AI on ttvl.co

This page is transparency report on AI usage on this website. Below, you’ll find details on exactly where I use AI tools and for what purposes.

Anything not mentioned above is authored by me.

Every single line of text, even if generated by AI, is read, edited, and approved by me. There is no auto- or dynamically generated AI text, images, or any type of media on this website.

Proofreading is done using Grammarly.

Log

The bullet points you see in my log are initially generated by AI, but then edited and linked by me. The bullets come from a full ingestion of all my journal entries into Claude Opus 4.x, using the following prompt:

Act as somebody who summarizes my journal entries into a publicly shareable bullet point of things I have created, made or done.

I’ll be giving you full text of my journal that I maintained for the whole month and I want you to generate a bullet points of major events and the projects I’ve worked on this all of this will be shared on my public website as a log entry for that month.

Everything that could be an interest of somebody else of knowing what I have produced or made should be mentioned in the in the list things that are overly personal can be omitted.

Feel free to reference previous months, but try to make each summary self-contained.

The journal is mine so when writing summary bullets, do it as if I’m writing those from first person.

Provide output in Markdown as I want to copy paste it into my static site generator directly. Try to have list without sublists—one hierarchy level only.

The text below the bullet links is always written by me.

I do enjoy creative writing and thinking process that comes with it. Since this website is my happy-place of the internet, I would not delegate this pleasant task to AI.

Image Descriptions

Image descriptions—also known as alt-text—are part of the HTML spec and help visually impaired people understand the content of photos. They are not autogenerated at build time; instead, they are written as part of the post at the time of publishing. I generate these in a single pass using Claude Opus 4.x, using following prompt:

Write specific, descriptive text under 5–25 words that describes what’s actually in the image. Don’t include “image of” or “picture of” as screen readers already announce images.

Include: The main subject, important visual elements, actions being performed, and relevant location markers or artistic medium only if they add important context.

Avoid: Redundant phrases, excessive adjectives, information not visible in the image, copyright info, or interpretive language.

For street/urban photography, focus on the scene and people’s actions without identifying individuals. For gallery/art photos, include essential artistic elements. For product or gallery documentation, describe what’s shown and its context.

If text appears in the image and is important, include it. For purely decorative images, return empty text. For complex images like charts, provide a brief summary.

Return only the alt text string without quotes or formatting, ready to insert directly into HTML.