In October 2026 I posted a third X thread that covered a set of emerging AI‑hardware, robotics, model‑optimisation and security developments. I highlighted the memory‑speed‑first Positron inference chip, the one‑year Voxel home‑robot prototype, a RL‑post‑trained Qwen3.8‑27B model that halved hallucinations, the open‑source DeepInfer speed‑layer engine, the Open‑Science research‑workflow workstation, and a US sanction episode affecting five Chinese entities linked to a former Google engineer.
The Positron chip family is built to write AI answers faster and cheaper by focusing on memory speed (the real bottleneck when AI generates text) instead of raw processing power.
Voxel (Tuesday Labs) reached prototype status within one year, demonstrating a home‑robot that can tidy rooms, fetch items, locate objects, and tend plants.
A new RL post‑train Qwen3.8‑27B model cuts the hallucination rate on AA‑Omniscience from 30.4 % to 15.1 %.
DeepSeek just open‑sourced DeepInfer, the inference engine behind its production services, under the Apache 2.0 license.
AIPOCH reports 79.05 with gpt‑5.6‑sol on BiomniBench‑DA’s public tasks, ranking first in its comparison on a benchmark that grades agents on full biomedical analyses taken from published papers.
The US sanctioned five Chinese entities linked to ex‑Google engineer Linwei Ding and his startup Shanghai Zhisuan Technology; Ding received a 12 months – 1 day sentence.