About
A research and teaching project for AI-assisted computer architecture
Architecture 2.0 is the engineering discipline of using AI, grounded in architectural representations, tools, and experiments, to formulate, explore, implement, evaluate, explain, and defend computer architecture decisions.
The idea
Computer architects formulate problems, explore alternatives, implement mechanisms, evaluate them with tools and experiments, explain why results changed, and defend decisions under constraints. AI can now assist across that work, extending what a team can consider and evaluate. Architecture 2.0 asks how to use that assistance without detaching a result from the architectural representations, mechanisms, evidence, and judgment that make it credible.
Design loops remain useful ways to organize execution and analysis, but they do not define the field. Much of the effort here defines the state an agent reasons over, the actions it may take, the rejection criteria that disqualify a bad result, and the evidence that makes a claim believable. The book develops that vocabulary; the tool registry collects the simulators, models, and harnesses used to carry it out.
Shared infrastructure
Architecture 2.0 is organized around the shared infrastructure required for credible AI-assisted architecture.
Turning machine learning loose on architecture needs shared datasets, comparable benchmarks, inspectable tools, and reproducible evidence. It also needs people trained to work across both fields. Architecture 2.0 grew out of research at Harvard and exposes a set of artifacts that other groups can inspect, test, and extend. The work organizes into a few strands:
Datasets & benchmarks
Shared, versioned data and comparable benchmarks so results mean the same thing across groups.
Methods
The learning, search, and agentic methods that generate, predict, and optimize designs inside bounded architecture studies.
Tools & infrastructure
Simulators, proxy models, and verification harnesses that connect methods to architectural evidence.
Reproducibility & evidence
Practices that let reviewers replay declared parts of a study and check the evidence behind an architectural claim.
Education & workforce
Courses and materials for a generation of architects fluent in both systems and machine learning.
Where it started
The community behind this work has been meeting for years. Since 2020, the MLArchSys workshop at ISCA (the International Symposium on Computer Architecture) has brought the machine learning, systems, and architecture communities together around learning for hardware and hardware for learning, including meetings in Tokyo (2025) and Raleigh (2026).
As foundation models and autonomous agents began to reshape how systems are designed, that community turned toward agentic approaches. MLArchSys 2026 added a dedicated A³ (Agentic Approaches to Architecture) track, and the Architecture 2.0 workshop at ISCA 2026 focused on agentic AI for computing-systems design, anchored in computer architecture and hardware/software co-design. An earlier gathering, opened by a keynote from Partha Ranganathan (Google), helped map the datasets, tools, and training the field would need.
Architecture 2.0 itself grew out of Harvard's CS249r graduate seminar (Fall 2025) and the book developed alongside it. It is part of the mlsysbook.ai family of open, community-built learning resources.
Take part
Use the hub as the front door for reading, contributing, and teaching.
Add a tool
Built an open simulator, surrogate model, or agentic loop? Propose it for the registry.
Discuss and review
Use Discussions for questions or proposals that should remain searchable, and contribute reviewed changes on GitHub.
Teach with it
Adapt the book and course materials for your own students and reading groups.