I think Bun helps with the memory pressure, granted this is relative to V8. I'd pushback on the certainty with the reality that TS provides a significant drop in entropy while benefiting from what is a sweet spot between massive corpus size and low barrier for typical problem/use-case complexity. You'll never have the fastest product with JS, but you will always have good speed to market and be able to move quickly.

Bun uses JSC (JavaScriptCore) instead of V8. From what I understand, whereas Node/V8 has a higher tier 4 "top speed", JSC is more optimized for memory and is faster to tier up early/less overhead. Good for serverless. Great for agents -> Anthropic purchase.

I wonder how much of a lost in the middle effect there is and if there could be or are tools that specifically differentiate optimizing post compaction "seeding". One problem I've run into with open spec is after a compaction, or kicking off a new session, it is easy to start out already ~50k tokens in and I assume somewhat more vulnerable to lost in the middle type effects before any actual coding may have taken place.

The tech/demand for the glasses didn't break through some threshold it hadn't reached before, all of the sudden. They became viable as a product again because real training data is more valuable now than ever.

If the answer is, "of course not". Pull that thread. Honestly, so much "therapy" for some of us boils down to confronting/examining that disconnect and exploring why it exists/how it came to be.