Now it gets to fight with Ragnarok Origin codes in the SERPS. Good due diligence. Ottomator should have stuck too, bolt.new reminds me of laundry detergent. Just picking fun. All of these tools are killing it so fast there's no time to quibble. I get it. I'm still painting on the walls of my cave compared.
True, Personally, I have never paid a company before in API because it is expensive for me compared to the capabilities. Deepseek really made an excellent offer and I topped up my balance with it too because it is really strong and cheap.
Not going to happen, not how these things work. I work and know people who work in big companies and they would not use DeepSeek because of data leakage. Microsoft Azure Cloud has been the winner here because of their data policies.
agree, this way they can lower their pricing and maybe add another plan like "Alternative Pro" with only cost $3-$5 with unlimited request since deepseek is very cheap comapre to sonnet
My storage array has ballooned over the past two years. Hard drive manufacturers are returning to platter drives just to meet the demand. Seagate is supposed to come out with a 32 TB platter drive soon. 1 huge step forward and a bunch of tiny steps backward.
It's the first model that achieves 3/70 correct answers on my internal CompSci set of test questions. I am pretty impressed by that =). That may not seem like a lot but it's full of questions about automata/ register machines/ ... for example transforming a regex to DFA to NFA and vice versa. It might be worth using it as coding assistant
I tried using roo cline (now roo code) along with ollama deepseek r1 hosted locally. The coding agent is not of much use. It keeps going into loops. Any suggestions on how to improve the experience ?
With DeepSeek and Roo Code, is there a way to tell the model not to learn more from our prompts and code? Something like the privacy mode in Cursor? Excuse my terminology if it's wrong, I am new to this.
I wanted to see more editing and problem solving with the plugin, I have tried a few of these plugins and they are great for some initial code generation but then are actually a hassle to use to try and work on it once the code-base gets a bit bigger but maybe I am using it wrong, but everyone just shows small demo apps which are cool but not super useful, I always end up just going back to a standard chat in the end.
ollama will work with the DeepSeek-R1-Distill-Llama-70B-GGUF:Q5_K_M locally on my M4 Macbook Pro, FYI. It's not as fast as the cloud versions but it's usable
Hello, I'm new to these LLM stuffs. I have a question, all the things you did in this video, are you creating a personal github copilot ? I'm reading docs about using LLM api, is it true that LLM api can just be used to make modified chatbot for companies or selling shop ? Thanks for answering.
I tried the R1 7B QWEN, and I benchmarked in some math problems I had. It is really nice. On the downside, is very verbose which makes it also very slow.
hey i'm actually pretty new to code generation and all of this, can you help me as a starter to know and understand how these things work ? it would be very much appreciated! thankyou so much in advance, this video is great
FYI: Roo Cline is now Roo Code. This new version was the renaming. Long live Roo Code
Now it gets to fight with Ragnarok Origin codes in the SERPS. Good due diligence. Ottomator should have stuck too, bolt.new reminds me of laundry detergent. Just picking fun. All of these tools are killing it so fast there's no time to quibble. I get it. I'm still painting on the walls of my cave compared.
I tried the R1 7B Ollama model with Open WebUI and it's good. It's great to see competitors to OpenAI's o1 model.
It's Qwen 7B R1 distill. Ollama fucked the naming. You're NOT using R1.
@@MrCorpsy6 Yeah upon checking the table in the Ollama page it says `qwen`. My bad.
OpenAI will be under a lot of pressure now to release something much better or else they will lose clients to deepSeek. Looking forward to it
True, Personally, I have never paid a company before in API because it is expensive for me compared to the capabilities. Deepseek really made an excellent offer and I topped up my balance with it too because it is really strong and cheap.
I think they will just ban deepseek in usa
Not going to happen, not how these things work.
I work and know people who work in big companies and they would not use DeepSeek because of data leakage.
Microsoft Azure Cloud has been the winner here because of their data policies.
i mean they have o3 but ofc this wil cost 100% more than 200$ in the future
They just announced 500 billion investment -- im sure ClosedAI will come up with something 😉
Windsurf or cursor needs to partner with deepseek to make custom model
yes
😊😊@@mr_ribin_123
agree, this way they can lower their pricing and maybe add another plan like "Alternative Pro" with only cost $3-$5 with unlimited request since deepseek is very cheap comapre to sonnet
For sure.
Yessss tag winsurf please
I switched yesterday.
Curious what will happen next week.
Switching ai's like my socks #day289
😂😂😂
My storage array has ballooned over the past two years. Hard drive manufacturers are returning to platter drives just to meet the demand.
Seagate is supposed to come out with a 32 TB platter drive soon. 1 huge step forward and a bunch of tiny steps backward.
LMAO yes, Only y'day I was searching the best model in the leaderboard, and here we are now :P
It's the first model that achieves 3/70 correct answers on my internal CompSci set of test questions. I am pretty impressed by that =). That may not seem like a lot but it's full of questions about automata/ register machines/ ... for example transforming a regex to DFA to NFA and vice versa. It might be worth using it as coding assistant
Finally, started my day with your video & coincidentally became the first one to comment 😂
Great one!
Cline has a plan/act mode now that would probably work the same.
Is it work for now ?
Thank you for the video, always good to find new updates and information, may i ask what you use as your whiteboard?
I would love to see this using ollama, Deepseek-R1 + RooCline, step by step all Local.
Me too !
same here guys 💪
I AM USING DEEPSEEK FOR EVERYTHING
how can I reference, with @, to a database, table, with Roo Cline?
128k context? Isnt that a limiting Factor?
Nice init. Thanks mate.
what do you think of the codeium extension?
I tried using roo cline (now roo code) along with ollama deepseek r1 hosted locally. The coding agent is not of much use. It keeps going into loops. Any suggestions on how to improve the experience ?
With DeepSeek and Roo Code, is there a way to tell the model not to learn more from our prompts and code? Something like the privacy mode in Cursor? Excuse my terminology if it's wrong, I am new to this.
I tried using Roo Cline with DeepSeek R1 Distill Qwen 7B running locally in LM Studio - but Roo Cline cannot access MCP tools. Not great.
Which model size do you use ?
I wanted to see more editing and problem solving with the plugin, I have tried a few of these plugins and they are great for some initial code generation but then are actually a hassle to use to try and work on it once the code-base gets a bit bigger but maybe I am using it wrong, but everyone just shows small demo apps which are cool but not super useful, I always end up just going back to a standard chat in the end.
Awesome, thanks man ❤
Its Roo Code now and not Roo Cline btw.
nice video thank u ! i have a problem : MCP hub not available, any solution ?
imagine how o3 will be like 😳
Great tips!
ollama will work with the DeepSeek-R1-Distill-Llama-70B-GGUF:Q5_K_M locally on my M4 Macbook Pro, FYI.
It's not as fast as the cloud versions but it's usable
Hello, I'm new to these LLM stuffs. I have a question, all the things you did in this video, are you creating a personal github copilot ?
I'm reading docs about using LLM api, is it true that LLM api can just be used to make modified chatbot for companies or selling shop ?
Thanks for answering.
I tried the R1 7B QWEN, and I benchmarked in some math problems I had. It is really nice. On the downside, is very verbose which makes it also very slow.
Verbose is what I wanted in the first place
I'm using DeepSeek's API . So what changes do I need to make for continue extension to make it use the R1 model as well ?
You need to use the deepseek-reasoner model
hey i'm actually pretty new to code generation and all of this, can you help me as a starter to know and understand how these things work ? it would be very much appreciated! thankyou so much in advance, this video is great
you can deep dive into his video, its about a years ago he explaining everything. just dig the video until you found it
@aculz which video ? Could you share the title or something?
Now inde game developers can make games easily 🎉❤
How is that Api free
use Chinese phone number
I don't see free 10$ credits on those sites
I switched to R1 for my AI agents because the value proposition is just too good to pass up.
Does openrouter has R1?
yes
Make the video about new cline 3.2
first