Unpopular opinion Qwen 3.8 is hard to understand

I find both Qwen 3.8 27b and Qwen 3.8 Flash Next difficult to read. Here's some examples of what I mean: Model-visible tool set per turn (assembled by the host at provider-request time): persona tool allowlist ∩ session tool surface ∩ tools not deny-classified under the active permission profile. In the above, Qwen uses the set intersection symbol as opposed to a human readable explanation. Maybe this is because it's been trained so hard on math, science, reasoning, so it's a little understandable but unnecessarily dense in my opinion. My thoughts are that this is the consequence of minimizing "tokens per intelligence" -- that it makes reading it as a human incredibly dense. It also uses the word "persona" which is an odd word to use when something like "mode" or "agent" would have made more sense. And another example: Consent is negotiable; enforcement is gravity. WTF does "gravity" mean in this sentence, just say "Consent is negotiable; enforcement is not." I know a lot of claude users have been talking about how claude 5 was a step backwards in terms of human-readability. I'm somewhat afraid Qwen is taking the same road. Qwen 3.6 was the last easy to read Qwen model IMO. What are your thoughts?