Articles

Articles and write-ups: process, decisions and practice.

  1. How to add an AI chatbot to a site: what it can do, integrations and cost

    What an AI consultant on a site can do: answer from a knowledge base, pick products, pass requests to a CRM or Telegram. And what the cost of implementation is made of.

  2. AI search and GEO: how neural networks choose sources and what depends on the site

    How AI search finds and reads pages, whom it cites in answers, and what here depends on how the site is built: indexing, HTML structure, markup, the author and dates.

  3. How to automate filling a site on Tilda with AI

    A chain from the source data to publication: how an AI agent prepares products and blog entries on Tilda, where a person’s check is needed, and what CSV import and MCP give.

  4. Claude Code, Codex or Cursor: what to choose for building sites with AI

    How Claude Code, Codex and Cursor differ in site development: where each one works, how it edits code, what a person checks, and which one to take for the task.

  5. How to create a website with ChatGPT: what it can do, the tools, and the real limits

    What ChatGPT can do when a site is created: structure, texts, page code. And where a person is needed: hosting, forms, a check on a phone, and finishing it into a working site.

  6. How to create product cards automatically with neural networks

    How to get product names, descriptions and meta tags in bulk from a table of characteristics with AI, check them for mistakes and not lose the accuracy of the catalog data.

  7. How to create a Telegram Mini App with neural networks

    A catalog, a request form or a personal account inside Telegram: what AI tools speed up when a mini app is developed, and what needs manual setup and a check.

  8. Vibe coding for business: which sites and apps can be created without a developer team

    What vibe coding is in plain words, which sites and internal tools a business can assemble with AI, and where a project has a hard time reaching launch without a developer.

  9. Automating an online store with neural networks: which processes can be handed to AI

    Which store work can be handed to neural networks: product data, the catalog, descriptions, answers to buyers. And where the check and publication should stay with a person.

  10. How website development on Tilda works: what the client does at each stage

    How creating a site on Tilda goes from the discussion to publication: seven stages, what the developer does, and which decisions by the client speed the process up.

  11. How to approve design and revisions without stretching the timeline

    How to approve the structure, the layouts and the revisions: one list of comments, one person in charge, and the difference between a mistake and a change to a decision already made about the site design.

  12. Which features your website actually needs

    Forms, quizzes, calculators, catalogs, filters, accounts and integrations: which functionality is really needed at once, and what can be added after the site launches.

  13. How to choose website references and explain your wishes to the developer

    Why phrases like “I want a modern website” help little. How to pick design examples, explain what you like, and approve layouts without endless rework.

  14. How to prepare website texts and not delay development

    Which information is needed for the home page, the services, the company and the contacts, and how to collect website texts even if a copywriter or an AI agent writes them later.

  15. How to prepare products for upload to an online store

    How to collect catalog data: names, SKUs, attributes, prices and photos. An example of an import table and a way to prepare products for an online store.

  16. What photos a website needs: covers, cards, galleries and sliders

    Why one shot rarely fits everywhere. Which photos are needed for covers, cards, galleries and sliders, and how a good source file speeds up preparing the site.

  17. How to hand photos to the website developer correctly

    How to send shots via Google Drive: size, names, folders and the selection for the cover. That way handing photos to the developer does not drag out the work on the site.

  18. How to prepare materials for a website so development goes faster

    Which source files to collect before the start, how to name the files and arrange the folders, so that preparing materials for the website speeds development up instead of delaying it.

  19. How to save on website development without harming quality

    The base price and the final estimate often do not match. I go through how ready materials and decisions made on time help lower the cost of creating a website.

  20. Exporting Tilda to your own hosting: what actually becomes independent

    What you get when a project is exported from Tilda to your own server: catalog, Feeds, forms and fonts after the move, rules for the code and how independent it really is.

  21. A quiz, a calculator or an ordinary form: different tools for different decisions

    A step-by-step survey, a cost calculation or an ordinary form: when interaction helps the client, where false precision appears and how many questions a person is willing to go through.

  22. Catalog filters and SEO pages: why convenient selection is not the same as search structure

    How selecting products by attributes affects search, when a combination of properties needs a landing page and what to set up on Tilda so you do not multiply duplicates.

  23. A catalog without a cart: why an online store is not for every business

    A catalog without a cart or a store with online payment: how to choose the format by whether the price is known in advance, what the buyer decides alone and what the manager does.

  24. A responsive site is not a shrunk desktop: what must remain on a phone

    A responsive version is not a squeezed desktop: what to do with tables and filters on a phone, why hiding blocks is risky, and how this works on Tilda.

  25. Animation on a commercial site: when an effect helps and when it gets in the way

    Animated page effects: which motion explains the interface, when parallax gets in the way of reading, and why to honor the system reduced-motion setting.

  26. A PageSpeed 100 or a fast site for people: why they are not the same thing

    Real-world site speed and a Lighthouse score are not the same: how lab tests differ from visitor data and what Core Web Vitals measure.

  27. Why AI pages on one site look different: the role of a design system

    Why AI-generated pages need a shared design system: what to lock down beyond colors, stress-testing components with long copy and what DESIGN.md gives you.

  28. Tilda Feeds 2.0: when a site needs a content model, not page copies

    Feed entries instead of duplicated pages: custom and reference fields, when a site needs a content model and what Feeds do not replace.

  29. Vibe page or standard Tilda blocks: same look, different life after launch

    A Vibe block and library blocks look the same on launch day but age differently: what changes when you edit content and why you can’t go back from Zero Block.

  30. A neural-network website: where generation ends and development begins

    How an AI-built site differs from an AI prototype after generation: where data, integrations and QA remain and what a one-line business brief cannot tell you.

  31. Online store on Tilda: what it can do, its limits and how to choose a platform

    Tilda online store: why product count matters less than data quality, 1C and MoySklad sync, wholesale checkout and what the ST340 update changed.

  32. SEO on Tilda: what you can and can’t do

    Can you rank a Tilda site? SEO settings, indexing, structure, filters, technical limits and speed.

  33. Tilda website development: pros and cons

    Tilda pros and cons in 2026: design, SEO, online stores, development cost and platform limits.

  1. AI search and GEO: how neural networks choose sources and what depends on the site

    How AI search finds and reads pages, whom it cites in answers, and what here depends on how the site is built: indexing, HTML structure, markup, the author and dates.

  2. How to create product cards automatically with neural networks

    How to get product names, descriptions and meta tags in bulk from a table of characteristics with AI, check them for mistakes and not lose the accuracy of the catalog data.

  3. Catalog filters and SEO pages: why convenient selection is not the same as search structure

    How selecting products by attributes affects search, when a combination of properties needs a landing page and what to set up on Tilda so you do not multiply duplicates.

  4. SEO on Tilda: what you can and can’t do

    Can you rank a Tilda site? SEO settings, indexing, structure, filters, technical limits and speed.

  1. How to add an AI chatbot to a site: what it can do, integrations and cost

    What an AI consultant on a site can do: answer from a knowledge base, pick products, pass requests to a CRM or Telegram. And what the cost of implementation is made of.

  2. AI search and GEO: how neural networks choose sources and what depends on the site

    How AI search finds and reads pages, whom it cites in answers, and what here depends on how the site is built: indexing, HTML structure, markup, the author and dates.

  3. How to automate filling a site on Tilda with AI

    A chain from the source data to publication: how an AI agent prepares products and blog entries on Tilda, where a person’s check is needed, and what CSV import and MCP give.

  4. Claude Code, Codex or Cursor: what to choose for building sites with AI

    How Claude Code, Codex and Cursor differ in site development: where each one works, how it edits code, what a person checks, and which one to take for the task.

  5. How to create a website with ChatGPT: what it can do, the tools, and the real limits

    What ChatGPT can do when a site is created: structure, texts, page code. And where a person is needed: hosting, forms, a check on a phone, and finishing it into a working site.

  6. How to create product cards automatically with neural networks

    How to get product names, descriptions and meta tags in bulk from a table of characteristics with AI, check them for mistakes and not lose the accuracy of the catalog data.

  7. How to create a Telegram Mini App with neural networks

    A catalog, a request form or a personal account inside Telegram: what AI tools speed up when a mini app is developed, and what needs manual setup and a check.

  8. Vibe coding for business: which sites and apps can be created without a developer team

    What vibe coding is in plain words, which sites and internal tools a business can assemble with AI, and where a project has a hard time reaching launch without a developer.

  9. Automating an online store with neural networks: which processes can be handed to AI

    Which store work can be handed to neural networks: product data, the catalog, descriptions, answers to buyers. And where the check and publication should stay with a person.

  10. How to prepare website texts and not delay development

    Which information is needed for the home page, the services, the company and the contacts, and how to collect website texts even if a copywriter or an AI agent writes them later.

  11. Why AI pages on one site look different: the role of a design system

    Why AI-generated pages need a shared design system: what to lock down beyond colors, stress-testing components with long copy and what DESIGN.md gives you.

  12. An AI agent runs the site: what Tilda MCP changes and where humans must stay in control

    Tilda’s MCP protocol gives an AI agent access to your project: what it can do today, why to limit keys by module and lifetime, and which calls stay human.

  13. Vibe page or standard Tilda blocks: same look, different life after launch

    A Vibe block and library blocks look the same on launch day but age differently: what changes when you edit content and why you can’t go back from Zero Block.

  14. A neural-network website: where generation ends and development begins

    How an AI-built site differs from an AI prototype after generation: where data, integrations and QA remain and what a one-line business brief cannot tell you.

Contact

Let's discuss
your project

  1. 01

    You briefly describe the task

  2. 02

    I reply personally and clarify the details

  3. 03

    I propose a structure and an estimate

Project request

Marina Stepnaya