# Bolek – full text for models (English) > All Bolek guides in one file. Index with short descriptions: https://bolekrobot.pl/en/llms.txt # How to write a CLAUDE.md file that Claude Code actually follows (2026) URL: https://bolekrobot.pl/en/guides/claude-md-file-claude-code-follows Type: Guide Date: 2026-11-07 Updated: 2026-11-07 Author: Maciej Głuszak **In short:** CLAUDE.md is the rules file Claude Code reads at the start of every conversation. The fastest way to write it is to let the assistant interview you and build a short file from your answers. Paste into Claude Code: ``` Interview me – at most 7 questions, one at a time – and based on my answers write a CLAUDE.md in this folder: what language to reply in, what not to do without asking, how to name files and how to check the result. At most 40 lines, one sentence per rule. ``` Updated: 2026-11-07 · tested on: Claude Code 2.1, Windows 11 Claude Code does not remember the previous conversation. It remembers what you write in **CLAUDE.md** – the assistant loads it at the start of every conversation and treats it as a manual for working with you. Bolek, our AI assistant for Windows, has stood on such a file since day one, so everything below comes from daily use. ## Where the file lives and which one wins | file | where | what for | |---|---|---| | CLAUDE.md in the project folder | the folder where you run claude | rules for one project: names, folder layout, how to check results | | CLAUDE.md in your profile | the .claude folder in your user profile | rules for all projects at once: language, tone, what not to do without asking | The assistant reads both. When rules conflict, the one closer to the work wins, i.e. the project file. Write language and tone once, in the profile. ## Step by step 1. Open a terminal in the project folder and run **claude**. 2. Paste the prompt from the box at the top of this page and answer the assistant's questions. 3. Read the finished file and strike every rule you do not understand or disagree with. 4. Start a new conversation and check that the assistant follows the rules – for example ask for a new file and look at its name. 5. When you correct the assistant a second time on the same thing, add it to the file in one line. The built-in **/init** command also creates such a file, but it mostly describes the code of the project. If you use it, review the result against the skeleton below. ## The skeleton we start from Four sections are enough to start: - **Language and tone** – "reply briefly, no preamble". - **What not to do without asking** – sending email, publishing, deleting files, payments. - **Names and order** – "file and folder names in plain ASCII", "save results in the results folder". - **How to know it is finished** – "before you say it is done, open the file and check the result". ## Five traps we hit - **A file that is too long.** The file is read at the start of every conversation and costs working memory. At several hundred lines the rules start to get lost. Move details to separate files and point to them with one line. - **Rules without a reason.** "Do not send email" works worse than "do not send email without my confirmation – a draft once went to a client". The reason helps the assistant decide cases you did not foresee. - **Vague rules.** "Write nicely" means nothing. "Sentences of at most two lines, no exclamation marks" can be checked. - **Conflicting rules.** Two rules that disagree end with the assistant choosing at random. Once in a while read the file top to bottom. - **Passwords in the file.** CLAUDE.md is plain text that often ends up in a repository. Keep passwords and keys in a separate private file outside the project. ## FAQ **Can I check which rules the assistant loaded?** Yes. The **/memory** command in a conversation shows the rules files in use and lets you open them for editing. **What if the assistant breaks a rule?** Most often the rule is too vague or got lost in a long file. Shorten it to one concrete sentence and move it higher. # How to run Claude Code on Windows without WSL (2026) URL: https://bolekrobot.pl/en/guides/claude-code-on-windows-without-wsl Type: Guide Date: 2026-11-07 Updated: 2026-11-07 Author: Maciej Głuszak **In short:** Claude Code runs natively on Windows: install Git for Windows, run the official installer in PowerShell, then start it in your project folder. You do not need WSL. Paste into Claude Code: ``` Look at this folder and tell me what is in it. Then create a CLAUDE.md with three rules: reply in short sentences, never send anything outside this computer without asking, and name files without spaces or non-ASCII characters. ``` Updated: 2026-11-07 · tested on: Claude Code 2.1, Windows 11 Claude Code is an AI assistant from Anthropic that works in your terminal and does the work itself: it reads and writes files, runs commands and keeps track of the steps. Bolek, our AI assistant for Windows, is built on it, so everything below was tested in daily use on a plain Windows machine, not in a Linux layer. ## What you need - a computer with Windows 10 or 11, - a paid Claude plan (Pro or Max) or an account billed by usage – the free plan does not include Claude Code, - Git for Windows – Claude Code uses the shell that ships with it to run commands. ## Step by step 1. Install **Git for Windows** with the default settings. 2. Install **Claude Code** with the official installer from Anthropic – one command pasted into PowerShell. 3. Open a terminal in the folder the assistant should work in and type **claude**. 4. Log in with your Claude account in the browser window that opens. 5. Type your first instruction, for example: "look at this folder and tell me what is in it". 6. Create a **CLAUDE.md** file in that folder with your working rules – the assistant reads it at the start of every conversation. If `claude` is "not recognized" after installing, close the terminal and open a new one so the PATH refreshes, then check with `where.exe claude`. ## The rules file: the most important five minutes Claude Code does not remember the previous conversation. It remembers what you write in the rules file. For us it started with a few lines and is now the foundation of everything. Three things are enough at the start: - **what language and tone to use** – "reply briefly, no preamble", - **what not to do without asking** – for example sending anything outside your computer, - **how to name files** – see the first trap below. ## Five traps we hit on Windows - **Non-ASCII characters in file and folder names.** Text in any language works, but names with accented letters can break inside commands. A rule "file names in plain ASCII, no spaces" saves hours. - **Garbled characters in a saved file.** Older Windows PowerShell saves files in an encoding that loses non-English letters by default. Ask the assistant to always save files as UTF-8 and put that in the rules file. - **Two shells, two command languages.** On Windows the assistant has both PowerShell and the Git shell. A command written for one fails in the other. When something "fails for no reason", first ask which shell ran it. - **Paths with spaces.** "My Documents" in a command without quotes is two separate words. Keep projects in folders without spaces. - **One long conversation about everything.** The longer the conversation, the more the assistant has to hold in its head and the worse it does. Our rule: one task, one conversation. Whatever must survive goes into the rules file or a note. ## What next Once the first prompts work, teach the assistant repeatable tasks with skills, and see how to write a rules file it actually follows in our guide on CLAUDE.md. # How to make an AI video with voiceover and burned-in subtitles on Windows (2026) URL: https://bolekrobot.pl/en/guides/ai-video-voiceover-subtitles-ffmpeg-windows Type: Guide Date: 2026-11-07 Updated: 2026-11-07 Author: Maciej Głuszak **In short:** A voiceover video is three layers: text, recorded voice and subtitles burned into the picture. Claude Code can run the whole pipeline while you only approve the script and the subtitle text. Paste into Claude Code: ``` Make a vertical 9:16 video with a voiceover. First write a 60-second script and show it to me for approval. After I approve, generate the voice, transcribe it with Whisper, show me the subtitle text for corrections, and after my ok burn the subtitles into the video with FFmpeg. ``` Updated: 2026-11-07 · tested on: Claude Code 2.1, Windows 11, FFmpeg 7 A video with a narrator is three layers: **text**, **voice** and **subtitles**. The picture can be a still, a screen recording or a clip from a generator. Below is the way we do it ourselves in Bolek, an AI assistant for Windows built on Claude Code. Unlike the hosted avatar tools that top the search results, only the voice is a paid web service here – assembly and subtitles run on your own computer. ## What you need - an account with a voice service, for example ElevenLabs – it has many languages and can clone your own voice, - FFmpeg – free, it joins picture, sound and subtitles, - the Whisper speech recognition model – runs locally, - Claude Code to chain these steps into one pipeline. ## Step by step 1. **Script.** Ask for a text of a specific length. About 150 words is roughly one minute of speech. Read it yourself – the voice repeats every mistake. 2. **Voice.** Paste the text into the voice service and pick a voice. Short sentences and numbers written out as words help, because digits are sometimes read in the wrong language. 3. **Picture.** Prepare a background or clips as long as the recording. Build a cheap preview from still frames first, and only turn the approved preview into expensive clips. 4. **Subtitles.** Whisper turns the recording into text with timing in an SRT file. It gets most words right but mangles names – always read the subtitle text before burning it in. 5. **Assembly.** FFmpeg joins everything. Burn subtitles with a command like this: `ffmpeg -i video.mp4 -vf "subtitles=subs.srt" -c:a copy result.mp4`. 6. **Check.** Watch the finished video on a phone. Make sure subtitles fit the frame and do not disappear too quickly. ## Traps we hit - **Lines that are too long.** Two lines of about 35 characters read well on a phone; longer ones are cut off at the sides. - **Dangling short words.** A one-letter word at the end of a line looks sloppy – break the line before it. - **Non-ASCII characters in paths.** Name input files in plain ASCII, otherwise FFmpeg can refuse to work. - **Full path in the subtitles filter.** The colon in a Windows path breaks the filter. Keep the SRT file next to the video and pass only its name. ## What it costs The voice and any generated clips are pay-per-use external services. Whisper and FFmpeg are free on your own computer.