# Post-AI Systems Turning a private network of AI-built projects into a living public record ## Project brief - **Problem:** Substantial work with AI agents remains difficult to evaluate when its architecture, decisions, failures, and evidence stay scattered across private workspaces. - **System:** Post-AI Systems turns approved project and article snapshots into a public, navigable evidence layer without taking ownership away from their source projects. - **Input:** Versioned Markdown snapshots, project metadata, articles, visual assets, design-system rules, and corrections from live review. - **Output:** Project landings, full stories, notes, related material, and agent-readable context generated from the same approved sources. - **Evidence:** The live site publishes multiple working projects and articles through shared content contracts, project routes, story pages, context endpoints, and its own design system. - **Status:** Working public portfolio under active development, proven on one owner-operated project network rather than as a general publishing platform. - **Full story:** /projects/portfolio-site/story/ ## Full story Work with AI agents is easy to underestimate because most of it is invisible from the outside. Architecture, research, prompts, corrections, tests, failures, backlogs, and decisions may produce a serious working system, yet the public description often collapses into "I used AI to build it." Post-AI Systems was built to expose the missing evidence. It is a public surface for projects developed with AI agents, showing not only what shipped but also how the system works, what decisions shaped it, which boundaries remain, and what can be verified now. The site is organized around one question: what happens when implementation stops being the bottleneck? When software becomes faster to produce, judgment becomes more visible. The difficult work moves toward choosing what deserves to exist, structuring the system, preserving context, rejecting weak output, testing against reality, and carrying useful lessons into the next iteration. ## From private work to public evidence A finished screenshot cannot explain an agentic project. It hides the operating model behind it: the agents involved, the artifacts they use, the external systems they connect to, the evidence behind decisions, and the difference between an implemented capability and a future direction. Each project on Post-AI Systems therefore has several levels of access. A short landing page establishes the problem and current value. A Full Story explains the development and architecture. Related notes expose the thinking around the project. An agent-readable context endpoint presents the same approved evidence in a compact machine-readable form. This structure lets different readers enter at different depths without creating competing versions of the project. ## One source, several public surfaces The site is built as a content-first system using Astro, TypeScript, Markdown Content Collections, and the `@portfolio/design-system` package. Markdown is more than a writing format here. It is the contract between private project work and public presentation. Source projects prepare approved snapshots containing the title, summary, current state, evidence, limitations, and Full Story. Portfolio validates those snapshots and renders them as project cards, landing pages, story pages, notes, related content, and context endpoints. The same source can therefore serve a casual visitor, a hiring manager, a technical reviewer, or an AI assistant without requiring a separate manually maintained explanation for each one. ## Ownership stays with the source project The operating principle is simple: source projects own the meaning, while Portfolio owns the public rendering. Research, decisions, requirements, and implementation knowledge remain with the project where they were created. Portfolio receives only an approved public snapshot. It does not become a second database of private operational truth, and it does not silently reinterpret the project. This boundary keeps publication useful without making the website responsible for maintaining every source system behind it. ## The website is part of the review loop Publication is not the end of the workflow. A live page reveals problems that remain hidden in a repository: a vague title, an unconvincing summary, weak evidence, excessive detail, a missing limitation, or a visual that fails to explain the product. Those corrections return to the appropriate source. A content problem updates the project snapshot. A repeated interface problem becomes a component, token, page pattern, or design-system rule. The revised source is exported again and reviewed in production. The working loop is: 1. A source project produces verified internal work. 2. Approved evidence becomes a public Markdown snapshot. 3. Portfolio renders the snapshot through shared product surfaces. 4. Live review exposes weak content or presentation. 5. The correction returns to the source project or design system. 6. The next publication carries the accumulated lesson. The site is therefore not only a destination. It is an instrument for improving how the work is understood. ## A portfolio that AI can inspect Most portfolio sites are designed only for visual browsing. Post-AI Systems also publishes concise project context that another AI system can read before evaluating the work. This matters because future discovery may not begin with a person opening every page. A recruiter, reviewer, collaborator, or client may first ask an assistant to compare projects, identify relevant experience, or explain how a system works. The public context layer gives that assistant bounded evidence instead of forcing it to infer the project from marketing copy and screenshots. Human and machine-facing surfaces are generated from the same approved source, reducing the chance that they describe different versions of reality. ## The site proves its own argument Post-AI Systems presents work across agent infrastructure, operational knowledge graphs, design systems, publishing workflows, decision simulations, content automation, and mobile communication systems. The projects differ, but each is treated as a working system with inputs, outputs, evidence, and explicit limits. The website itself follows the same model. Its content arrives through governed snapshots. Its interface runs on a design system shaped by live corrections. Its articles and project stories are maintained outside the rendering layer. Its public pages feed new lessons back into those source systems. The site does not merely describe a post-AI workflow. It is one. ## Current boundary Post-AI Systems is a working public portfolio under active development. It already renders project catalogs, project landings, Full Stories, notes, related content, responsive visuals, design-system documentation, and agent-readable context. It is not yet a generalized publishing platform, a multi-author product, or a commercial content management system. The current evidence comes from one owner-operated network of projects. More projects exist than are currently exposed, and the public surface expands as their evidence becomes ready. Its value is not that it proves someone can build another website. It demonstrates how fragmented work with agents, code, research, design, and writing can become a coherent public system that remains readable, testable, and open to further development.