On this channel we think, plan, and build.
Right now on the channel, we're on the path to evolve into Agentic engineers.
Engineers that build software that works for them while they sleep.
Here are principles of my engineering philosophy and ideologies this channel holds as facts.
- Avoid hype, focus on real valuable tools, technology and products.
- Build real products. There are enough news, hype, trend channels. Let's make something real.
- Listen to feedback but always think for yourself.
- Use the best technology for the job period full stop.
- Keep learning, forever.
- Do what you share. I don't share anything on this channel I'm not betting on with my time, energy and money.
- Cancel out the noise and focus on the signal of value creation.
- Great things happen in the flow. Look for it everyday.
I'm not the perfect programmer, designer, or creator but to succeed you don't have to be perfect, you just have to try, over and over in success or failure.
Пікірлер
wow had no idea curser had this.....
I watch all KZread videos at 2x speed, like many I assume, but, unfortunately there isn't a feature that also shortens the time between words, for some KZreadrs. Because. I. would. really. like. to shorten. the time. between. Dan's. words.
Are you terminally ill? What’s the rush, chill out and enjoy the video
Someone did a VS Code / Cursor extension for aider. It will know where your files are to edit them. To open it when installed, from the top command input enter >open aider Or do crtl+shift+p then write open aider and enter. It will open aider in your IDE terminal with the right path to files when you use add/.
This is escalating pretty fast 😮
Phenomenal
Cool 👍👍👍,thanks a lot .
Aider and ClaudeDev have entered the chat. I like Cursor. The nicest thing is not hitting token limits as quickly but I’m an insane developer… I mean I’m literally insane. Here is an example. I used Swagger to generate a PHP client of the entire OpenAI api using their recent spec and am already half done converting it to PHP 8.3 with strict typing. Earlier, I created Pydantic in PHP and translated the Ollama Python client to PHP. Do you realize how much code that is?
are you saying just in general you're an insane dev, or are you saying with the assistance of these tools you've maxed out your code gen ability?
love it. Unfortunately i'm the co-pilot for Cursor. lol
I feel like aider is streaks ahead of this hey? It's free, only pat for the API, and it has three features already, can yo usee yourself using Cursor composer over aider?
I was going to ask the same question. Aider seems to cover this already.
@@goforit5 And with far less pointing and clicking. Hardly a need to pop open a text editor, let alone an IDE.
My thoughts exactly. Aider handles multi file edits really well. There are still some quirks but it’s night and day better than this.
100%. I believe this is a paid ad by Cursor.
Try using cursor and composer along with a voice transcriber tool (such as SuperWhisper). You can just talk out loud to code entire apps!
Just use aider already ? I occasionally use voice prompts but people react odd when you start to talk to your computer ...😂
If I had a company I would hire anyone here 😅
Top content, I'm an avid learner yearning for more quality videos.
I have been using this, Claude Dev and Aider Claude Dev has already implemented prompt caching. I sent you an email. Do you do consulting?
ClaudeDev is great. 🤘🏻
Love the content Dan! Always waiting eagerly for your next video's insights, keep it up.
Love your content. BTW, what do you think of agent Q, swe agent, sakana AI and genie?
I started using markdown but after looking over the anthropic workbench I started using xml. Havent looked back.
I am a huge fan of 4o-mini for its effectiveness vs cost. What if I wanted to use prompt chaining to make a chrome addon, would I modify the 'Minimal Prompt Chainables' gist and use that? Or is there a more abstract, universal kind of template
Same here - 4o-mini let's you do so much for so little. Yes, you can modify Minimal Prompt Chainable to your use case.
that's neat, I've been trying to get TTS to work for open-interpreter which I use a lot with gpt-4o-mini
😮👏🏻👏🏻👏🏻
BRO you are good teacher
This is awesome, being a loyal fan of your work. You are a great teacher, thanks for everything.
Dan. Great video.. could you do a quick start video? Setup + tips, using a basic application. Include how to build your code dunctions by referencing other existing code from other folders or repos. That would be really useful.
As cs first year scares me
Do you realize OpenAI and other monopolies are doing this to become the largest private data registrars in the world? Hence why OpenAI is not even open. Their business is data not Ai. You thought fb was bad? Oof. Your data. All data. Stop using this crap. Use free and local and stop promoting these companies.
What’s the other channel?
I know this isn't the point of this demonstration, but who wants an agent that can make the worst images possible?
Those are some really crappy images.
agreed lol - do you think openai has given up on image editing? This agent will be a lot more useful with a flux pro function.
thank you a lot for all your valuables videos
Thank you for demonstrating this for us. I am a writer and a digital artist. I have no experience or ability for writing code. Yet as a creative I would like to have an agent that could assist me in promoting my creative work on various social media outlets. I see this as a huge step forward for employing an llm to act as a promotions manager on my behalf. It is my hope that this process will become super simple in the near future.
What about calling two functions ar the same time
yeah, but I kinda don't want to use a company thats about to develop AGI on a brandname that is in itself a lie. They are ClosedAI, and call themself OpenAI.. So I want to learn opensource LLMs
One thing I noticed is this is very accurate with just a few tools, but struggles with anything more than 5. Thoughts?
I haven’t tested/experienced this, but along the same lines i am wondering whether the LLM should be at the „top“ of the toolchain, as it is good for some things, but too complex for others. Maybe the way is then to start with the toolchain and then see what we can use LLMs for and where other solutions are more efficient
Here’s how it could work: You can set up the assistant with a “crews” tool, where you define different crew types, their keys, and what each crew is responsible for. The main assistant analyzes the task and decides which crew to assign. You can configure these workflows using frameworks like LangGraph or CrewAI. Once a crew is selected, it starts the task, no matter how complex. The crews can communicate their status back to the main agent during the process, keeping you informed. Finally, the main agent compiles the results, ensuring a smooth and transparent workflow.
And what makes it beautiful is these frameworks support human-in-the-loop interaction. If you’re not satisfied with the output of a certain step, you can provide feedback, and the agent will restart from that point in the flow.
@@bartolli74 human in the loop has its purpose. But accuracy is key when doing automation.
@@DanielBowne I agree! I was trying to explain that, in this use case, it’s about a personal assistant performing tasks on demand. For example, it could pull your new emails, read them to you, and then you could instruct it to reply. This would start a process where the assistant drafts a reply, incorporating your feedback in the loop. The accuracy of the output depends on how we design the tools. My approach is always minimize reliance on the LLM’s general knowledge, focusing instead on logic defined by me, data fetched from trusted sources and databases.. llm just decides which function/process to call. And you can always implement custom validation and guarding methods to make sure output is what you expect to be.
Did you try Elevenlabs Websockets? Where you stream the output directly into Elevenlabs and an immediate response comes out?
Thanks for the video. Did you mention you have another channel?
I've switched from ChatGPT-4, to Claude 3.5 Sonnet, but now have been trying Gemini for coding due to it's recent Leaderboard score on lmsys Arena, and finding it to be pretty good at coding. Have you tried it? What do you think?
I know someone you did a similar agent, to lower the cost it's 100 % local. You dowload rhe LLM, Whisper. The TTS is from Included Windows voices and lately he added the possibility to open a GUI interface with Streamlit (with full option like choosing models ect..) . No need to do enter on keyboard to execute thask because it a continual conversation and it's always listenining unless you ask to stop. It cost nothing to chat with it and make it execute actions. (The only problem, the guy ask to get 5000 youtube subs. to make it open source on Github.) I think you can do the same with your incredible skills and AI knowledge. Question. What do you thing about Open Interpreter ? (Open source on Github)
how does this intertwine with your previous XML video? Can the structured outputs achieve 100% accuracy with XML? Are you switching to .json now?
You can use both. Your prompt can be structured in XML while giving output examples in JSON format that you then use structured output on to force precise JSON outputs. XML is still king for high accuracy prompts.
wondering if you are taking on any project work, let me know how to get in touch. As always, love your videos and #1 release I look forward to each week.
Possibly starting 2025 Q1. Let me know what you're looking for. Email: [email protected]
So do annoying voice !
I started to implement this in my apps. Works Awesome; just very slow.
have you tried using gpt4o mini instead of gpt4o? maybe that could accelerate your usecase
@IndyDevDan What tool do you recommend to build low-code/no-code multi-agent orchestrations? Preferably a tool that integrates well with the Microsoft ecosystem. Do you have a class/video series?
Always impressed with your videos.
On top of that, could be interesting to provide an xsd (xml schema definition) so that the response format is fully predictable.
Love the content but I think it would be helpful to at least mention relative costs. You've said in the past that the API costs are worth the investment. You are likely correct but probably still worth mentioning. Thanks again for the content.
Amazing content - keep up this combo of practical + high level videos 💪
Code base will become full of bug and impossible to maintain, no one will be expert of his features and will be able to tell how to fix an issue without even looking at the code base
This is why I love your stuff -> "I think one of the most important things you can do in the age of AI is make sure that your information diet is as clean as possible. There's going to be so much content generated; there already is so much content you feel on a daily basis. You know what this is like. I think it's important to build and use tools that help you filter out the noise and keep yourself in a low-noise, focused environment."
Always great insights, need to give promptfoo a shot!
Lovely
What would be the main difference between the custom voice assistant and the OpenAI voice mode?