C.W.K.
Stream
Lesson 01 of 06 · published

The Layered Stack — Backbone, Training, Inference, Product

~10 min · framework, axes, literacy

Level 0Scout
0 XP0/41 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete

The single biggest source of LLM-discourse confusion

Every LLM exists at the intersection of four layers — backbone, training, inference, product. Marketing language deliberately collapses them. "AI-native architecture" sounds like it means something architectural; almost always it means a thin product layer. "Reasoning engine" sounds like a new wiring; almost always it means RL post-training plus extended-thinking inference. Until you separate the layers in your head, every announcement reads like noise.

The four layers, explicitly

LayerWhat it isWhat changes here actually means
BackboneThe wiring of the network — attention, FFN, embeddings, position encodingA genuine architectural change. Rare and usually clearly disclosed.
TrainingPretraining + post-training (SFT, RLHF, GRPO, DPO, distillation)Where most "breakthroughs" of 2024–2026 actually live.
InferenceHow the model is invoked — standard, extended thinking, speculative decoding, beam searchSame weights, different runtime. Cheap to switch.
ProductScaffolding — RAG, tools, agents, system prompts, guardrails, memoryApplication code. Backbone unchanged.

The reading test

For any claim about a model, ask: "Which layer does this change live on?" If the answer is "backbone", you should be able to find it in the architecture section of the technical report — different attention type, different FFN structure, different position encoding. If you can't, the claim is on a higher layer.

Why this lesson exists

The next several lessons pick apart specific common claims — RAG, tool use, agents, long context, speculative decoding, GQA/MQA/MLA — and show which layer each lives on. None of them are backbone changes, even when the marketing copy insists they are. Once you see the pattern in concrete examples, you stop being fooled by the abstract version.

External links

Exercise

Take three model announcements from the last month. For each, list every claim it makes about the model. Tag each claim with which of the four layers it actually lives on. You will find most announcements are 80% training/inference/product and only 20% backbone — yet the language often makes it sound the other way around.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.