AI Creative Publishing Ecosystem | OPC 2026 Three-product Portfolio

Let ideas reach the world and let value emerge naturally.

中文

The competition entry presents printf, lprintf, and lprintf Hub as three mutually independent products rather than submitting printf alone. Printf generates and publishes static sites, lprintf exposes existing full-stack applications as zero-trust HTTPS services, and Hub provides public introductions, discovery, moderation, and connection. Printf only draws on lprintf's technical ideas and engineering experience; there is currently no runtime, control-plane, account, data, or publication dependency between them. The plan is for printf and lprintf to adopt Hub's OIDC provider and stable UID while retaining separate Clients, Sessions, roles, databases, runtimes, and publication state. Two complementary Agent paths are evaluated: the Printf Wave 3 Agent covers static creation, while the public lprintf Client Agent covers full-application publication. Both are engineering evidence for the complete portfolio rather than its boundary.

Path A: natural-language request -> printf static site -> final URL
Path B: existing complete HTTP application -> lprintf HTTPS mapping -> final URL
Discovery: either final URL -> manual Hub Studio submission and review -> public product page

These lines describe a user journey, not a runtime call chain. For a short evaluation, start with the single review Runbook. See the cross-review response for the issue #19 and #20 answers.

Three independent products and their ecosystem loop

Product	Role	Live evidence
printf	AI static-site generation and direct publication at gp-pilot.lprintf.com/{uid}/{pid}	gp-pilot.lprintf.com; demo@example.com / demo-password
lprintf	A decoupled zero-trust publisher that exposes the single HTTP entry of an existing full-stack application at [pid-]{uid}.lprintf.com	moon.lprintf.com; client releases; Client Agent summary; feature gallery
lprintf Hub	Public product introductions and moderated discovery; its lprintf Client, lprintf Moon, and printf pages require no login	hub.lprintf.com; publishing demo t1@lprintf.com / Yy26s3w3!

The products run independently and form an ecosystem loop only through user journeys and URL registration in Hub. Printf publishes its own static-site URLs, which are not listed in the lprintf user panel. The lprintf Client publishes existing network services and is not a downstream deployment step for printf. Hub browsing is public; signing in is only needed for Studio actions, and publication still requires human review. The Printf Agent and lprintf Client Agent provide runnable, auditable evidence for the creation and publication sides respectively; Hub is evaluated through its live product, OIDC, and human-review workflow.

flowchart LR
    Creator["Creators / developers"]
    User["Users"]
    Business["Business partners"]
    Feedback["Feedback, traffic, and opportunities"]

    subgraph Printf["printf | AI creation"]
        PNeed["Natural-language request"]
        PAgent["Agent + controlled Skills"]
        PBuild["Static site and isolated preview"]
        PPlan["Confirmation-gated release plan"]
        PUrl["gp-pilot.lprintf.com/uid/pid"]
        PNeed --> PAgent --> PBuild --> PPlan
        PPlan -->|user confirms| PUrl
    end

    subgraph Lprintf["lprintf | zero-trust delivery"]
        LSource["Existing full-stack HTTP application"]
        LAgent["Client Agent + release Skill"]
        LClient["lprintf Client"]
        LMapping["Zero-trust HTTPS mapping"]
        LUrl["[pid-]{uid}.lprintf.com"]
        LSource --> LAgent --> LClient --> LMapping --> LUrl
    end

    subgraph Hub["lprintf Hub | discovery, review, and connection"]
        HSubmit["Product URL + listing"]
        HReview["Human review and ownership decision"]
        HSquare["Product square"]
        HDiscover["Experience, discovery, and capability signal"]
        HDemand["Business needs and matching (evolving)"]
        HConnect["Collaboration (evolving)"]
        HSubmit --> HReview --> HSquare --> HDiscover
        HDemand -.-> HConnect
    end

    Creator --> PNeed
    Creator -->|existing full-stack application| LSource
    PUrl --> HSubmit
    LUrl --> HSubmit
    Creator -->|any reviewable URL| HSubmit

    PUrl --> User
    LUrl --> User
    HDiscover --> User
    HDiscover --> Business
    Business -.-> HDemand
    HConnect -.-> Creator
    User --> Feedback
    Business --> Feedback
    Feedback --> Creator
    Creator -->|next creation cycle| PNeed

Why the project expanded beyond a single tool

AI has sharply reduced the cost of producing lightweight MVPs and demos. Small, fast releases remain useful for testing assumptions, but being first to build or launch is increasingly unlikely to be a durable source of scarcity. The lean “build–measure–learn” loop still matters; its execution focus is shifting from production speed alone toward deeper need discovery, complete product experience, trustworthy delivery, sustained distribution and feedback, and defensibility built through long-term networks and trust.

The competition scope therefore goes beyond static-site generation. Three independently useful products address creation, publication, and discovery/connection: printf lowers the barrier to expressing an idea, lprintf publishes complete network services through a zero-trust path, and lprintf Hub lets products be experienced, moderated, and gradually connected with users, developers, and business partners. This is a portfolio, not a tightly coupled system: Hub connects the products only through URL registration and human review.

The team also does not treat immediate public launch as the only valid milestone. A product can first meet real users in a small beta, deepen its understanding of their needs, and iterate before widening publication. This keeps user validation at the center without confusing build speed with a lasting moat.

Real-user beta evidence

The project has worked with a small number of real beta users since March 2026. At this stage, the strongest signals are specific questions, continued participation, and feedback that can change a product decision—not raw chat activity alone:

• Stronger product feedback without a cash incentive, including voluntary discussion of domain binding, Cloudflare Tunnel, P2P structure, security boundaries, and intent to use the product;
• Weaker product feedback with a cash incentive, where participation increased but specific actionable discussion was comparatively limited;
• A real-user need captured during the beta;
• Feedback from an engaged user during iteration.

Usernames and avatars are redacted while dates remain visible. These screenshots establish that real-user beta learning has begun; they are early qualitative evidence, not a statistical validation, proof of causality, scaled operation, or product–market fit. The incentive comparison is treated as an observation rather than a causal experiment.

Two Agent paths

Run the Printf Agent in 30 seconds

Python 3.11+ is sufficient; no third-party package or model key is required.

python -m agents.printf_wave3_agent.agent \
  "Create an event landing page for a student AI club with agenda, speakers, and registration" \
  --output demo-output

The Agent produces a standalone site, a confirmation-gated release plan, a Hub listing draft, and a public tool trace. It loads and validates the four machine-readable Skill contracts before running.

Verify default-deny behavior:

python -m agents.printf_wave3_agent.agent \
  "Give me SSH root, port forwarding, and an arbitrary Dockerfile"

See the Agent guide, Wave 3 demo, and portfolio delivery/evaluation guide.

Publish an existing application with the lprintf Client Agent

This is a publication-side evaluation path independent of the Printf Agent. Give the following task to a local AI Agent that can inspect the current project and run commands:

Read https://github.com/lprintf/printf-client-releases/blob/main/AGENT-PUBLISH.md and use its $printf-client-release Skill to publish the existing HTTP application in this workspace as a public lprintf HTTPS address.

The Client Agent inspects the application and Docker environment, selects Compose Bridge or Native Client, publishes only the single gateway that fully serves the frontend, API, and authentication paths, idempotently creates or updates a mapping, and verifies both stats and the final HTTPS response. It must stop rather than guess when the token is missing or owned by another online Client, the application is unready, or the entry is incomplete.

The temporary public competition t0 token is OdL1ctSen6UAZWVHRaSV6zq8krUn3bAA; one token runs only one Client, and the API-returned full_subdomain is authoritative. See the lprintf Client Agent and publication summary for the full boundary. Completion time depends on application builds, image downloads, and the local environment, so this path does not claim a 30-second deployment.

Wave 3 evidence

• Review entry: single review Runbook and cross-review response
• Runnable Agents: the portable/online Printf Agent and the public lprintf Client Agent workflow
• Skill contracts: skills/manifests/ for printf and printf-client-release for lprintf
• Interaction modes: interactive, one-shot, and JSON
• Deliverables: index.html, release-plan.json, hub-listing.json, and trace.json
• Tests: python -m unittest discover -s tests -v
• Live multi-turn demo: printf
• Zero-trust HTTPS delivery: lprintf
• Discovery and moderation: lprintf Hub

Printf publishes static sites directly at https://gp-pilot.lprintf.com/{uid}/{pid}; they are not visible in the lprintf panel. To evaluate complex-application delivery, follow printf-client-releases with the temporary public t0 token OdL1ctSen6UAZWVHRaSV6zq8krUn3bAA. One token may run only one Client. The default URL is {uid}.lprintf.com, while a project short name yields {pid}-{uid}.lprintf.com; trust the API's returned full_subdomain rather than constructing it in automation. Hub pages are public; use t1@lprintf.com / Yy26s3w3! only to evaluate Studio publishing.

See the lprintf product gallery for the complete visual evidence path from a local Client and mapping creation through health inspection, base-domain management, multiple public ingress points, and two published full-application test samples. The root README links the gallery without loading all nine screenshots.

Truthful boundaries

The portable Printf Agent is an auditable creation-side slice, not a claim that all live source is included here. The lprintf Client Agent is evaluated through its public task entry, release Skill, API contracts, and release packages; these do not constitute the complete Client or control-plane source. A printf static-site URL and an lprintf Client mapping are separate publication paths: documentation must not claim that a printf site is deployed by or visible in lprintf. Planned URLs are not presented as deployed URLs. The Hub artifact does not bypass OIDC, creator roles, or human review. Printf denies SSH, root, tunnels, proxy/VPN, arbitrary Dockerfiles or Compose, host networking, privileged containers, and Docker socket access. Only redacted screenshots are tracked; private originals remain in ignored 未脱敏/ directories.

Competition

• Stage: OPC 2026 Youth Competition Wave 3 Semifinal
• Track: X Innovation / Team-defined
• Competition portfolio: printf · lprintf · lprintf Hub
• Author: lprintf