Eric Elliott

Eric Elliott is the creator of SudoLang, a language for expressing programs and constraints to AI language models, and the author of Composing Software. His AI work centers on combining natural-language instructions with programming structures—interfaces, functions, and declarative constraints—to describe complex behavior and state.[1][3][6][4]

In his June 25, 2024 O’Reilly article, Elliott described SudoLang as “designed, discovered, and curated with the help of GPT-4.” He explained that the main effort involved discovering and testing language models’ existing understanding of pseudocode, then selecting and documenting familiar concepts from programming and natural language.[3]

Elliott credits Ivan Sutherland’s Sketchpad as the inspiration for SudoLang’s natural-language constraint programming. Rather than specify every implementation step, a programmer describes desired behaviors and relationships, such as keeping a calculated value updated when a variable changes. Typed interfaces define data structures and operations, with types often inferred. Function composition connects smaller operations: g |> f applies g first, then passes its output to f. These instructions rely on a language model’s interpretation.[3]

Elliott illustrated the approach with a text adventure based on an interaction with Claude 3.5 Sonnet. Composed interfaces represent characters, quests, and objectives. When the player acquires an Ancient Scroll, the demonstration adds it to inventory, marks the fetch objective finished, and starts the objective to return to the village elder. The example makes his intended use concrete: coordinating narrative responses with structured state updates.[3]

The SudoLang repository draws a practical boundary: “For most simple prompts, natural language is better. Use it.” It recommends SudoLang for interactions that require following a program, obeying constraints, tracking complex state, or implementing algorithms. It also documents /commands for chat or programmatic interactions and editor syntax highlighting.[4]

Before his 2024 explanation, Elliott advocated structured prompting in the Latent Space episode published June 14, 2023, about OpenAI’s Functions API. He recommended defining an interface inside a prompt for the function calls a developer wanted. He reported that pseudocode had helped models obey constraints and produce intended outputs in his experience, and invited comparisons of function-call reliability under different prompting approaches. His recommendation was exploratory: he explicitly said he had not yet had much opportunity to try the new API.[1]

Elliott describes himself on his GitHub profile as founder and CEO of Parallel Drive, building AI agents for software. He links that work to the AIDD Framework, which combines reusable agent skills, orchestration, code libraries, and CLI tools. Its documented workflows pair specifications and systematic task planning with test-driven development and automated code review.[2][5]

Elliott also describes a former engineering-management role in generative AI for Adobe Firefly, saying he ran Firefly Client for most of 2024.[2]

Elliott’s career account also reaches back to 2012: he associates work on the Wall Street Journal’s WorldStream with a Webby nomination that year and identifies himself as tech lead on the social-video AI app spun out of Stanford Research Institute. He also lists JavaScript consulting and security work, including security architecture at MyCrypto and dApp integration leadership at Exodus Wallet.[2]

His book Composing Software, subtitled An Exploration of Functional Programming and Object Composition in JavaScript, teaches how to build complex applications from simple components. Its coverage of function and object composition provides context for the composition concepts he explained in SudoLang. The book began as a blog-post series; its publisher lists topics including pure functions, immutability, higher-order functions, and object composition.[6][3]