After 8 months of running everything local, ive accepted the productivity tools also have to be local

Disclosure: Some links in this article are affiliate links. AI Maestro may earn a commission if you make a purchase, at no…

By AI Maestro May 12, 2026 2 min read
After 8 months of running everything local, ive accepted the productivity tools also have to be local

“`html




After 8 Months of Running Everything Local

After 8 Months of Running Everything Local

I’ve been running everything locally for the past eight months, and I’ve come to a few realizations. The key one is that moving my LLM (large language model) to a local environment has inadvertently made other tools feel less necessary.

Context:

  • M3 Max 64GB
  • Relying on ollama for inference, llama-3.3-70b-q4 for general tasks, and qwen3-coder-30b for code-related work.
  • nomic-embed-text-v2 used for vectors.
  • local-whisper.cpp for transcription, which runs on CPU but requires a GPU for better performance.
  • Obsidian, plain markdown, encrypted volume on disk for notes and memory management.
  • AirJelly for screen and cross-app memory capture. While not purely local, it stores the captured data locally with an inference round-trip to their backend.

I was most skeptical about AirJelly. It captures screen history but sends a cropped frame to their backend for analysis. The raw screen history never leaves my machine once I set this as my threshold. This aligns more with the goal of no packets leaving the system.

What I Use AirJelly For:

  • Cross-repo comparisons: examining how different inference server implementations handle continuous batching and speculative decode.
  • Maintaining a reference for quick lookups, such as which Slack thread corresponds to a specific commit during code review or debugging.

While the local setup isn’t as smart as some of the external APIs like Claude Opus 4.7, it works well enough for most tasks. For sensitive information, I still send those requests through an isolated environment to avoid any potential leaks.

Key Takeaways:

  • Moving your LLM to a local environment is just the beginning; other tools need to be addressed one by one.
  • The transition from relying on external services like rewind.ai or meta-owned apps to running everything locally can reveal how much data and context are being shared with these platforms.
  • No single tool is perfect, but the right mix of local tools can significantly improve productivity for most tasks.

“`

Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.

Name
Scroll to Top