Update README.md

Leader failover, 5-retry reconnect then drop, dataset picker, pretraining framing

Mid-run joins (resume bundle: weights+Adam state), auto-reconnect with backoff, per-message fragment reassembly

Fix 3+ device weight fork (canonical roster-order averaging), ICE candidate race, add TURN relays, send backpressure

Web trainer upgrade: sync guard (leader roster + weight-hash divergence check), 3xINT8 fast-accurate GEMM, DP4A hardware path (verified vs units), Spikewhale tokenizer (16.5k vocab), FineWeb-Edu streaming, gradient/checkpoint fragmentation, inference kit, generation tester, contribution logs

Mini transformer LM through the verified units; width/seq-len/batch/steps sliders (scale with devices)

Room-first lobby (Create/Join, host approval), group training settings sliders

DaisyChain-Web demo: P2P browser training through verified INT8 units

initial commit