
Conviction1000%
2.5K posts

Conviction1000%
@ConvictionFAQ
Community Lead @MindoAI / Hermesmaxxer



Today we release Token Superposition Training (TST), a modification to the standard LLM pretraining loop that produces a 2-3× wall-clock speedup at matched FLOPs without changing the model architecture, optimizer, tokenizer, or training data. During the first third of training, the model reads and predicts contiguous bags of tokens, averaging their embeddings on the input side and predicting the next bag with a modified cross-entropy on the output side. For the remainder of the run, it trains normally on next-token prediction. The inference-time model is identical to one produced by conventional pretraining. Validated at 270M, 600M, and 3B dense scales, and at 10B-A1B MoE. The work on TST was led by @bloc97_, @gigant_theo, and @theemozilla.













not enough people know about /side and it's underrated. Think claude's /btw but add: - multiple message followups allowed - you can spin up > 1 /side chats ask questions while the agent is working on your main session, or ask questions to explore other directions, and it won't affect your session's context!








ALIVE v3.2.0 🐿️ 🏗️ /alive:demo - scaffold a full world from one paragraph 🔧 /alive:system-upgrade rewritten - 13 phases, --dry-run, --resume, --rollback 🔍 version detection via content fingerprinting ⚡ alive CLI for log writes, doctor checks, task promotion 🛟 failure recovery - partial runs resume from last good stage 66 commits. 79 files. Plain files, still yours. github.com/alivecontext/a…











