The content argues against the notion that new AI tools like CLD code have rendered older automation platforms like N8N obsolete, instead advocating for a hybrid approach that leverages the strengths of both traditional automation and advanced agent-based workflows.
Mind Map
Click to expand
Click to explore the full interactive mind map • Zoom, pan, and navigate
Friends, hello everyone. So, has CLD code killed NV8 man? Of course not. I think many of
us have already seen popular YouTube videos about how people use code or antigravity to create
automations they had on NM using CLD code, antigravity, or other tools.
They spend literally seconds on it instead of hours, as before. But in reality, there are, naturally,
many exaggerations, omissions, and nuances here. And by the end of the video, we'll delve into this
issue a little, because, as with any issue in our lives, I think one should never go to
extremes or take a completely radical view. One must monitor
the strengths and weaknesses of each point of view, analyze them, keep them in mind, and choose
a middle path that leverages the strengths and mitigates the weaknesses. To
put it simply, we've essentially reached a crossroads that forces us to understand
the difference between AI automation, which we've been building constantly up until now, and agent-based workflows,
which everyone is building left and right. Naturally, there's no clear terminology here,
but what we were doing when we built automations on NVCm, Zapier, Make,
and other automation tools was essentially building applications with
tools, resources, and data to work with. That is, we defined what
the application used, what capabilities it had, what to do with the data
it received or processed, and sent incoming data as input, which was processed according to
these rules. Agent-based workflows operate conceptually differently. This truly is the next,
very powerful level of development and automation. Essentially, agent-based workflows have a set of scripts.
They can execute code, run commands in the terminal, and use the vast number
of libraries that humanity has written over time, install new libraries,
and remove old ones. Essentially, the user simply writes instructions that
determine which libraries might be needed, instruct them to install them if they're missing,
which scripts should be executed, and in what order. In other words, if we simplify things
as much as possible, AI automations only perform what we've taught them. Agent
workflows, however, can do absolutely anything Nerona can come up with. Our job is to curtail
its variability and creativity. As you can imagine, these are conceptually different things. Let's
look at how we can combine the best of both worlds. I think few would
argue that NM is incredibly good at integrations and authentication. Its ability to receive
data and send data to a huge number of tools is simply mind-blowing and works at
a very high level. It also includes the ability to quickly build simple automations in a visual editor and
quickly achieve clear results. And, perhaps most importantly, something we don't think about
when we talk about agent workflows. While our development knowledge is still limited,
NV CM provides excellent logging and executes its operations in a way
that allows us to quickly visually identify errors, correct automation, and
quickly deploy a new version. And perhaps the most basic, powerful thing NVCN allows
us to do very quickly is roll out our automation to the public and deploy it to servers,
which becomes a real challenge with limited programming knowledge if we want to deploy
our agent OFLOWS anywhere. Let's look at a real-world example. Let's say you
have a Telegram bot that constantly receives incoming requests. For example, people ask about
advertising on your channel or inquire about specific materials. You want to
quickly process all these requests, not respond immediately, but rather perform some complex analytics.
Let's say you want to do some manipulation, collect all the requests for the day in a specific
place, analyze what people are asking for, and then generate personalized messages in response,
perhaps generate some PDFs, personalized files, so that they are sent to
specific people who left incoming messages. So, if you have even a little
understanding of NV CMN, then you can create an automation that accepts incoming messages, incoming
If you 're using data from virtually any channel, which can, in principle, accept incoming messages,
for example, even transcribe them if they're voice messages, and store all these files or
incoming messages in a specified directory, then this kind of automation will take you 10 minutes. And then all the complex
analytics with generating PDF files, possibly other materials, generating tables, and
other things in N8 could take another couple of hours. But the combination of the two worlds
of AI automation and agent workflows allows you to perform all the very complex manipulations and tasks using
a neural network on a local machine. As you remember, in this case, the neural network is completely unlimited
, only by our rules that it must adhere to. And, accordingly,
we can ask the neural network to perform a task until it's completed,
using all the available tools it can think of. Let's take a look at how automation works right away
. I launch the execution and go to my Telegram channel.
Hi, I have a software development company. I'd like to know
if I can post on your channel, and how much it costs. I'm interested in 30-60 seconds somewhere around
the middle of the video. Thank you. I'm sending it. I see my automation processing the voice message
and saving the file directly to my drive. I go to my inbox, see a text file, open it, and see
my message: "Hi, I run a software development company. I'd like to know
if I can post on your channel, and how much it costs." So, as you can see, the automation
is incredibly simple. Its task is trivial: collect all incoming messages for the day, compile them into
a minodisk for subsequent analysis. Let's briefly review what it consists of. There's a switcher
that determines whether it's a voice message or a text message. If it's a voice message, it downloads
the audio file and transcribes it using neural networks. Then it combines
the message itself and the Telegram ID of the user who sent me the message. Then it converts everything
into a file and saves it to my drive in a specified folder. We can build this automation in
5-10 minutes. And then we can move on to our agent Overflows. This time, we'll be working
in Cloud CD, but the principles are the same. Whether you're working in Antigravity or the Gemini CLI cursor,
the principles will remain the same. We'll open Visual Studio CD,
launch the terminal, and add our inbox, which we use all day, where NV CM stores
all messages. Here we see the messages we've received throughout the day. Here's my idea.
I want Nerona to analyze all incoming messages and then generate a table for me. I want
it to be a table consisting of three columns: the ID of the user
who wrote to me, the response message I'll send based on the message
they sent me, and the PDF file on MediakiIT, which is customized for that
specific user. And I want all of this to be done with literally one command. So, we
type clД in the terminal. I press slash and see that I have my very first command, called
Review Messages. This is a custom command created specifically by me. Let's take a closer
look. By the way, friends, at the end of the video, I download all the automations on Envo Cement that we create on YouTube
and post them in my free Telegram group. You open this group, find
a relevant video you liked, and under it, find the NVM template. Download it,
open your N8N, and using those three dots, click "import from file," upload this file,
and your automation will open, and you can continue experimenting with it. And if
you're just starting out with NV CMEN, be sure to drop by the pro group. We have a fantastic
master class on NVC MEN there, which will quickly introduce you to all the basics. And here
in this pro group, next week we're starting to release a fantastic master class on ClD
code, which is absolutely mind-blowing. So, if you fly in before it's released, you can lock in
the current entry conditions. So, be sure to check the group information as well. Well,
I'm going back to my CLD code and my team. All the teams in CLD code are stored in the CLD folder.
Commands. Commands can consist of both a descriptive section and specific scripts
to be executed. This allows us to significantly save on tokens and Neural Network usage.
Let's open this command. I specifically created it in Russian for
clarity. This command scans the inbox, parses all messages, generates responses, and creates media kits,
including PDF generation. It then combines all of this into a single CS file. This explains
where to get the data for the media kit, meaning our channel's statistics. It also specifies which
scripts should be run, translated examples, the response generation logic, and the media kit structure itself.
This will contain a simplified media kit, but most importantly, it also includes a script
that will be executed by this neural network when the command is requested. You can see
that the script defines the installation of all the necessary resource dependencies that we
might need to generate a specific CS and a specific PDF document. So,
here you need to understand the key difference between using automation and Wordflow's agent-based implementation.
When I ask the neural network to execute this command, it will follow my command to execute
all the scripts. But if, for example, a script fails, or the terminal responds
that a certain library is missing, Python isn't installed, or some environment isn't available,
the neural network will try to resolve these issues itself, installing everything it needs, naturally
asking our permission, and will attempt to complete my task using all available means,
iterating several times, while we can only observe. So, I run
my Revw Messenges command. And let's see what happens. Yes, it loaded the required skill,
found all the documentation, and is trying to execute the commands. Look at this. It encounters
an error. After that, it makes new attempts to install dependencies, and succeeds.
It sees that the library for analyzing and creating PDFs is already installed, continues executing the command
, and terminates. He says, "Your CS file has already been created. Now all we have to do is open our inbox, scroll
down, see the relevant color file, open it, and see that it contains exactly what we asked for:
all the IDs that were written to us, all the messages we will be sending, and even the generated PDFs
that correspond to these requests. They are located here in the Media Kids brand folder. We see a huge
number of PDF files. We open a random one, and see a simplified diakit. Here we already have the ID of the partner
who writes to us, describes the pricing, channel statistics, that is, creates materials
directly in PDF files, which we can send anywhere. And, of course,
here we come to the basic question that probably arose in your head. You might
say, "Well, we can write the receipt of messages from Telegram right here, directly in the input code, that
is, build an application. " Or, conversely, we could process all text messages in N8N:
generating PDF files, generating CS files, saving them to disk, sorting them. I'll say, "Yes,
we could have done everything in N8N or in clot code." But in this case, we very quickly did what
we can do really quickly in NCMN. We focused on collecting all incoming messages, leaving us
with the ability to visually debug, make quick edits, and understand what's going on, even if we're
not developers. And we left the real work to the cld code. We want file generation to be
placed in the appropriate directories, deciding what should be stored where, where, and
which specific libraries to use to generate PDFs and other files. We want
to leave this to the neural network, which has much more comprehensive functionality and will
try to complete the task until it achieves a positive result. For us, as
people who are deeply passionate about automation, we certainly need to avoid following the hype and
shouting at the top of our lungs that N8N is dead. We need to take the coolest and most positive things from here. And,
of course, we need to start implementing code and agent workflows into our work, because
the potential here is simply limitless. As for what exactly can be done in cloud code and how, even
If you're a complete beginner, we'll be covering it in our masterclass, which will be released in
the pro group next week. The N8men masterclass is already on the channel. Drop in, take it, and
implement artificial intelligence in your work and your life. That's all for now. Bye, and see you soon.
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
