Not a hit piece on AC, rather about how non coordinated a thorough policy change can have adversarial effect.
Also, yes AC is essential, but it is on track to consuming so much electricity that is has to be pushed in the right direction to not become a new problem replacing the old.
In startups with less than 50 people (and I am being generous on the number), everyone talks to everyone all the time, so there is no need for these moments to extract key info to fix/improve situations, identify topics to push, ...
But once the company is just large enough, there is no way you're going to interact with everyone in a meaningful manner (n^2 relationships and all that), and the simplest solution is intermediaries and 1-1s.
To me this rule is not necessarily something each individual must do on their own, but rather an objective for systems and societies.
In this frame the outcome is more: Companies providing chatbots should not encourage anthropomorphism by giving cute names, making them witty, using human like pictures, ...
It's not about the cyclists wearing ANC headsets (which is already prohibited at least in Euro countries), but about pedestrians wearing them. Another problem altogether.
What these 'channels' do is essentially why I was running a nanoclaw at work: triggering a claude code based on events and getting feedback/review/analysis which nicely closes the loop with other agents.
Not sure why it has to be an mcp, but will be trying this out asap.
I am not an expert, but from the examples in the article I think the author is looking for a bit more than read-your-writes.
E.g. They mention reading a list of attachements and want to ensure they get all currently created attachements, which includes the ones created by other processes.
So they want to have "read-all-writes" or something like that.
Also, yes AC is essential, but it is on track to consuming so much electricity that is has to be pushed in the right direction to not become a new problem replacing the old.