The problem nobody talks about

I work across four worlds simultaneously. This is not an accident - it is a deliberate choice rooted in the Quadruple Helix model of innovation, where industry, government, academia, and civil society reinforce each other to create lasting impact.

Kamer Vishi - Quadruple Helix Industry Financial sector Government Public institutions Academia Research - teaching Civil society Knowledge - forensics Kamer Vishi PhD - Associate Prof. Co-founder, CDFC Quadruple Helix - four-sector collaboration

In practice that means: working in the financial sector at the intersection of government and industry, lecturing and developing curriculum in academia, co-founding a centre dedicated to cybercrime and digital forensics, and engaging with the broader research and knowledge community.

Each world has its own rhythm, its own language, its own demands.

The bottleneck is never knowledge. It is time and cognitive load. Drafting the same type of email for the fifth time this month. Restructuring a lecture outline from scratch when the shape is always similar. Translating a complex technical finding into plain language at 11pm before a meeting.

I kept thinking: I already know how to do all of this. I just need to stop doing it manually every single time.

What I actually built

I set up a personal AI workspace using self-hosted LLM, structured around three dedicated projects:

  • ONH / CDFC - lectures, research, curriculum, and everything related to the Cybercrime and Digital Forensics Centre I am co-founding with my colleague Fredrik Harboe.
  • Blog and thought leadership - drafting posts like this one, LinkedIn content, and research writing in both English and Norwegian.
  • Kamer Core - personal decisions, long-term thinking, and anything that does not fit the other two.

Each project has a detailed system prompt - a master profile that tells the AI who I am, how I communicate, what I care about, and what I am currently working on.

It does not replace my thinking. It handles the scaffolding so I can focus on the parts that actually require me.

The setup took one afternoon

Here is what I did, in order:

  1. Created a subdomain at ai.vishi.no as a personal portal
  2. Set up three self-hosted LLM Projects with full professional context loaded into each one
  3. Built a prompt library for recurring tasks - one click to copy, paste into self-hosted LLM, done
  4. Set up this blog using Hugo - a static site generator with zero attack surface, no database, no plugins to patch

The blog setup deserves a separate mention. As someone who works in security, I am not going to run a database-backed CMS and hope for the best. Hugo generates pure static HTML. Nothing runs on the server. Nothing to exploit. Loads in milliseconds.

What changed immediately

The first thing I noticed was the reduction in startup cost per task. Before, starting a new piece of work meant mentally loading all the context - the framework, the tone, the audience, the history. Now I open the right project and the context is already there.

The second thing was the quality of thinking. Having a well-prompted AI as a thinking partner - one that knows your background, pushes back when something is weak, and does not just agree with everything - is genuinely different from using a generic chatbot.

The third thing was time with my family. That one is simple. Less repetitive work in the evening means more presence at home. That was the original goal.

What this is not

It is not autopilot. Every output gets reviewed, edited, and often substantially rewritten. The AI handles the first draft and the structure. The judgment, the tone calibration, the decision about what to actually say - that stays with me.

It is also not a replacement for expertise. If anything, it amplifies it. The more context and domain knowledge you bring, the better the output. A generic user gets generic results. A specialist who loads their full professional context gets something much closer to a genuine thinking partner.

The bigger point

We are at an early moment where the people who invest in setting this up properly will have a meaningful advantage over those who use AI casually or not at all.

Not because the AI is magic. Because systematic, context-rich use of good tools compounds over time.

I spent one afternoon on this. I expect to get that time back within the first week.

If you are a practitioner juggling multiple roles and contexts - research, teaching, operations, writing - this is worth doing. The setup is not complicated. The return is immediate.

I write about cybersecurity, AI, and research - in English and Norwegian. Follow along at vishi.no/blog or connect with me on LinkedIn.