JOURNAL
Why I only work on Claude.
Specialization as a moat, and as a constraint.
Most AI consultancies hedge. They claim model agnosticism, list a dozen logos on the website, and quietly use whichever tool the salesperson is most comfortable with on the day of the call.
Brinvik does not hedge. Brinvik works on Claude only. Here is why.
The argument for focus
Focus produces depth. Depth produces results. Results produce trust. Trust closes deals.
A specialist who has shipped twenty Claude deployments has seen the failure modes. They know the cost-of-context tradeoffs at the third decimal. They have an opinion on prompt caching that is not theoretical. They can tell you which Anthropic features are fake-stable and which are real-stable.
A generalist who has shipped two Claude deployments and three GPT deployments and one Gemini deployment knows none of this. They cannot. The hours add up the same way for everyone.
Multi-model is the consultant's hedge against having a thesis. It looks like neutrality. It is actually indecision dressed up as range.
Why Claude specifically
A few reasons that are not technical, because the technical comparison is not the part that matters most.
Long context. A real working session for a B2B knowledge worker might be a 60-page contract, a 30-page playbook, last quarter's call notes, and the prospect's last six emails. Claude reads the whole thing in one go. The decision quality reflects that.
Reasoning quality on serious work. Not toy problems. Not the test that gets photographed for a benchmark. The kind of work where the wrong answer has consequences and the right answer requires holding three constraints at once.
Behavior alignment. Claude refuses things in ways that match how serious B2B work is supposed to be done. It says "I don't know" when it does not know. It cites its sources when it can. It pushes back when the question is malformed. These are not features. They are temperament.
EU residency. Available. Real. Working. Brinvik uses it.
The Anthropic posture. A constitution-driven company that publishes its safety thinking. That posture matters less in some industries and more in others. In legal, financial, and B2B-trust-led work, it matters a lot.
These are not the only reasons. They are the ones that come up in every intake call.
The honest disclaimer
Claude is not always the answer.
If you need a consumer-facing chat experience tuned to a teenage audience, ChatGPT is probably the better starting point. If you need deep Google Workspace integration that a third-party API cannot reach, Gemini is the better starting point. If you need open-source self-hosted models for compliance or budget, you need Llama or Mistral, not Brinvik.
Brinvik is upfront about this on the site. I will be upfront about it on the call. If your problem is not a Claude problem, I will say so and point you somewhere useful. The audit fee is the only thing I charge before that conversation, and the audit fee is credited against any engagement that follows.
The competitive argument
If you are a Claude-only specialist competing against a generalist consultancy, you will out-deliver them on every Claude project. They will out-deliver you on the projects that are not Claude. The trade is clean.
The market is more than big enough for both shapes of consultancy. What is not big enough is the market for generalists who claim to be specialists. Buyers are getting better at the difference. They will keep getting better.
Closing
Specialization is a moat. It is also a constraint. Both are true.
Brinvik chose the constraint. The moat is the result.
If you need Claude, Brinvik is the shop. If you need something else, I will tell you. That is the entire pitch. It is short on purpose.
Written by Kim Olsen, principal of Brinvik. Get new essays by email.