The closing part of season one: a practical roadmap for people who start without background knowledge and want to reach local models, adapters, gateways, and software development.
Two trends in optical infrastructure are about to run into each other. Hyperscaler AI fabrics push aggressive disaggregation — programmable transport, vendor-neutral interfaces, customer-controlled SDN. Carrier DWDM, where most of the
Discover the importance of hard boundaries in AI security with MAGATAMA and ShieldX. Learn how to prevent prompt injection and protect sensitive information.
Explore the real timeline for IEEE 802.3 Ethernet standards beyond optimistic projections. Plan your network deployments with accurate information.
Building software with LLMs requires careful verification of changes to maintain quality and control.
Explore the hurdles preventing co-packaged optics from reaching volume production before 2028. Understand yield issues, lengthy qualifications, and cautious customer behavior.
After each training run, take a cooldown to evaluate your results. This helps avoid unnecessary training runs and ensures better outcomes.
Adapters and LoRA in plain language: not a completely new model, but a small learned behavior layer on top of one.
An LLM gateway is a front door for AI work: one place for routing, safety checks, audit, fallback, and cost awareness.
I just got back from three days of a regional NOG meeting. By the end of day one I had fourteen good ideas. By the end of day three I