When I started working on MCP for X in April of this year, I started with a personal bias/conviction and tried to argue against myself.
My bias is that I want to preserve the primacy of human expertise and judgment. In other words, I recognize AI to be a tool like we have never seen before, but it is a tool nonetheless.
Humans have been making tools for hundreds of thousands of years. Never once has a tool become a complete replacement for the tool-maker. And I don't think that's going to change.
I have plenty of technical reasons to believe that we will never get to a point where out-of-box AI is exceeding human expertise & judgement, especially in niche domains. But an intellectually honest approach is to try and find the strongest opposing argument, a so-called steel-man.
In this case, I think the strongest argument to demote human intellect is a self-improving AI that is given access to data that isn't publicly available and reinforced with human experts. Ironically, this is exactly what is happening right now where companies are willingly lowering drawbridges over their last remaining moat, proprietary data and judgment, by having Claude or Codex etc. cycle through their workflows and having their expert talent provide all kinds of feedback back to the AI labs or their "forward deployed engineers".
Firms are thinking that their contracts will save them from the AI sucking the actual value out of their orgs and making it generally available in the next update.
So why do I think this won't actually happen? It has nothing to do with the AI companies acting ethically and morally and being a fiduciary of their customers' assets. If you believe that, I have an AI-picture of a bridge to sell you. It goes back to how the thing actually works. Next token prediction. You can see it in all the models. They have no conviction of their own, and the output is entirely dependent on how you prompt them. They will find the best possible sequence of words given your prompt, no matter how much you try to bury human directions to confuse yourself that some sort of sentience has emerged from the weights. It's just not there.
Anyway this is a long debate and I might write a longer post on it sometime. The point is that if I'm wrong, well - then there is hardly any point to engaging in any knowledge work. Why get a second rate opinion from a consultant charging $500/hour, when an AI can give you a far more competent deliverable for $5 and 10,000x faster.
If I'm right, then companies need to take urgent measures to 1) protect the most important asset they have, which is human expertise and judgement, and 2) leverage this as the only ultimate differentiator that it really is in the age of AI, and encode and use it with any model + harness. Treat the AI-layer as fungible across the many providers, and with zero data retention, zero dependence, and zero loyalty. Take control of the AI economics, your strategic roadmap, and recognize that even, or perhaps especially, in the age of AI - the expertise and judgment of your people is your moat.
And so was born MCP for X. A platform that lets experts and orgs encode their judgment and expertise, and own it completely. Where these can be used with any agent and harness. Where the run is programmatically gated (and not just via prompts), determinism dial is pushed higher, and each run is governed and traceable.
I am looking for beta-testers of this concept. I think this is a super-advanced and frontier application of how to have an AI do real, useful, complex work. If interested, write to team@mcpforx.com and I'll set you up.
I always come back to this. We shouldn't lose our agency as we mess around with agentic AI. We choose what we build. There was a great piece of dialogue in the movie Jurassic Park, "the scientists were so preoccupied with whether or not they could, they didn't stop and think if they should". If I'm wrong and supremacy of artificial intelligence over natural intelligence is achievable - it is up to the pursuers of that goal to convince us that investing trillions into that project is beneficial - to us! They need to make a convincing argument as to why we should be happy about a data center in our backyard while losing our jobs.