Maybe I just lack the experience but I don't understand what situation this helps with. Is choosing access roles really such a common process that you need to automate it? I'm not seeing how this helps security at all if the user can just tell it what you need to access. Who are the users? Agents, employees, or customers?
I agree completely. I struggled throughout the post on how this is useful. If the role catalog is so vast that natural language processing is needed to filter them, I would consider addressing this from a different angle. Maybe RBAC is the wrong approach and other methods would lend itself better to the usecase? Hard to say.
On a larger scale I suspect we’re heading in a dangerous territory if AI for IAM is overshadowing fundamentals. I may be judging the post harshly, but more info on what is being solved here would be nice at least.
Presumably the only customers buying Pis right now are companies who use them in their products. There are much better options on the used market for people needing small computers/servers.
Legally you're right contract law does not assume that. But realistically your average Joe does not have the resources or time and is not willing to take the risk to sue a megacorp. Contracts only work when lawsuits can happen.
300 years ago average Joes were routinely signing tenant contracts for renting a plot of land from the local aristocrat, the imbalance of power was similar.
I agree. MCPs have the massive problem of not being composable, which is a big strength of shells/CLIs allow.
People make the argument that it doesn't matter for non-programmers. I use Bluebeam (a pdf editor which is standard in the construction industry) daily, and it has an MCP, but a big problem is that if the agent wants to import or export a large amount of data from a document, then it has to regurgitate a bunch of tokens into or out of the MCP. If it were a CLI, then it would just pipe things in and out as it needs.
I ended up just having the agent make a script that wraps the MCP so it can use it as a CLI instead.
As a trivial counterpoint, streaming outputs and proper signaling support is required for being able to compose efficiently under any setup/harness. The more layers of abstractions you add, the more likely it is your MCP server didn't support the usecase that some later agent came up with and so you can't stop your stream when you find your top 3 hits. If you don't even own the server, you can't even modify the source and restart it to fix it.
There are usecases of MCPs (e.g. serving outside customers via proprietary tools), but composability has not been in the minds of the early users of MCP and it is one of the hundreds of little practical drawbacks they have when used in house.
Information is practically free now. And many people that are fresh out of college are still not ready for the careers their degrees supposedly prepared them for. I expect either for colleges to adapt to provide more real-world knowledge or for companies to care less about whether someone has a degree and more about their work history and personal projects. For me, 4 years of work in the field (commercial construction and service calls) was a much better decision than 4 years of college would have been. My current company hired me for my experience in the field.
I strongly agree. Not only is information free, but application of that knowledge is also either free or incredibly cheap. The case for university is very poor now, and looking ahead even two years, I question its existence. Even before AI, it had basically become a signalling mechanism: "I have the right contacts, socio-economic status, grades, and foresight to get into this specific university and study this specific course. I have no idea how to apply it in the real world." There is some value in that signalling mechanism, but it appears most of the value derived was not even the degree itself. Let's cut out the middleman and on our CVs just put our parents' income, our zip code, our IQ, grades, and connections. I know that's crass and for the record I'm not advocating for this. I'm just deconstructing a sick industry which desperately needs upending.
I agree. Someone who is willing to learn will use the information they have available to do that. If someone has access to the world's information and is unwilling to learn, that is their problem, and college probably won't help.
This entire page (except for the footer) is made with images. Is there any reason they would do this? Are they making these images by hand or are they generated somehow? Just seems like an unusual choice.
I work for a construction design company, and marking up on PDFs is a very common task, so we use Bluebeam. I've found it's better just to have the agent call the MCP with python or edit the pdf directly, since (at least with Claude) there's no way to pipe data into the MCP directly; the agent has to output everything itself. For this reason alone I feel like command line tools are just the better way forward for most things.
reply