Real-world
Regulated work
Concrete workflows, including the tradeoffs and limitations reported by their authors.
this collection
Local-only by requirement: tracing binaries for cybersecurity research
Compliance, not preference, rules out any cloud infrastructure for this cybersecurity researcher's day-to-day binary analysis.
Drafting medical reports from clinical notes, kept fully local for privacy
Turning clinical notes into medical reports is privacy-sensitive enough that this person built a dedicated local skill for it rather than risk a cloud tool.
A HIPAA-style internal LLM platform, built on a cluster of DGX Sparks
Beyond personal daily use, this person stood up a Spark cluster for a client so their employees get an internal, compliance-friendly LLM platform.
A cardiologist's referral system, fully air-gapped
Referral triage that extracts details, standardizes them, and suggests categories and tests — with the final call always left to a doctor, and nothing ever leaving the network.
A local parser and RAG for ITAR-constrained federal documents
Aerospace work with a legal requirement to keep data in-house led to a custom Qwen document parser and RAG, after a stock model struggled with an 785-page federal regulation.
Legal work: a cloud planner driving local anonymisation, OCR and RAG
A legal user has Claude build task plans while a local Qwen 27B executes them, with anonymisation, OCR for difficult documents and a RAG pipeline using nomic embeddings.
RAG and schema classification on data that must not leave the network
At work, local models handle bounded tasks on sensitive data: document RAG and classifying material into schemas, while frontier models only help build the systems.
A solo lawyer wants local on the old laptops practitioners actually own
A lawyer interested in local models for privacy and offline use points out that most solo practitioners run very old, low-spec laptops and asks how to make local work there.
Unit tests and reviews for NDA-bound client code on a single 3090
A contractor whose clients forbid commercial providers from reading their source code runs unit tests and code reviews locally on a 3090, with a long-context llama.cpp setup.
Pentest work and code review split across an M5 Pro laptop and a Strix Halo
A security professional runs uncensored Qwen models for client work on a 64GB MacBook and a 128GB Strix Halo, using a containerised OpenCode setup with ported code-review skills.
Uncensored Qwen 3.8 Flash Next tops a 19-task hacking benchmark at 80.6% on 2x RTX Pro 6000
On a 19-task cybersecurity benchmark, an uncensored Qwen 3.8 Flash Next solved 80.6% first try, but the run used someone else's 2x RTX Pro 6000 and only 2 runs per task.
A lawyer keeps client documents off any cloud API
Confidentiality obligations, not a specific incident, are enough reason for this lawyer to do all document review and drafting locally.
Three rigs for freelance work an NDA won't let leave the building
Client confidentiality rules out cloud APIs for this freelancer, who spreads different model sizes across three separate machines.