Splash 1.3.0 cuts local agent first-token time from 19s to 1s via SSD offloading
songhan_mit · x · 2026-10-08
Splash 1.3.0 targets the wait before a local agent starts answering: SSD offloading drops first-token time from 19s to 1s when switching back to a project, and new tasks start 1.5× faster thanks to the Neural Engine — all on a 24 GB M6 Mac.
More from coding & agent
- Recursive Game Creator: an experience-oriented agentic harness hits SOTA 77.89 on GameCraft-Bench — hkuhk · 2026-10-08
- Polyphonic companion agent capability list: screen tracking, long-term memory, autonomous pursuits — RileyRalmuto · 2026-10-08
- 469-question eval of Qwen3.8-27B fine-tunes: best local model 28x slower than Opus 5.5 at 99.6% — norenEnmotalen · 2026-10-08
- Farseek: an agentic-built 3D puzzle game across four worlds made in under two days — const_reborn · 2026-10-08
- Claude builds an Ableton plugin and tests its own audio output in a loop — ctjlewis · 2026-10-08
- Free Inspector Tool Generates IGA-Style Review Packets for MCP/A2A Agent Protocols — ContextIQ · 2026-10-08