SYSTEM Cited by 1 source

Claude 2.1 (Anthropic)

What it is

Claude 2.1 is Anthropic's frontier large language model released November 2023 — the successor to Claude 2 and predecessor to the Claude 3 family (Haiku / Sonnet / Opus). Its 200K-token context window was the headline capability at launch.

Claude 2.1 is the LLM Slack used in their production Enzyme-to-RTL codemod pipeline as of the 2024-06 retrospective.

Seen in

• sources/2024-06-19-slack-ai-powered-conversion-from-enzyme-to-react-testing-library — Slack integrated Claude 2.1 via the internal DevXP AI team's API endpoint. Disclosed failure modes for naive single-call prompting: 40-60% conversion success with high variance. Mitigated by the AST + LLM hybrid pipeline + DOM context injection + structured prompt template, which lifted the success rate on selected evaluation files to ~80%. Canonical datum that hybrid architectural scaffolding around a 2024-era frontier LLM is required for deterministic- correctness code-transformation tasks at production scale.

Related

• concepts/llm-hallucination — generic failure mode
• concepts/llm-conversion-hallucination-control — specific problem class for code transformation
• patterns/ast-plus-llm-hybrid-conversion — the pattern Slack used