2025 is the year of AI. Whether you are a developer or a creator, integrating AI into your daily workflow is no longer optional. It is mandatory. If you started applying AI properly from the beginning of 2025, chances are you have already seen significant outperformance compared to traditional workflows.

I use CursorAI as my primary coding IDE. I occasionally experimented with Codex and Claude Code, but only briefly. Cursor has remained my main tool throughout 2025. Below are my personal usage statistics from this year.

I used 2,624,290,510 tokens throughout the year. I am not sure whether this number is objectively high or low, but what I am most proud of is consistency.

My token usage is evenly distributed across weeks, which shows that I built continuously throughout the year.

In 2025, I sent 8,412 chat messages across 1,535 chat sessions. Vibe coding became my primary way of building.

My workflow is simple:

• Design a clear plan

• Let the agent implement it

• Verify results with unit tests and end-to-end tests

Vibe coding allows me to control requirements, execute ideas quickly, and still maintain code quality and maintainability.

I will continue following this approach until a more effective paradigm appears.

To be honest, I am a big fan of Claude models, from Sonnet 3.5 to 4.5. Most of my usage this year was on Claude. Sonnet consistently outperformed other models in the same category.

When OpenAI GPT-5.1 and GPT-5.2 were released, I moved some workloads to them. They are not strictly better than Sonnet, but when considering both cost and output quality, they can be the better choice.

For simple tasks, I occasionally rely on automatic model selection.

In my opinion, we will eventually reach a point where most coding models are smart enough to handle both basic and advanced requirements. At that stage, competition will focus mainly on price and speed rather than intelligence.

I coded 334 out of 365 days, reaching a 91.5% consistency rate. For comparison, a disciplined employee working 20 days per month reaches roughly 65%.

If you are building a startup, working hard is not enough. You need to treat coding as part of your life.

To summarize, I am currently in the top 17% of Cursor users. AI has fundamentally changed how we code and how we work. If you have not started using AI seriously yet, you are already behind.

The next step is simple: open your laptop and start learning today.

#AI #AICoding #VibeCoding #CursorAI #ClaudeAI #GPT

Share your AI journey in 2025 in the comments below. I will read and respond to every single one 👇

2025 is the year of AI. Whether you are a developer or a creator, integrating AI into your daily workflow is no longer optional. It is mandatory. If you started applying AI properly from the beginning of 2025, chances are you have already seen significant outperformance compared to traditional workflows.

I use CursorAI as my primary coding IDE. I occasionally experimented with Codex and Claude Code, but only briefly. Cursor has remained my main tool throughout 2025. Below are my personal usage statistics from this year.

I used 2,624,290,510 tokens throughout the year. I am not sure whether this number is objectively high or low, but what I am most proud of is consistency.

My token usage is evenly distributed across weeks, which shows that I built continuously throughout the year.

In 2025, I sent 8,412 chat messages across 1,535 chat sessions. Vibe coding became my primary way of building.

My workflow is simple:

• Design a clear plan

• Let the agent implement it

• Verify results with unit tests and end-to-end tests

Vibe coding allows me to control requirements, execute ideas quickly, and still maintain code quality and maintainability.

I will continue following this approach until a more effective paradigm appears.

To be honest, I am a big fan of Claude models, from Sonnet 3.5 to 4.5. Most of my usage this year was on Claude. Sonnet consistently outperformed other models in the same category.

When OpenAI GPT-5.1 and GPT-5.2 were released, I moved some workloads to them. They are not strictly better than Sonnet, but when considering both cost and output quality, they can be the better choice.

For simple tasks, I occasionally rely on automatic model selection.

In my opinion, we will eventually reach a point where most coding models are smart enough to handle both basic and advanced requirements. At that stage, competition will focus mainly on price and speed rather than intelligence.

I coded 334 out of 365 days, reaching a 91.5% consistency rate. For comparison, a disciplined employee working 20 days per month reaches roughly 65%.

If you are building a startup, working hard is not enough. You need to treat coding as part of your life.

To summarize, I am currently in the top 17% of Cursor users. AI has fundamentally changed how we code and how we work. If you have not started using AI seriously yet, you are already behind.

The next step is simple: open your laptop and start learning today.

#AI #AICoding #VibeCoding #CursorAI #ClaudeAI #GPT

Share your AI journey in 2025 in the comments below. I will read and respond to every single one 👇