Embracing Local Solutions: A Game Changer in AI and Productivity
In recent years, the tech community has increasingly recognized the benefits of adopting local or self-hosted tools, and my experiences exemplify just how transformative this shift can be. As someone who has been experimenting with various AI models and development tools, I’ve discovered that going local not only enhances efficiency but also fosters a more sustainable and user-centric approach to technology.
A Personal Journey with Cursor and Open-Source AI
My journey began with Cursor, a versatile platform I’ve been using intermittently, predominantly subscribing to the $20 tier, occasionally upgrading to the $60 plan, and exploring the latest frontier models each month. My use isn’t rooted in high-stakes coding or dreaming of building the next massive SaaS; rather, I leverage these tools for practical purposes—preparing code examples for talks, assembling demo projects, and contributing to open source initiatives.
Recently, a fortunate opportunity arose when a friend sold their 48GB MacBook Pro as part of an upgrade. Seizing the chance, I acquired it to explore what all the fuss was about with the AI model Qwen-3.6-35B. My goal was straightforward: set up benchmarking tasks and establish a local environment to run these models effectively.
Harnessing the Power of Local Models
Using Cursor, I instructed Composer 2.5 to run benchmarks and configure Qwen-3.6-35B on my newly acquired MacBook Pro. Remarkably, the system responded with a detailed MLX profile and a ready-to-use code configuration optimized for my MacBook Air to serve as a baseline. This seamless interaction demonstrated that, while the model isn’t GPT-5.5 or even as advanced as 27B, the local deployment still holds significant value.
Since integrating the MCP servers for documentation and routine skill/rules management, my workflow has been remarkably smooth. I no longer worry about hitting usage limits—tokens simply process quickly, and if something doesn’t work as expected, I rerun the prompt without concern for additional costs. This flexibility encourages experimentation and documentation, making my repositories more organized and well-documented than ever before.
The Benefits of Local Environments
Opting for local setups brings several advantages:
- Cost Control: No surprise charges—just re-run your prompts as needed.
- Data Privacy: Keep sensitive code and data within your own environment.
- Increased Reliability: Less dependency on external servers or endpoints.
- Enhanced Productivity: Quickly iterate and test without waiting for cloud responses.
This approach might not be groundbreaking for many in the tech community, but for me, it’s a delightful way to interact with AI and coding. It shifts the experience from a transactional, money-driven interaction to a more playful and productive engagement with technology.
My Configuration Snapshot
For those interested, here’s a glimpse into my current AI model configuration—admittedly simple, but effective:
json
{
"models": {
"qwen3.6-35b-4bit": {
"name": "Qwen3.6 35B",
"reasoning": true,
"limit": {
"context": 65536,
"output": 8192
},
"options": {
"temperature": 0.6,
"top_p": 0.95,
"extraBody": {
"max_tokens": 8192,
"enable_thinking": true,
"chat_template_kwargs": {
"enable_thinking": true
}
}
}
}
}
}
While I’m not certain if this setup is perfect, it works for my needs, and that’s what matters most.
Final Thoughts
Adopting local AI tools represents more than just a technical decision; it’s a philosophical shift towards greater ownership, control, and sustainability in our tech workflows. It makes the process less about transactional money exchanges and more about empowering ourselves to innovate on our terms.
If you’re contemplating experimenting with local AI, I highly recommend it. The experience is invigorating, and the freedom it offers is genuinely life-changing.
