Research
Documents, interviews, project memory, real photographs and public sources.
WORKING · WEB · BOOK · VISUAL · VIDEO
AI accelerates production.
Results pass verification.
In digital production neural networks are not a single-generation button but a set of specialised tools: for research, structure, copy, code, images, video preparation and quality control.
Approach
One project, several model types and checks
A text model helps assemble structure and drafts, a coding agent implements and tests the site, an image generator creates individual visual assets, and the video system prepares files, covers and metadata. Raw facts, material rights and final publish are each controlled separately.
Production workflow
From sources to published output
Copy, code and visuals
Structure, editing, programming, images and production variants are built with different tools.
Verification
Facts, links, responsiveness, visual screens, file formats and task compliance.
Physical or digital output
Website, printed book, graphic asset or prepared and published video.
Confirmed results
Four production tracks
Websites and interfaces
Research, information architecture, copy, HTML/CSS/JS, SEO, responsive and interactive checks.
Printed book
AI was used in research and the production process of the book "Build Your Own Assistant," which reached print and sales.
Images
Generation and processing of individual illustrations, covers and visual briefs, keeping real source files where evidence matters.
Video materials
Script, metadata, cover, file preparation and subsequent delivery to VK Video and YouTube.
Site control
Static guard, link and interactivity checks, desktop/mobile screenshots and overflow control.
Publishing system
Ready materials pass to the agent and MCP connectors without exposing tokens in the cloud environment.
Separation of responsibility
Content creation and publishing are different systems
Production owns content, visuals, code and the result file.
Quality gate checks facts, rights, technical quality and behaviour on devices.
Content operations adapts finished material for channels and returns publish links.
Human approves new sensitive content and decisions outside agreed scenarios.
Case boundaries
Real results are shown, not model authorship
AI helps research faster, assemble options and do technical work, but it does not verify facts on its own. Real photographs and approved materials are not replaced by decorative generation when they serve as proof of the project.