An argument piece. Operating system, agentic mesh, AI workforce — the three metaphors competing for the agentic-AI category are not equivalent, and the metaphor a category settles on determines the developer mental models, the commercial structures, and the kind of company that ends up winning.

A cornerstone essay. The Model Context Protocol and the Agent-to-Agent protocol now sit under the Linux Foundation, and the agentic internet has, for the first time, a neutral substrate. Web4 is the layer that gets built on top of it.

A cornerstone essay. Web4 is the working name for the operating layer that replaces the app — agentic infrastructure, autonomous services, and the post-app internet — and the Bulletin's editorial position is that it is already partly built.

An essay on the difference between two of the most-used category names in the autonomy conversation. The terms are not synonyms, and the Bulletin's editorial position is that flattening them costs the category clarity it cannot afford to lose.

Web3 was a financial-rail thesis. Web4 is an autonomy thesis. They are not the same category, and treating them as a continuation is the most common reason early-stage Web4 work gets miscategorized.

A comparative essay. The two most-watched coding-agent companies have radically different engineering cultures, and the cultural difference does more to explain the products than the product roadmaps do.

An opinion piece. The name has more friction than its detractors expected and more durability than its skeptics predicted, and the Bulletin's editorial position is that the reasons it is sticking are structural rather than rhetorical.

A profile. n8n closed a $60M Series B in late 2025 against a thesis that should sound familiar to anyone who reads the Bulletin: workflow automation is the application layer of the autonomy stack.

A profile. Cognition's $25B valuation rests on a thesis the rest of the field is still arguing about — that the next moat in coding agents is the ability to hold a task open for hours, not minutes.

A case study. Replit's effort-based pricing model is one of the cleaner working examples of how the autonomy layer prices itself. We work the math on what 'autonomy per dollar' actually means for a small team shipping real work.

A case study. LangChain has spent the last six quarters quietly executing the most consequential repositioning in the autonomy stack — from an open-source orchestration framework to a managed evaluation, observability, and deployment platform.

A field guide to the companies the Bulletin's directory tracks across the autonomy stack — protocol, operating system, workforce, integration substrate, interface, and the services layer above all of them.

A field guide. The trade-press framing treats the autonomy-layer category as a small set of named winners. The real picture is a long tail of working platforms — Buildship, Pipedream, Tray, Make — that are quietly absorbing the field's adjacent commercial value.

A working diagram piece. The autonomy layer is not a single product category — it is a layered stack, and the companies that confuse the layers tend to ship poorly.

An architecture piece. The Vercel AI SDK is the developer-side primitive most of the autonomy layer ends up building on, and the recent download milestone is the right moment to evaluate what the SDK has done and what it has not.

An architecture piece. The Agents API is the most under-discussed of OpenAI's recent platform releases, and the integration surface it has chosen — Responses, structured tool use, Assistants persistence — is the part the trade press has not yet caught up to.

A working bibliography for the autonomy layer. The pieces the Bulletin's contributors return to most often when they need to argue the Web4 case from primary sources.

A regional piece. Southeast Asia is producing a structurally different kind of Web4 company than the one the venture-backed Bay Area has been producing, and the region's role in the next phase of the category is larger than its current coverage suggests.

A field report. India's autonomy-layer cohort is producing structurally different companies than the Bay Area or Southeast-Asian patterns the Bulletin tracks. Sarvam, Ola Krutrim, Sarvam-of-the-stack, Devrev, and others are worth watching closely.

A speculative essay. Predictions are the Bulletin's most-criticized format and our most-cited one, and this is the consolidated version of what the editorial team thinks the autonomy layer will look like eighteen months from now.

A field report. The next decade of Web4 startups will be Asia-anchored, and the structural reasons — a neutral global AI hub in Singapore, a founder-cluster forming in Chiang Mai and Bangkok, and the conference and visa infrastructure pulling both together — are already in place.

A field report. Anthropic's Claude Code has spent twelve months becoming the de facto background-agent for a specific kind of engineering team. The pattern of adoption is more interesting than the headline numbers.

A speculative piece. Ten numbered predictions for where the Web4 stack lands twelve months from now, each tied to current evidence from the field. Bookmark for future scoring.