SCOI Atlas · L2 Models

L2A Foundation & Multimodal Models

Frontier models from OpenAI, Anthropic, Google, Meta, xAI, Mistral, DeepSeek and their open-weight peers.

TL;DR · Direct answer

L2A Foundation & Multimodal Models is a sublayer of L2 Models in the Supply Chain of Intelligence™ (SCOI) by Anand Arivukkarasu. It is a required capability but not on its own a durable moat. Do not build a moat here. Do build the discipline to swap the underlying model quarterly without rewriting your product.

What actually matters at L2A

  • Frontier capability is now a commodity within ~6 months of release.
  • Model choice is a reversible decision; architect for that.
  • Multimodal (audio, vision, video) shifts the surface layer more than the model layer.

The startup lens

Do not build a moat here. Do build the discipline to swap the underlying model quarterly without rewriting your product.

Vertical lens — how this plays across categories

Vertical AI SaaS

Prompt + eval abstraction over multiple providers; buy on price/quality.

Regulated verticals

Prefer providers with SOC2, HIPAA, DPA-friendly terms.

Frontier-dependent products

Have a documented plan for 'what breaks when GPT-N+1 ships'.

How to defend L2A

  • Model-agnostic architecture.
  • Eval suites that decide model choice, not vendor relationships.

Other sublayers in L2 Models