Project Stardust From Adobe

Stardust represents a new class of AI driven web transformation systems that can redesign migrate and deploy websites with a level of automation that previously required large teams and long timelines. It is built as a sequence of specialized skills that read a site generate a design direction prototype a new version migrate the full tree and deploy it to AEM Edge Delivery Services. The system is designed to take any existing website and bring it back redesigned fixed migrated or all three. Its capabilities span extraction of design evidence creation of brand aligned directions generation of high fidelity prototypes clean rebuilds of template archetypes static HTML migration block based deployment and full site rollout with quality assurance. The breadth of these capabilities positions Stardust as a tool that can compress months of work into days while maintaining a level of craft that feels human.

The first major capability is extraction. Stardust reads a live site in a real browser rather than relying on view source. This allows it to capture the true layout palette type scale motifs voice and copy. It produces structured outputs that include a product specification and a design specification. These documents serve as evidence for later steps and prevent hallucination. The extraction step can read five pages by default or the entire site when needed. It is the foundation of the redesign workflow.

The second capability is direction. Stardust can take a single sentence from a user and interpret it as a design intent. The user can ask for a warmer feel or a more editorial tone or a more modern layout. Stardust writes a target specification that explains the reasoning behind the direction. This allows teams to debate decisions and ensures that the redesign is grounded in intent rather than guesswork. The direction step keeps brand palette and fonts unless the user explicitly requests a rebrand.

The third capability is prototyping. Stardust generates a redesigned version of a page as a single HTML file. It runs a craft loop that includes critique audit viewport adaptation and motion checks. It rejects template looking output and aims for a level of polish that feels custom. The user can annotate the prototype and rerun it until it is approved. This step produces a high fidelity representation of the future site.

Once the prototype is approved Stardust can migrate the entire site to static HTML. It applies the design to every page and ensures consistency across the full tree. This is the fourth capability. The result is a clean static version of the site that reflects the approved design.

For sites that need a clean rebuild Stardust offers replica. Replica reads the live site and recreates one page per template using clean HTML and CSS. It gates each page against the original at two breakpoints using pixel diff. It allows only design changes that the user explicitly lists. Once archetypes are approved migrate clones the rest of the pages. This produces a clean foundation for future deployment.

The next capability is deployment to AEM Edge Delivery Services. Stardust converts each section of a static HTML page into an Edge Delivery block. It generates CSS for each block and converts copy into authorable content through the Document Authoring API. It rewrites internal links to the new origin previews the page publishes it and verifies that the live version matches the expected output. A page counts as deployed only after verification passes.

The final capability is rollout. Stardust runs deploy across the full migrated tree maintains a ledger of shipped and pending pages shares blocks across pages adds sitemap and redirects audits every link and performs a QA sweep of the live site. It reports findings and ensures that the site is fully functional.

These capabilities make Stardust a powerful tool for redesign and migration. However a realistic analysis of what it takes to perform a lift and shift migration of a medium complexity website reveals that the process is not trivial even with automation. A medium complexity site often includes ten to twenty templates and fifty to two hundred pages. It may include forms interactive components dynamic content and integrations with backend systems. Stardust can handle the front end transformation but backend and API functionality require separate consideration.

Stardust is not a backend migration tool. It does not rewrite server logic or rebuild APIs. It can read the rendered output of a page and reproduce the front end but any functionality that depends on server side logic must be migrated manually or through separate tools. For example a site that includes authenticated dashboards custom search endpoints or transactional workflows cannot be fully migrated by Stardust alone. The tool can replicate the front end and integrate with existing APIs but it cannot replace them. This means that enterprises must plan for backend continuity or backend modernization as a parallel track.

Token costs are another factor. A medium complexity site might require extraction of fifty pages direction prototyping of ten templates migration of fifty to two hundred pages and deployment of the same number. Each step consumes tokens. Extraction and replica are the most expensive because they involve rendering pages and producing structured outputs. Prototyping also consumes significant tokens because it generates full HTML and runs critique loops. Migration and deployment are lighter but still meaningful. A realistic estimate for a medium complexity site might range from two hundred thousand to one million tokens depending on the number of pages and the number of iterations. This cost is manageable for enterprises but must be budgeted.

The question of integrating less expensive third party agents is important. Stardust is built on Adobe AI skills and uses a specific architecture. It is possible to integrate third party agents for tasks such as content rewriting translation or QA. However the core redesign and migration steps rely on Stardust’s own skills. Third party agents can complement the workflow but cannot replace the core capabilities. For example a cheaper agent could rewrite copy or generate alternative design directions but it cannot perform extraction or block based deployment. Integration is feasible but must be done thoughtfully.

There are additional aspects worth including in an article about Stardust. One is governance. Enterprises need to understand how Stardust fits into approval workflows. Another is accessibility. Stardust can enforce accessibility standards during redesign. Another is globalization. The system can support regional variants and translations. Another is compliance. Enterprises may need to ensure that extracted content is handled securely. Another is long term maintenance. Once a site is deployed to Edge Delivery Services teams must maintain blocks and authorable content.

A complete article should also address organizational change. Stardust compresses timelines and reduces the need for large front end teams. This changes how enterprises plan redesigns and migrations. It also changes how agencies deliver work. The article should discuss how Stardust fits into the future of enterprise web operations.

In summary Stardust is a powerful system that can redesign migrate and deploy websites with a level of automation that feels transformative. It can handle front end lift and shift migrations but backend functionality requires separate tools. Token costs are manageable but must be planned. Third party agents can complement the workflow but cannot replace core capabilities. A complete analysis should include governance accessibility globalization compliance and organizational change. This gives enterprises a realistic understanding of what Stardust can do and what it means for the future of web transformation.