/blog
Case studies, product news, and research from the Subterra team building production AI systems.
ResearchA practical benchmark comparing a local NVIDIA GB10 AI agent running Qwen3.6 35B through Hermes against GPT-5.5 on real office workflows, including code repair, document synthesis, web edits, travel research, and concurrency limits.
ResearchA practical RTX 4090 vs NVIDIA GB10 local AI benchmark comparing Qwen3.6 35B Q4, Q5, MXFP4, long-context workloads, and dense 70B inference. The results show where the RTX 4090 dominates, where GB10’s memory headroom matters, and why the VRAM cliff is one of the most important limits in local AI hardware.
Case StudiesSubterra uses custom-forked AI agent systems built around Hermes Agent and OpenClaw to support blog writing, CRM workflows, documentation, software development, proposals, research, and internal operations. These systems create real leverage, but improper deployment can be catastrophic without strong security, permissions, and human review.
ResearchA detailed benchmark of Qwen3.6-35B-A3B on NVIDIA GB10 using 243 llama.cpp runs, comparing BF16, Q8, Q5, Q4, and MXFP4 GGUF variants for local AI inference, RAG, agents, and long-context workloads.
ResearchA hands-on benchmark of Gemma 4 on NVIDIA GB10 comparing 31B dense and 26B-A4B MoE variants across speed, memory, thinking mode, and practical deployment tradeoffs.
Case StudiesA proof-led look at how Microsites turned static document delivery into interactive, trackable web experiences at scale.
Case StudiesHow Subterra turned fragmented geotechnical inputs into a clearer, more searchable intelligence workflow through GeoIQ.
What Subterra learned while building an observability platform around anomaly detection, signal interpretation, and faster response.
A proof-led breakdown of how Subterra approaches private AI architecture, deployment boundaries, and governance for sensitive workflows.
Subterra AI Ask anything about our services, tech, or how we can help.