Principles of AI-Native System and Chip Design
Start with the card · Website · Synthesis lecture · Labs · Tool registry · Readings · Workshops · Discussions
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. This public research and teaching project develops that discipline through a synthesis lecture, companion labs, community resources, and a versioned design-loop card.
Important
Work in progress. This material is under active development. Arguments, terminology, examples, labs, and machine-readable interfaces may change as evidence and reader feedback improve. Tagged releases are stable snapshots; if you cite or build on the project, record the version you used. Corrections and technically grounded disagreement are welcome.
AI can assist across an architecture study, from framing a question and constructing alternatives to testing mechanisms and defending a bounded recommendation. A design loop organizes that work around architectural state, tool feedback, evidence, and human judgment. The card records the question, inputs, tool runs, results, and decision for one study; it does not define the discipline.
Architecture 2.0 is part of the mlsysbook.ai family.
| If you want to... | Go here |
|---|---|
| Draft and validate a context-profile design-loop card in 30 minutes | Start workflow |
| Run a tool-backed student study and inspect its run archive | Companion labs |
| Download the machine-checkable blank card | YAML template |
| Read the core argument | Architecture 2.0 synthesis lecture |
| Find tools, simulators, benchmarks, datasets, and loop infrastructure | Tool registry |
| Orient yourself with papers, posts, talks, and datasets | Reading list |
| Find venues where this work is being discussed | Workshops |
| Submit a tool, dataset, benchmark, or artifact | Submit |
| Ask questions or propose ideas | GitHub Discussions |
The project accepts proposed tools, readings, workshops, corrections, and examples through structured GitHub forms. Maintainers review each proposal before it becomes part of a public registry.
| Contribution | Submit |
|---|---|
| Tool, dataset, benchmark, model, or workflow | Submit an artifact |
| Paper, post, talk, dataset, or reference | Submit a resource |
| Workshop, tutorial, CFP, or venue | Submit a workshop |
| Correction or clarification | Suggest a correction |
| Broken link | Report a link |
Accepted contributions can be credited in this README through the All Contributors convention.
The public site connects the following maintained surfaces.
| Surface | Role |
|---|---|
| Synthesis lecture | Develops the discipline of using AI to formulate, explore, implement, evaluate, explain, and defend computer architecture decisions. |
| Design-loop card | A versioned twelve-field review record with machine-checkable claims, evidence, independent profiles, replay, and decision rights. |
| Companion labs | A separate tool-backed practice path for producing evidence and replayable run archives without replacing the synthesis lecture's executed study. |
| Tool registry | A maintained index of simulators, proxy models, verification harnesses, benchmarks, datasets, and data representations, each tied to a role in an architecture design loop. |
| Reading list | A curated path through papers, SIGARCH posts, talks, datasets, and workshop writeups. |
| Workshop archive | Verified active calls when available, with prior events retained as an archive. |
| Discussions | A public place to propose questions and ideas when durable discussion is useful. |
| Path | What it contains |
|---|---|
book/ |
Quarto source for the synthesis lecture, including chapters, appendices, figures, and references. |
labs/ |
Tool-backed exercises, the executed study, replay support, and tests. |
design-loop-card/ and schemas/ |
Human-readable templates and machine-checkable contracts for recording a study. |
examples/ |
Completed cards and supporting evidence that show how the contracts are used. |
www/ and tools/ |
Source for the project site, community resources, and tool registry. |
data/source-receipts/ and compliance/ |
Source notes, permissions records, and publication checks. |
The repository is developed in public so that claims, examples, interfaces, and teaching material can be inspected and improved together. A tagged release captures a reviewed snapshot; work on the development branch may change before the next tag.
Use the HTML or EPUB edition for reflowable reading and screen-reader semantics. The current Springer-layout PDF is untagged and may not work reliably with screen readers.
People who contribute tools, readings, workshops, corrections, examples, datasets, or artifact pointers help make Architecture 2.0 useful to the discipline.
Contributor guidance, environment setup, local build commands, branch policy, and validation checks live in CONTRIBUTING.md. The public site is published at arch2.mlsysbook.ai.
Cite the tagged version you used rather than the development branch. The
repository carries a CITATION.cff entry, which GitHub renders
through the Cite this repository control in the sidebar. Tagged snapshots
are listed under Tags.
Copyright 2026 Vijay Janapa Reddi. The manuscript, figures, and supporting materials are readable in public so that claims and teaching material can be inspected, but that visibility is not a grant of reuse rights. See LICENSE.md for the terms that currently apply before copying, redistributing, or building on this material.