STS Collective — cybersecurity apparel and gear

Shop Now — Save up to 20% Today Code: SIMEONONSECURITY

Eye Spy — passive surveillance detector for Bluetooth, WiFi, ALPR, and AirTag awareness from STS Collective

Shop Now — Save up to 20% Today Code: SIMEONONSECURITY

Other Posts Tagged "Qwen3.5"

Self-Hosted Hindsight: Local Agent Memory and Private Model Routing

2026-11-08 (Modified: 2026-11-08) — Written by SimeonOnSecurity — 7 min read

A case study of self-hosted Hindsight, observed local Qwen routes, and request-scoped GPU work without a hosted model subscription on the inspected paths.

An abstract private memory service routes work to local CPU and GPU model servers — Memory storage, retrieval, and the inspected model routes run on owned compute

#local AI #local llm #Qwen3.5 #embeddings #Docker

Hermes Local Model Benchmarks: The GPU and Model Combinations I Would Use

2026-10-24 (Modified: 2026-10-26) — Written by SimeonOnSecurity — 26 min read

My Hermes model recommendations from long-context, tool, streaming, and vision tests on RTX 5060 Ti, RTX 2000E Ada, RTX 3060 Ti, and RTX 3070 systems.

#Hermes Agent #RTX 5060 Ti #long context #llama.cpp #Ollama

Inside Our Hybrid Agentic AI Fleet: Hermes, LiteLLM, and Tailscale

2026-10-21 (Modified: 2026-10-22) — Written by SimeonOnSecurity — 15 min read

See a verified lazy local AI fleet and its staged T0 to T4 control plane with Hermes, LiteLLM, Tailscale Aperture, and Pearl mining.

An anonymized private AI network linking cloud models, a secure gateway, and local GPU servers — A tiered agent fleet sends each task to an appropriate model while idle GPUs return to mining

#Hermes Agent #tailscale aperture #Tailscale #llama.cpp #Ollama

2x RTX 5060 Ti vs GB10: Five-Model Local AI Benchmark at 256K Context

2026-10-10 (Modified: 2026-10-15) — Written by SimeonOnSecurity — 17 min read

A five-model RTX 5060 Ti versus GB10 comparison, plus a same-host Qwen3.8 Flash-Next Strata benchmark through 256K context.

An illustration showing two NVIDIA RTX 5060 Ti graphics cards on the left and an NVIDIA GB10 system on the right, both highlighted against a dark background with vibrant colors. — Local model performance depends on memory architecture, model choice, and context length.

#RTX 5060 Ti #DGX Spark #qwen3.8 #strata #Qwen3.5

RayHunter — IMSI catcher and Stingray detector devices from STS Collective

Shop Now — Save up to 20% Today Code: SIMEONONSECURITY