So far this series has covered the data going into a model and the technical choices made while training it. Now: what happens when your organization doesn’t build a model at all, but fine-tunes someone else’s. ...
The first four parts of this series covered where AI training data comes from, what gets permanently baked into a model, how training and optimization choices introduce risk, and how fine-tuning inherits all of it. This part covers a harder problem: you often can’t check any of it yourself. ...
This series started with a simple idea: every AI model is a direct product of its training data, and most organizations deploying AI have no real visibility into what that data actually was. Over five parts, we traced that problem from raw data collection all the way through to the documentation that’s supposed to make it transparent. ...
Paperclip.ai is a self-hosted AI project management platform. It gives you a Linear-style issue tracker where AI agents can actually pick up tasks, reason through them, and take action — all running on your own infrastructure. No cloud dependency, no data leaving your environment. For a homelab running a SOC stack, threat intelligence pipelines, and a growing collection of automation projects, having an AI agent that can work through a backlog is genuinely useful. ...
The multi-agent security pipeline we built earlier produces useful reports — structured risk analysis, mitigations, recommendations. But it has a ceiling: every agent reasons from what the model learned during training. It doesn’t know about the domain that started hosting malware last week, the C2 infrastructure tied to a campaign your MISP instance just ingested, or the specific indicators your feeds have flagged today. ...
Most AI project tutorials end the same way: deploy to AWS, add an OpenAI API key, pay monthly. That works, but it sidesteps the more interesting question — what does it look like to build something real without handing control to a cloud provider? ...
If you’ve followed the Self-Hosted AI Stack walkthrough, you’ve got Ollama running locally, OpenClaw as your agent UI, and the whole thing locked down behind Tailscale. That’s a solid foundation. But a chat interface, however useful, isn’t the ceiling of what this stack can do. ...
There’s a certain appeal to running your own language model. No API costs, no data leaving your network, no rate limits, no terms of service to worry about when you feed it sensitive context. For anyone who works in security — or just values privacy — the idea of keeping inference local is worth the setup cost. ...
AI tools are showing up in analyst workflows whether your organization plans for it or not. A Tier 1 analyst dealing with a hundred alerts a day will find ways to work faster — and if the org hasn’t provided sanctioned tools, they’ll use unsanctioned ones. That’s not a criticism, it’s human nature under pressure. ...
Tier 1 analysts using AI for alert triage is one problem. Security engineers integrating AI into automated pipelines is a different one — and in some ways a harder one. When AI is in the pipeline, the decisions it influences happen at scale, without a human in the loop on every call, and the code you write today becomes the attack surface your team defends tomorrow. ...