Twenty-five people whose frameworks, research and infrastructure the agent era is being built on. Unranked within the list — ordering is alphabetical by contribution area, not a league table.
| # | Name | Base | Why they are on it |
|---|---|---|---|
| 1 | Harrison Chase | LangChain | Agent frameworks used across the industry |
| 2 | Simon Willison | Independent / Datasette | LLM tooling and the most-read practitioner blog in agents |
| 3 | Chip Huyen | Author, AI Engineering | The reference text for building LLM systems |
| 4 | Swyx (Shawn Wang) | Latent Space | The AI engineer community's town square |
| 5 | Andrej Karpathy | Eureka Labs | LLM engineering education at massive scale |
| 6 | Jerry Liu | LlamaIndex | Data frameworks for LLM applications |
| 7 | João Moura | CrewAI | Multi-agent orchestration |
| 8 | Hamel Husain | Independent consultant | LLM evals practice, taught to thousands |
| 9 | Eugene Yan | Applied scientist and writer | Applied LLM patterns and evals |
| 10 | Shreya Rajpal | Guardrails AI | Guardrails and validation for LLM apps |
| 11 | Ankur Goyal | Braintrust | Eval and observability infrastructure |
| 12 | Jason Liu | Instructor / 567 Labs | RAG and structured extraction practice |
| 13 | Lilian Weng | Thinking Machines Lab; ex-OpenAI | The 'LLM Powered Autonomous Agents' survey everyone cites |
| 14 | John Yang / Carlos Jimenez | SWE-agent / SWE-bench (Princeton) | Benchmarks that made coding agents measurable |
| 15 | Graham Neubig | CMU; OpenHands contributor | Open coding-agent research |
| 16 | Ofir Press | SWE-bench / SWE-smith | Agent evaluation datasets |
| 17 | Robert Nishihara | Anyscale / Ray | Infrastructure agents and models run on |
| 18 | Ion Stoica | UC Berkeley; Anyscale, Databricks co-founder | Systems for AI at scale |
| 19 | Matei Zaharia | Databricks co-founder; Stanford | ML platforms and compound AI systems |
| 20 | Omar Khattab | Stanford; DSPy creator | Programming — not prompting — language models |
| 21 | Tri Dao | Princeton; Together AI | Efficient attention and open model systems |
| 22 | Arthur Mensch | Mistral AI | Open-weight frontier models |
| 23 | Thomas Wolf | Hugging Face | Open model ecosystem |
| 24 | Clem Delangue | Hugging Face | The hub the agent ecosystem ships through |
| 25 | Nathan Lambert | Allen Institute (Ai2) | Open models and post-training research |
We weighted shipped, widely-used work: frameworks in production, benchmarks the field trusts, research others build on, and infrastructure that quietly carries the ecosystem. Vendor marketing reach counted for nothing. Disagree? Nominations for the 2027 list are open — see also Emerging and Young builders.
We weighted shipped, widely used work: frameworks in production, benchmarks the field trusts, research others build on, and infrastructure that quietly carries the ecosystem. Vendor marketing reach counted for nothing.
No. The 25 people on it are unranked within the list — ordering is by contribution area, not a league table. It is a map of who the agent era is being built on, not a countdown.
The list is reviewed quarterly and nominations for the 2027 list are open. Self-nominations are considered on the same evidence as any other: shipped work the field actually uses.