Framework · 2023–now · AI Present
LLM app frameworks
LangChain-class glue gave way to vendor SDKs and thin wrappers. The durable pieces are boring: evals, retrieval, and product UX.
LLM frameworks mutated because the models underneath changed faster than semver. The fad was treating prompt glue as architecture. By 2026 the category consolidated — framework repos going quiet, first-party tool APIs winning. What sticks looks like observability and eval pipelines, not chain-of-thought cosplay.
Case studies
Patterns
Context
Demos consolidate; evals remain
AI pair programming changed how code is typed faster than how it is reviewed. Agent frameworks and standalone vector stores sorted into demos versus durable plumbing; mid-market RAG folded back into Postgres. The permanent layer is familiar: evals that gate deploys, model routing for cost, tool protocols instead of plugin snowflakes, and humans who own production. Autopilot rewrites and vibe-shipped auth middleware are still big-bang migrations with better slides — and a longer on-call.
Compare with
Related
Platform · 2022–now
Vector DB gold rush
Specialized embedding stores sold as the default. Mid-market folded back into Postgres — after the vendor tour.
Language · 1991–now
Python
From scripting glue to the default language of data and AI. Readability as a growth strategy.
Practice · 2021–now
AI pair programming
Autocomplete that understands the file — now table stakes. Generation outran review capacity; ownership is the bottleneck.
Framework · 2023–now
LLM orchestration frameworks
Chains, tools, memory, and agents as a framework — LangChain-class glue sold as architecture. By 2026: LangChain-fatigue and thinner SDKs.
$ Teams adopted chain/agent frameworks before they had a single reliable tool call. Abstraction layers multiplied while prompt quality stayed flat.
Practice · 2023–now
RAG as default architecture
Retrieval-augmented generation as the answer to every knowledge problem — then "just stuff the window" as the counter-fad. Often right; often Postgres with pgvector would suffice.
$ Teams stood up vector pipelines, chunking strategies, and rerankers before asking if fine-tuning or a SQL query would answer the question. Retrieval infra became the product.