This is solved by the agent having its own identity and credentials. Why would you share your login and identity with your AI agent?

Access control and permissions should be handled on the backend by enforcing IAM on well-defined principals, not with MCP middleware. Claude can already bypass MCP and call APIs or use CLIs if it runs into blockers using MCP, so it’s not an effective point to implement the control.

IAM is generally binary (allowed yes/no) whereas MCP usually supports more nuance (always allow/ask if risky/always ask/no)

You can have agents that serve multiple different users with varying levels if permission

Anti-pattern imho. Agents should operate within granular identity and permission scopes, with audit and log trails for all data operations (read, write, etc).

Copilot: https://learn.microsoft.com/en-us/entra/agent-id/identity-pl... | https://learn.microsoft.com/en-us/purview/audit-copilot (for example)

TLDR Maintain an identity boundary whenever possible.

I’ve solved this by keeping my laptop keyboard uncustomized, while using Colemak-dh and home row mods on my desktop split keyboard.

I’m proficient enough that I can work on anyone else’s machine well, and I have the comfort and speed benefits on my main workstation.

If your scale is large enough and you don’t care about identifying individuals, synthetic data does this fairly effectively.

Vietnamese also has it.

The spec requires negotiation between the devices to deliver any nontrivial amount of power. A data connection is necessary to enable charging.

Author here. I agonized over the dates in the post when I was writing it. It’s correct to the best of my knowledge, but details fade with time.

I’m glad you had an easier time than we did!

Author here. At the time, Angular relied a lot on the AOT compiler and tree shaking to keep bundle sizes down. No idea if this is still the case.

If we built the app with the stable branch the bundle size was orders of magnitude smaller: less than 200kb. Still a bit of a chonker, but more reasonable than the ridiculousness the experimental SSR branch spat out.

Author here. That’s the takeaway I was going for.

It could have been any technology. The silver bullet is choosing the right tool for the job.

I don’t have an attachment to any particular tech. At the time React was what I knew, and I was coming off the back of building a server side rendered React site when I joined this company. I had a team of JavaScript-focused engineers to work with.

Author here. It’s actually my Dad!

Author here. It was a gamble, but I was fighting against a very strong sunk cost fallacy in leadership at the company at the time, and there was a general lack of trust in the entire technical team. I *would not* recommend this approach as a typical way of doing business, and maybe I didn’t do a good enough job at communicating that in the post. This is the only time in my career I’ve delivered this kind of ultimatum.

Everything about this particular situation was exceptional. I focused on the decision to do a rewrite in the post because I thought it was the more interesting part of the story. In hindsight I might have gotten that wrong.