Garry Tan: Traditional SaaS has two options aheadRead the article English (automatic translation) |
SummaryGarry Tan (YC CEO) argues that traditional SaaS will either die as a simple 'System of Record' or evolve into a 'Domain-Specific Skeleton.' This skeleton is a messy execution layer that wraps general-purpose large language models (LLMs) with layers for tool routing, authorization, rollback, and legacy system integrations. The article expands on this single tweet to present a broader thesis: the moat of old data lock-in is dead, and the real moat is migrating toward sector-specific edge cases and fault-tolerant design. |
Why it mattersThe article places a dual distinction at the center of the 2026 SaaS investment thesis: cheapening record-keeping versus hard-won domain execution. For founders and VCs, it sharpens the question of where durable value lies at the point where LLMs turn single-point understanding into a service. It also flags a second-order risk: thin orchestration layers being absorbed by model providers, pushing the defensible layer even deeper into domain knowledge under ownership. |
Highlights
|
Critical takeThis text is essentially a single provocative tweet presented as a sector verdict. The 'third layer' (Becoming a Skill) is quoted from an anonymous comment, not from Tan. The dramatic framing ('Batman-style death sentence') and the 49/100 AI score indicate that this is editorial embellishment rather than a robust strategic analysis. This binary 'die or become a skeleton' dichotomy also ignores the many SaaS products that will continue selling records to enterprise customers who are not interested in agents. |