It is probably the most important word when it comes to large language models (lllms), with the release of chatgpt. Chatgpt has been made so successful, mainly because of its convenience Pre-training Opena did, making it a powerful language model.
After that, Frontier LLM Labs started measuring i Training in the background, With good guidance and good planning and RLHF, where the models start better in the following teaching and performing complex tasks.
And just when we thought LLMs had reached a plateau, we started doing it measurement of time With the release of consulting models, where the money is spent Thought Tokens gave a significant improvement in the quality of the results.
Now I argue we should continue this measurement with a new measurement paradigm: usage-based ratingwhen measuring how much you spend on LLMS:
- Run multiple code agents in parallel
- Always start an in-depth research on the topic of interest
- Start the workflow
If you don't praise the agent before lunch, or you go to bed, you're wasting time
In this article, I will discuss why scaling LLM usage can lead to increased productivity, especially when working as a program. In addition, I will discuss some strategies you can use to measure the effectiveness of your LLM, personally, and in the companies you work for. I will keep this article high, it aims to promote how you can make the most of AI to your advantage.
Why you should rate the use of LLM
We've already seen scaling have great potential before with:
- Pre-training
- Training in the background
- Measuring time
The reason for this is that the more powerful the computers you use for something, the better quality output you will achieve. This, of course, assumes that you can use a computer effectively. For example, with prior training, being able to measure more confidence in
- Adequate large models (adequate training instruments)
- Adequate training data
If you measure the count without these two things, you won't see any improvement. However, if you measure all three, you get amazing results, like the LLMS front we see now, for example, with the release of Gemini 3.
Thus I argue that you should look to scale your LLM fees as much as possible. This, for example, can fire several agents to compile the code in parallel, or to start an in-depth gemini research on a topic of interest.
Of course, consumption must still be the right amount. There is no point in starting a coding agent on some abstract task that is unnecessary. Instead, you should start the code agent at:
- It's a straightforward argument that you never feel like you've had time to sit down and do for yourself
- A quick feature was requested on the last sales call
- Some UI improvements, you know, today's coding agents handle easily
In a world with an abundance of resources, we must look to increase our use of them
My main point here is that the threshold for doing jobs has dropped significantly since the LLMs were released. In the past, when you had a bug report, you had to sit down for 2 hours with deep concentration, thinking how to solve that bug.
However, today, there is no longer a case. Instead, you can log into the directory, file a bug report, and ask Claude Sonnet 4.5 to try to fix it. You can come back after 10 minutes, test if the problem is fixed, and create a pull request.
How many tokens can you use while doing something useful with the tokens
How to Apply for an LLM
I talked about why you should measure the use of LLM by using multiple coding agents, deep research agents, and any other Agents ai. However, it can be difficult to imagine what you should put out the fire. Therefore, in this section, I will discuss some agents that you can fire to measure your LLM usage.
Coding agents are similar
Parallel coding agents are one of the easiest ways to install LLM implementations for any program. Instead of working only on one problem at a time, you start two or more agents at the same time, or use Cursor Agents, Claude code, or any other aventic coding tool. This is usually made very easy to do by using the git staff.
For example, I usually do one big job or project that I'm working on, where I have a pointer and a program. However, sometimes I get a bug report when logging in, and I automatically submit it to Claude Code to make it search why the problem occurs and fix it if possible. Sometimes, this works in a box; Sometimes, I have to help it a little.
However, the cost of starting this bug is very low (I could just copy the linear magazine from the Cursor, which I learned the problem using the linear MCP). Similarly, I also have a script that automatically searches for relevant opportunities, which I have been working on in the background.
Advanced Search
Deep research is a functionality you can use in any of the Frontier model providers such as Google Gemini, Open Chatgpt, and Anthropic's Claude. I prefer Gemini 3 research for deep research, although there are many other deep research tools out there.
Whenever I'm interested in learning more about a topic, finding information, or anything similar, I fire up a deep research agent with Gemini.
For example, I was interested in finding out some of the prospects offered by a certain ICP. I then quickly set up the ICP information on Gemini, gave it some content information, and started researching, so it could start while I worked on my main programming project.
After 20 minutes, I had a short report from Gemini, which turned out to contain loads of useful information.
Creating a workflow with N8N
Another approach to LLM design is the use of LLM to create workflows with N8N or any similar workflow tool. With N8N, you can create a specific flow, for example, read the highlighted messages and perform a specific action based on those lazy messages.
To do, for example, have a workflow that reads the Bug Report Group group on slack and automatically starts the Claude code agent for the given bug report. Or you can create another workflow that combines information from many different sources and presents it to you in an easy-to-read format. There are endless opportunities with career building tools.
More
There are many other strategies you can use to measure the use of your LLM. I have only listed the first few things that came to mind when working with LLMS. I recommend that you keep in mind what you can change using AI, and how you can position it to be more effective. How you can apply for an LLM will vary greatly across different companies, job titles, and many other factors.
Lasting
In this article, I have discussed how to scale your LLM to become a more successful engineer. I argue that we have seen performance scaling very well in the past, and it is very likely that we are seeing more powerful results by scaling our use of LLMS. This can fire multiple code agents in parallel, running deep research agents while eating lunch. In general, I believe that by increasing our use of LLM, we can be more productive.
👉 Find me in the community:
📚 Get my FREE language ebook models
💻 My webinar on visual language models
📩 Subscribe to my newsletter
🧑💻 Get in touch
🔗 lickEdin
🐦 X / Twitter
✍️ Medium
