Viewing profile — pqtr2

Yeah but what are they for?

Couldn't agree more. Coding benchmarks are just a score. Benchmark the harness.

Agree on the general principle but you take a long time to get to the point and the reliance on MCP just shifts the problem somewhere else.

[dead]

Did you see token-warden, could be a good match. Be interesting to see how the models compare on the same task especially with different thinking levels...

Interesting, will take a look. Any published results on actualized savings over time? Would be very useful to know.

Useful but I'm Linux too :)

Exactly right, code review time has gone up 500%, code writing time has decreased to almost zero

This is great and definitely under-served right now, will take a look. I have done a lot of primary research on MCP observability so know what I am looking for :)

[dead]

Good explanation here https://alex000kim.com/posts/2026-03-31-claude-code-source-l... its a server side flag set by the distillation detection, shouldn't affect most user's session…