Pick two: The Impossible Triangle of LLM Infra
[HPP] SwyxNovember 19, 20259 min
18 connectionsΒ·25 entities in this videoβUnderstanding Impossible Triangles
- π‘ The impossible triangle is a mental model for understanding developer tooling, where only two out of three competing priorities can be optimized simultaneously.
- π― This concept is seen in various fields, including macroeconomics (impossible trinity), the CAP theorem in databases, and the evolution of frontend frameworks.
- π The speaker, an economist by background, applies this perspective to analyze developer tooling and frameworks as an investor.
Impossible Triangles in Tech
- π Examples extend to FinTech (e.g., Mercury, Ramp, Brex competing for features) and data engineering (e.g., DBT and Fivetran merger due to TAM limitations).
- π§© Even fundamental cloud components like networking, compute, and storage form an impossible triangle, with various cloud offerings representing combinations of these.
- π These triangles can nest within each other, from broad market segments to specific tech choices like browser automation, web search, and web scraping.
LLM Infrastructure Trade-offs
- π§ The impossible trinity for LLM infrastructure comprises Model/API Hosting, Frameworks, and Observability.
- β οΈ Companies in this space, including model labs and API providers, often face trade-offs and expand into adjacent areas to offer more complete solutions.
- π οΈ An LLM API platform bundles various jobs to be done; if a platform doesn't offer a feature, developers must build it themselves or wait for the platform to integrate it.
Mastra's Role and Developer Choices
- π± Mastra, initially a TypeScript framework, has predictably expanded its offerings to include observability, evals, and model routing, aligning with the impossible triangle concept.
- β Developers are encouraged to identify the bundle of tools they need by deeply understanding the underlying problems.
- π§ The core principle is that problems remain constant, while solutions evolve, providing a robust basis for evaluating new technologies and making informed choices.
Knowledge graph25 entities Β· 18 connections
How they connect
An interactive map of every person, idea, and reference from this conversation. Hover to trace connections, click to explore.
Hover Β· drag to explore
25 entities
Chapters5 moments
Key Moments
Transcript37 segments
Full Transcript
Topics15 themes
Whatβs Discussed
Impossible TrianglesDeveloper ToolingMacroeconomicsCAP TheoremFrontend FrameworksData EngineeringCloud ComponentsLLM InfrastructureModel HostingAI FrameworksObservabilityMastraTypeScriptAI EngineeringMonetary Policy
Smart Objects25 Β· 18 links
ConceptsΒ· 14
CompaniesΒ· 6
EventΒ· 1
PersonΒ· 1
ProductsΒ· 3