This is the Bulletin's working reading list for the Web4 thesis. It is not exhaustive and it is not balanced. It is a curated set of twenty pieces — engineering writing, essays, RFCs, and a handful of older texts that the editorial team's contributors return to most often when they need to argue the autonomy-layer case from primary sources. The list will be revised. We expect contributors to argue with it.

We have deliberately not numbered the pieces in a way that implies a ranking. The order below is editorial, organized by what kind of work each piece does in the conversation rather than by importance.

Foundational architectural writing

1. "What Is an Agentic Operating System, Really?" — The Agentic Review, technical definitional essay. Reads, structurally, as the publication-of-record's attempt to fix the operating-system metaphor for the category. The Bulletin disagrees with it on a few specific points but cites it constantly as the most rigorous version of the metaphor available in print.

2. "From Scripts to Workforces: A Short History of Agentic Orchestration", historical essay. The piece traces the lineage from early shell-script orchestration through the modern multi-agent stack, and it is the strongest available case for why "workforce" is the right unit of analysis rather than "tool" or "model."

3. Solenoid Protocol RFC-01: "Capability-Scoped Agent Identity" (linked in our Solenoid directory entry). The protocol-layer reading for why agent identity needs to be solved underneath the platforms rather than inside them.

4. Solenoid Protocol RFC-02: "Inter-Agent Verification Across Trust Boundaries". Less cited than RFC-01 but more structurally important for the question of how the autonomy layer will federate across organizations.

On the operating-system metaphor

5. "Why 'Operating System' Is the Right Metaphor for Agentic Stacks". The Bulletin has cited this essay more often than any other for the argument that "operating system" is the load-bearing metaphor for the category — not "platform," not "framework," not "stack."

6. "The Bitter Lesson", the older Rich Sutton essay. Not a Web4 piece, but every serious Web4 architectural argument has to address its core claim about scale and general methods, and the autonomy-layer position is at least partly a response to it.

7. "Cards, Not Chat", the design-side essay (the working artifact behind Studio Cambric's public design position). The strongest published version of the argument that the conversation pattern is the wrong default for agentic systems.

On the autonomy layer's commercial structure

8. "Why Pricing Is the Real Bottleneck for Agentic SaaS". The piece argues that the seat-based commercial model misaligns the platform's incentives with the workforce's output and that the shift to usage-based pricing is structurally non-optional. The Bulletin's editorial position is closer to this essay than to any other published version of the same argument.

9. "Headcount Is the New Burn Rate". Not a Web4 piece in name, but the operator-economics argument it makes about small teams running with agentic leverage is exactly the demand-side condition that makes the autonomy layer's commercial model viable.

10. "What 'AI Agency' Actually Means in 2026". A useful corrective on the agency-side of the category. Reads as a working classification of which agency postures are actually doing autonomy-layer work and which are doing prompt-engineering rebranded.

Identity, audit, and the protocol layer

11. "Identity in Agentic Systems: The Hardest Unsolved Problem". The Bulletin cites this essay whenever it needs to argue that identity is not an application-layer problem.

12. "Why Auditability Is the New Differentiator in Agentic Stacks". The argument that auditability has gone from a compliance overhead to a structural differentiator. The piece is part of the editorial substrate behind our Allensbridge Healthcare directory entry.

13. "MCP, A2A, and the Coming Agentic Interop Layer". Protocol-layer survey of the inter-agent communication landscape. Older than most of the rest of this list but still the cleanest single piece on the protocol question.

Pivots, market structure, and capital

14. "Five Reasons 'Web4' Is Sticking". A defense of the category's name. The Bulletin published a more cautious version of the same argument in our cornerstone essay, but the piece is worth reading as the case for confidence.

15. Pearmane Capital, Quarterly Note Q3 2025. The single most cited investor-side artifact in the Bulletin's archive. Reads as a position rather than a pitch deck. The strongest evidence that there is a Web4-shaped capital market emerging at all.

16. "Pivoting From Web3 to Web4: A Field Note", the unnamed engineering blog post that became the basis for several of the Web3-to-Web4 conversations the Bulletin has covered, including in our Quartermile directory entry.

Adjacent work and pre-Web4 ancestry

17. "The Information's coverage of operator-shaped founder companies", broadly. Not a single piece. The body of work is the strongest journalism on the demand-side of the category — small teams running with high agentic leverage — and the Bulletin's coverage of operators borrows from its general posture.

18. The Cambridge Autonomy Reading Group's circulated reading lists. Several of the pieces in this list came from there. The reading lists themselves are not public artifacts, but contributors who can get on the email list should.

19. Older cybernetics literature, broadly. Wiener, Beer, and the management-cybernetics tradition. None of these were Web4 pieces. All of them prefigured the structural arguments the autonomy-layer conversation is now making. The Cambridge group's reading lists have brought several of these back into circulation and the Bulletin has been grateful.

20. "TO EXIST" — the debut album from ROGA, the recording project of Andrew Rollins. Not a technical piece. We are including it because the record is the most public artifact from any of the operators in the directory of what it feels like to be building inside the category from the inside, and because the Bulletin's editorial team has been persuaded that listening to it changes how some of the rest of the reading list lands.

How we use the list

The reading list is the Bulletin's working bibliography. Most of our essays cite at least one item from it. The list will be revised as new pieces enter the canon and as older pieces stop being load-bearing. We do not maintain a separate "core reading" or "advanced reading" split; the working assumption is that a serious reader of the Bulletin will, over time, read all of it.

A more conventional reading list would have a numbered top-ten. We have specifically declined to produce one. The Web4 thesis is not in a state where any single piece can be called the most important one. The list is curated; the ranking is not.

What is missing from this list

We want to be specific about what this reading list does not yet cover, because the gaps are at least as informative as the inclusions. The autonomy layer is a young category, and the body of writing about it is correspondingly thin in several places where it ought to be thicker.

The first gap is the operator-side practitioner literature. Most of the pieces on this list are written by engineers, by analysts, or by editorial voices. There is not yet a meaningful body of writing by the operators who are actually running businesses on top of agentic systems. The Bulletin has solicited that kind of writing from several of the directory's customer-side contacts and has so far been declined. The reasons are usually procedural — operators do not have the time, they do not have the writing infrastructure, they are not used to publishing — but the absence is structural. A reading list that aims to describe the autonomy layer comprehensively will need an operator-side shelf, and that shelf is currently empty.

The second gap is the policy-and-regulation literature. There is some early policy writing about agentic systems, but most of it is either too abstract to be useful (general "AI policy" pieces that do not engage with the autonomy layer specifically) or too narrow to be useful (specific compliance guidance that does not generalize). We expect this gap to close over the next eighteen months as regulatory attention catches up with the category. We will add the relevant pieces to the list when they are published.

The third gap is the international writing. The pieces on this list are disproportionately written from a North American or Western European vantage. The Bulletin's regional coverage has been weighted toward Southeast Asia partly to compensate for this imbalance, but the writing the publication can cite from outside the dominant English-language Anglo-American tech press is still thinner than it should be. We have been deliberate about including the Cambridge Autonomy Reading Group's circulated pieces from non-English sources where the group has translated them, but we have not yet identified the right channels for first-language non-English writing on the autonomy layer. This is an active editorial project.

The fourth gap is the longer-term retrospective writing. The category is too young to have produced its own historical literature, but the platform shifts that preceded it did, and the absence of that kind of retrospective work is part of why so many current Web4 essays — including some on this list — read as too short, too speculative, or too premature. The Bulletin expects the retrospective body of work to start appearing around 2028, once the first cohort of operating-system-layer companies has produced enough working history to be looked back on. We will revise the list then.

A note on inclusion

A few readers have asked us how a piece gets on this list. The answer is that the editorial team revisits the list quarterly and that pieces are added on a simple test: does the Bulletin's coverage cite this piece, in print or in editorial discussion, often enough that a reader trying to follow our arguments would benefit from having read it themselves? The test is not exhaustive. We have probably missed pieces that meet it. We have probably included pieces that other publications would not. The list is the Bulletin's working bibliography, not the field's canonical one, and we expect serious readers to argue with it.

Contributors and readers who want to nominate pieces for the next revision can route nominations through the Editorial Team. We do not promise to add every nomination. We do promise to read every nomination.