fix(sae): correct pre-trained SAE hook ID to blocks.6.hook_resid_pre in gpt2-small-res-jb to enable real active inference

feat: complete CircuitScope v3 upgrades, Pearson validation checks, and elite README

feat: implement Upgrade A - dataset-wide mean activation caching hooked on blocks.{layer}.attn.hook_z

feat: implement A+ automatic semantic prompt composition filtering, dynamic active feature legend, sequence-preserving mean ablation hooks, and elite alignment science hiring dashboard

feat: implement automatic semantic feature clustering composition in SAE lab

feat: upgrade mechanistic interpretability workstation to elite 10/10 research-grade

chore: add root GET endpoint returning online status to resolve HF Space 404 logs

feat: add dynamic sparsity threshold control slider and raw telemetry export tool

feat: implement Dynamic Sparse Autoencoder (SAE) Latent Extractor and Live Activation Lab

feat: complete live cpu inference, research sweep, blog & production build