Cosmo Santoni @cosmosantoni.bsky.social · 11/03/2026By removing host-device round-trips for the autoregressive loop, this runs unmodified across platforms while saturating up to 64% of memory bandwidth on a single batch on Cloud TPU v6e. #DeepLearning #OpenSource #Flax #Python 020
Cosmo Santoni @cosmosantoni.bsky.social · 11/03/2026#MachineLearning state-space models are incredible but custom CUDA kernels lock its performance to NVIDIA hardware. My latest #JAX port maps the SSD algorithm to XLA passes, achieving true O(1) on-device caching across CPU, GPU & #TPU. Pre-print 👉 huggingface.co/papers/2603.... #AI #SSM 220
Cosmo Santoni @cosmosantoni.bsky.social · 20/02/2026🧵Pt3. Demo of what it looks like in practice you can see context usage of a snapshot with and without trimming below. 50% reduction on this particular task. 000
Cosmo Santoni @cosmosantoni.bsky.social · 20/02/2026🧵Pt2. A typical 150k token session is 60-70% tool result dumps and base64 signatures Claude already processed. Trim strips that, keeps every message. Observed 50% reduction in real sessions. Unlike /compact, nothing gets summarised away. Analysis breakdown here github.com/CosmoNaught/...github.com 110
Cosmo Santoni @cosmosantoni.bsky.social · 20/02/2026🧵 Built git-style versioning for Claude Code sessions. Snapshot context, branch for different tasks, trim the bloat. 40 mins of codebase analysis reused across 5 tasks instead of re-explaining from scratch. github.com/CosmoNaught/... #ClaudeCode #DevTools #Anthropic #Claude #LLM #AIAgents 140
Reposted by Cosmo SantoniNilani Chandradeva @nilanicdeva.bsky.social · 06/12/2024Took a much-needed screen break this week to answer the age-old tale: how many epidemiology PhD students does it take to build a gingerbread house 🤔 @nderqui.bsky.social @olliesimmons.bsky.social @cosmosantoni.bsky.social Sam Hemmings @mrc-outbreak.bsky.social 073
