't refuse and give you a lecture on the ethics of caging snakes and making them move around looking for food in a little box until they run out of room and die. The censorship on AI is getting insane.

[@kevyyar]

can you create a video on how to setup these LLMs on VSCode with the extension like Continue, Twinny, etc? I have downloaded Ollama and have downloaded the models i need but im not sure how to configure them to run on the extensions on vscode

[@zr0af1nity]

I find it annoying how constantly "shocked" you are that the company who open sourced React, GraphQL, PyTorch, RocksDB and more would continue this philosophy in the age of AI. Zuck's complex. Maybe evil. Maybe just a misunderstood autistic. But his commitment to open source has been consistent for a long time.

[@RevMan001]

If I can download the model from TheBloke, why do I have to apply for access from Meta? Is it just for the license?

[@MetaphoricMinds]

Thank you! Remember everyone, download while you can. Regulations are on their way!

[@TheReferrer72]

Don't be silly. These LLM's are not AGI

[@michaeldarling5552]

@@TheReferrer72 You're assuming the government knows the difference!

[@TheReferrer72]

@@michaeldarling5552 Governments are much smarter than people give them credit for.

[@karolinagutierrez4383]

Sweet, this llama model crushs GPT4 at coding.

[@JohnTanner-o7s]

coding as we know it will be replaced and a new programming paradigm will emerge . This is absolutely wonderful. I'm glad I lived to see this. I've only have been experimenting with AI for about 2 months and I can't get enough of it.

[@voncolborn9437]

And then what? There will only be "programmers" that know how to ask questions and hope they get what they need? That doesn't sound very promising for the futhre of computing.

[@chineseducksauce9085]

@@voncolborn9437 yes it does

[@fuzzylogicq]

A lot of these models seem to assume everything is python, for most other low level languages no model can beat GPT 4. Yet !

[@makesnosense6304]

This isn't open source. There is no source for creating the model.

[@allenbythesea]

awesome videos man, I learn so much from these. I wish there were models tuned for c# though. Very few of us create large applications with python.

[@mrquicky]

It is surprising that the DeepSeek Coder 6.7b model was not listed in the Rubrik, though I recall Matthew reviewing & confirming that it did create a working version of the snake game. That was the most interesting part of the video for me. Seeing that it was not even being ranked anymore. I'm assuming a 70 billion parameter model would use more memory and perform more slowly than the 6.7 billion parameter model.

[@JohnTanner-o7s]

I'm running LMStudio on my system running Debian bookworm 12 and it's running good. Really want to be able to run models locally on this system to to my work when I'm home. Any ideas about local models etc. would be helpful

[@brunobergami6482]

"I think this will make programming obsolete" - Matthew lol why people still believe that full trust on code will be passed to AI?

[@chunbochen9976]

it looks good with simple piece of code, but how will it work and help in the real project or product?

[@vectorhacker-r2]

This is cool, but I don’t honestly think it will make programmers obsolete.

[@mickelodiansurname9578]

So the virtual rig Matt set up there was $4 USD for 3 hours... (from the link in his description) So lets say your working day coding is '8 hours' total where you need on demand LLM, and remember you could HOST all three Llamas and Mistral and Stable Diffusion I suppose, whatever open source model you want, and likely Llama 3 when it's released on that rig, but you couldn't run them all at once. You might get 2 running concurrently if you don't dump them both into RAM. Hell a few coders could get together at this point and do time share deal. BUT ITS FOR LESS THAN A BUCK FIFTY AN HOUR! I spend more on Coffee in a week than that!

[@Derek-bg6rg]

This video has me wishing I was a coding LLaMA too.

[@ManuelCastrellon-d2q]

If this thing starts making TikToks we are all doomed.

[@aboghaly2000]

Hello Matthew, great job on your work! Could you please compare the performance of Large Language Models on Intel, Nvidia, and Apple platforms?

[@LuckyLAK17]

...please a test with installation/access insteuctions will be great. Tks

[@elon-69-musk]

give more examples and thorough testing pls

[@zkmalik]

yes ! please make more on the new llama model!

[@DailyTuna]

Your videos are awesome!

[@matthew_berman]

Glad you like them!

[@CV-wo9hj]

Love to see you running locally. What Specifications are needed to run it locally?

[@footube3]

At 4 bit quantisation (the most compression you'll really want to perform) you'd need a machine with 35GB of memory in order to run it (whether its CPU RAM, GPU RAM or a mixture of the two). For it to be fast you need that memory to be as high bandwidth as possible, where GPUs are generally the highest bandwidth, but where some CPUs have pretty high memory bandwidth too (e.g. Mac M1/M2/M3 & AMD Epyc).

[@CV-wo9hj]

@@footube3 gah when I got my Mac Studio M2, I couldn't imagine why I needed more than 32 gigs 🤦

[@janalgos]

still underperforms gpt4 turbo though right?

[@voncolborn9437]

Matt, you mentioned you were using a VM from Mass Compute with the model pre-installed. Who are they? So to be clear, you were not running the VM locally, right?

[@stickmanland]

Me, looking on with my lovely 3GB Geforce GT 780

[@mazensmz]

Hi Noobi, you need to delete the old prompt before prompting again, because it will consider the old prompts part of the context.

[@avi7278]

yeah because compated to GPT-4 it has the intellect of a chipmunk.

[@mirek190]

mixtral 8x7b doesn't have such limitations. You can ask so completely different code later and is no a problem. I whing llma2 architecture is too obsolete now.

[@avi7278]

yay... another round of code llama hypeeeee! GPT-4 would never write a javascript function but then actually identify it as jsx (which is just pure garbage because obviously jsx is react template syntax that you can embed javascript into, so there's only a tangential relationship there) and then for whatever reason just decide to write it again in python. The hype is stupid, these open source models suck and are not close to gpt 4 no matter what benchmark they managed "beat gpt4" on and any serious dev who works with AI on production code bases daily understands this.

[@HUEHUEUHEPony]

Yeah yet another groundbreaking ai that beats gpt4, omg why do we still compare them to gpt4 if they're all better than gpt4, rly makes u think

[@avi7278]

@@HUEHUEUHEPony it truly is a mystery how a model could be really good at a popular benchmark but suck at literally everything else, isn't it? BeTtEr thAN GpT 4 derrrrr!

[@protossmr2344]

The model needs a huge amount of data to be trained. And GitHub and azure devops are 2 places people put their repos. Only Microsoft can feed data to OpenAI. So I don’t see Code llama can beat gpt-4 in term of coding

[@mirek190]

Interesting mixtral works much better witch coding of wizard coding 34b v1.1

[@MrDoobieStooba]

Thanks for the great content Matt!

[@dgiri2333]

I need olamma text(nlp)to sql query or nlm to django orms is there any llms for that.

[@dungalunga2116]

I’d like to see you run it on your mac.

[@fabiankliebhan]

Do you plan to test the new mistral-next model available on the LLM Chatbot Arena? It is crazy good. Possibly better than GPT-4.

[@vishnunallani]

What kind of machine is needed to run these type of models?

[@gaweyn]

but why in LM Studio, why not in an open-source project?

[@JohnTanner-o7s]

I don't see the link to mass compute.

[@cacogenicist]

Sure would be cool to be able to run this on my own hardware. So, what are we talking VRAM-wise? 92GB do it? ... sadly, I don't have a couple A6000s sitting around.

[@janfilips3244]

Matthew, is there a way to reach out to you directly?

[@matthew_berman]

my email is in my bio

[@ryshask]

First rule of SQL is it's pronounced S Q L. Not Sequel.

[@stargator4945]

I used a mixtral instruct model 8x7B and it was quite good, especially with other languages than English. So would this 70B model actually be better?

[@AliYar-Khan]

How much compute power it requires to run locally ?

[@MelroyvandenBerg]

The RAM was already stated. And regarding GPU. You need 48GB VRAM to fit the entire model. That means 2x RTX 3090 or better. You could also use CPU only, depending on the CPU but I think that will result into 1 token a second or something. Hopefully we soon have ASICS. Since I think GPUs can't hold up.

[@synaestesia-bg3ew]

​@MelroyvandenBerg, this is quite sad .

[@theresalwaysanotherway3996]

I'd be very interested to see this compared to the current best open source programming model (exlcuding the recent alpha mistral medium leak), deepseek 33b. As far as I can tell it's not as good, but maybe this 70b really is the new front runner

[@MelroyvandenBerg]

If you are in investor in the project, also put in the video on screen in the future. Not just in the description. ok?

[@1986hr]

How well does it perform with C# code?

[@hqcart1]

it's trained on pht, might not be as good for c#

[@nannan3347]

*cries in RTX 4080*

[@MrWizardGG]

On the other hand, I have a 4090 but still probably won't use 70b version because it'll be slower than 30b version

[@kyrilgarcia]

same but in 3060. There is no such thing as enough vram 🤣

[@anthonyshort8957]

I don't see the basis for the claim in the title. Running Code LLaMA 70b and GPT4 side by side with the same questions, every time GPT4 seems to produce clean working code and Code LLaMA 70b not so much. Have you personally found any examples where Code LLaMA produced a superior answer to GPT4?

[@endgamefond]

What virtual computer you use?

[@montserrathernandezgonzale6856]

Looks like GPT-4 is getting put out to pasture.

[@StephenRayner]

Not watched yet. But really want to fine tune this bad boy. This will be so nuts!

[@IleanaAlday]

LLama is super impressive at coding.

[@LanceJordan]

What was the secret sauce to "get it to work" ?

[@onoff5604]

yes please try it out! (and let us know the results of your experiments with snake please...)

[@vladvrinceanu5430]

bro llm studio i guess fu cked up something with new updates. i cannot run even old models on my mbp14 pro m1 ( m1 pro with hightes core count ) as i was able before. improvements to make: - Beeing able to use model for scientific propose as generating molecule formula and so on. ( there is not a single LLM scientific tool supported on llm studio even if the model is available on huggingface) - Fix the gpu metal for m1 mbp14, in fact i was able to use it now not anymore.

[@Leto2ndAtreides]

Wonder if the Macbook guy was running a quantized version or not. The maxed out M3 Macbook has a 128GB option also.

[@frankjohannessen6383]

unquantized 70B would probably need around 150GB ram.

[@dungalunga2116]

Would it run on an m3 max 36gb RAM ?

[@BlayneOliver]

That’s you just flexing 😅

[@william5931]

can you make a video on mamba?

[@sumitmamoria]

Which version will run reasonably fast on rtx3090 ?

[@michaelcdoty]

You've decided to do this anyways, but I want you to test code llama 70 b

[@K9Megahertz]

LLM's so far have been horrible at programming. Sure for cases like snake game or sort an array of elements in reverse order it can bust those out pretty fast. Why? Because there's probably hundreds if not thousands of examples of this code on the net, so it can certainly text predict a program similar to this. It's more or less copy and paste of what it was trained on. Something programmers have been doing for quite a long time already. However, if you give it a problem that it has not been trained on or has seen before, it just fails. I've got a couple functions I've asked ChatGPT to write and it just cant get it right. Example: Sort a randomized list of coplanar vertices in 3d space into either clockwise or counterclockwise order. It gets most of the way there, but not 100%. Also the solution presented fails to take into consideration an edge case that if not handled, will give incorrect results. So at this point, the programmer has to intervene and understand the problem and code the appropriate solution anyway. It's not a new problem either, I had to solve this 20+ years ago when writing a BSP compiler for the Quake (id software game) map file format. It's not a difficult problem to solve, but there's just not copious amounts examples floating around on the web for an LLM to train on it. LLM's do not think through code, they do not reason, they do not do logic, they cannot create new material and this is why they will never fully replace programmers.

[@hishtadlut1005]

Did the snake game worked at the end? What was the problem there?

[@michaelestrinone2111]

Does it support c# and .net 8?

[@hqcart1]

no, just phyt and js

[@michaelestrinone2111]

@@hqcart1 Thank you. I am using GPT3.5 with average success, but it is not up to date with .net 8, and I don't know if open-source LLMs exist that are trained on this framework.

[@hqcart1]

@@michaelestrinone2111use phind, it's online coding ai and free, its level somewhere between gpt3.5 and 4

[@osamaa.h.altameemi5592]

can you share the link for mass-compute? (the ones who provided the VM)

[@kenhedges]

It's in the Description.

[@jackonell1451]

Great vid ! What's "second state" though ?

[@MelroyvandenBerg]

Let's go Code LLama!

[@ReligionAndMaterialismDebunked]

Early crew. Shalom. :3 Noice!

[@researchforumonline]

What pisses me off is no one is really talking about if you can run codelama 70b using a vps server with 20gb ram 400gb disk and 5 4ghz CPU's, I can run lama 2 13 b but anything more does not seem to work, please provide a video for poor people who cant afford gpu's and for people with servers who have no gpu's which is a lot of people.

[@emmanuelgoldstein3682]

Usually those older server CPUs don't support AVX instructions, especially Xeons. I have a Dell server from 2013 with dual Xeons and 73GB of RAM but it's practically useless for local models since it can't run any of the frameworks that are built on these instructions.

[@blady87ziom]

​@@emmanuelgoldstein3682 Could You please elaborate little more? What services do you have in mind? I have lots of core i5 and i7 laptop's that support AVX/AVX2 instructions. Could it be possible to have them in some sort of cluster with Gigabit Ethernet?

[@vcekal]

Hey Matt, will you do a vid on the leaked early version of mistral-medium? Would be cool!

[@NimVim]

How did you manage to get a checkmark? I thought only 100k+ channels and pre-existing businesses could get verified?

[@vcekal]

@@NimVim I found a security vulnerability in KZbin which allowed me to do that. It’s patched now, though.

[@harisjaved1379]

Matt how do you become an investor in LM studio? I am also interested in becoming an investor

[@samuelcatlow]

It's on their website

[@miguelangelpallares8234]

Please test in Macbook Pro M2 Max

[@DevPythonUnity]

how do one bomecs an inverster in LLM studio?

[@michaelpiper8198]

I already have a setup that can code snake that I plug into AI so this should be amazing 🤩

[@TomM-p3o]

So you are an investor in LM Studio, perfect. Can you please tell them to allow increasing font size. My vision vascilates between good and poor and sometimes I'm having problems reading LM Studio text. BTW I'm seeing LM Studio release frequency ramping up 👍

[@reynoeka9241]

Please, you should test it in macbook pro m2 max

[@HaraldEngels]

Please test Code LLaMA 70b locally. Thank you 🙂

[@liketheduck]

TEST IT! :)

[@GaelNoh]

Llama is impressive!

[@fbravoc9748]

Amazing video! How can I become an investor in LMStudio?

[@BrianCarver]

Hey @matthew_berman, love your videos, this one sounds a little different. Are you using AI to generate any parts of your videos now?

[@matthew_berman]

Nope! What sounds different about it?

[@knowhrishi]

We need test video pleaseeeeee