This kind of AI-assisted software development is sometimes called \u201cvibe coding\u201d. The term was coined by Andrej Karpathy, a co-founder of OpenAI and former AI leader at Tesla.<\/p>\n\n\n\n
However, the strictest definitions of vibe coding state that the developer neither reviews nor manually edits the code, relying solely on AI tools to evaluate and improve it.<\/p>\n\n\n\n
Under this strict definition, \u015eenol said his experiment was not pure vibe-coding. \u201cTo be able to make this, you still need to have a general idea of what a computer is and what is going on,\u201d he said.<\/p>\n\n\n\n
\u201cAI is not that good. But still, I don\u2019t know systems development and I did not write any code at all. It was mostly breaking things up into smaller steps for Claude.\u201d<\/p>\n\n\n\n
AI needs a pep-talk<\/h2>\n\n\n\n![\"\"](\"https:\/\/mybroadband.co.za\/news\/wp-content\/uploads\/2026\/01\/VibeOS-desktop.jpg\")
The VibeOS desktop<\/figcaption><\/figure>\n\n\n\n\u015eenol said Claude generally falls into certain patterns of failure with this level of project difficulty, and you need to notice and catch them.<\/p>\n\n\n\n
\u201cThe biggest problem with a project like this is managing the context window. AI will forget about the scope of the project, or what the project even is about, occasionally,\u201d he said.<\/p>\n\n\n\n
\u201cMy initial idea was updating claude.md every run, but Claude turned that into a formatted session log inside claude.md, which is why the session logs exist. They served well.\u201d<\/p>\n\n\n\n
Claude.md is a text file that the large language model automatically pulls into context when starting a conversation.<\/p>\n\n\n\n
Although Anthropic states that there is no required format for the file, it suggests using it to document core files and utility functions, code style guidelines, and testing instructions.<\/p>\n\n\n\n
\u015eenol said the inadvertently created session logs were especially useful to motivate Claude to tackle a scary feature. \u201cIt could read a few of the session logs, get impressed by the project and work harder.\u201d<\/p>\n\n\n\n
\u015eenol said that motivating the large language model to do the work was a real thing and that everyone should do it.<\/p>\n\n\n\n
\u201cAI agents work better in already created, pretty large codebases; otherwise, they fall back into \u2018demo\u2019 mode where they make garbage things from the first episode of a YouTube tutorial series,\u201d he said.<\/p>\n\n\n\n
\u201cYou need to impress it to get better results. This is less necessary for building iOS apps and more necessary for VibeOS. It\u2019s also not a bulletproof solution, but it worked well.\u201d<\/p>\n\n\n\n
He said the main motivational method was being nice to the assistant and having it read a bunch of irrelevant code or the session logs at the start of the session to understand how cool the project is.<\/p>\n\n\n\n
\u201cThen you tell it, \u2018See, you built all this, you can make a browser, an OS is harder than a browser,\u2019\u201d he explained.<\/p>\n\n\n\n
\u201cThe irony is when you vibe-code it, the browser is harder than the OS because you need to parse one million things, and that\u2019s why the VibeOS browser sucks, but the technique is useful.\u201d<\/p>\n\n\n\n
Our experience with the browser confirmed that it does, indeed, suck. Although it could download and display plaintext webpages, any modern website would cause the whole OS to crash.<\/p>\n\n\n\n
Using Gemini to explain bugs to Claude<\/h2>\n\n\n\n![\"\"](\"https:\/\/mybroadband.co.za\/news\/wp-content\/uploads\/2026\/01\/VibeOS-browser-and-sysmon.jpg\")
VibeOS browser and system monitor<\/figcaption><\/figure>\n\n\n\nAnother technique \u015eenol used was to use Google\u2019s Gemini large language model to help Claude understand and fix bugs.<\/p>\n\n\n\n
\u201cWhen debugging, looking at multiple files that are potentially bigger than Claude\u2019s context window is necessary,\u201d he explained.<\/p>\n\n\n\n
\u201cFor this, I repomixed a bunch of files in subfolders and asked Gemini to find out the reason why the code was not working, and explain it in English for Claude to fix.\u201d<\/p>\n\n\n\n
\u015eenol said Claude generally falls into certain patterns of failure with this level of project difficulty, and you need to notice and catch them.<\/p>\n\n\n\n
\u201cThe biggest problem with a project like this is managing the context window. AI will forget about the scope of the project, or what the project even is about, occasionally,\u201d he said.<\/p>\n\n\n\n
\u201cMy initial idea was updating claude.md every run, but Claude turned that into a formatted session log inside claude.md, which is why the session logs exist. They served well.\u201d<\/p>\n\n\n\n
Claude.md is a text file that the large language model automatically pulls into context when starting a conversation.<\/p>\n\n\n\n
Although Anthropic states that there is no required format for the file, it suggests using it to document core files and utility functions, code style guidelines, and testing instructions.<\/p>\n\n\n\n
\u015eenol said the inadvertently created session logs were especially useful to motivate Claude to tackle a scary feature. \u201cIt could read a few of the session logs, get impressed by the project and work harder.\u201d<\/p>\n\n\n\n
\u015eenol said that motivating the large language model to do the work was a real thing and that everyone should do it.<\/p>\n\n\n\n
\u201cAI agents work better in already created, pretty large codebases; otherwise, they fall back into \u2018demo\u2019 mode where they make garbage things from the first episode of a YouTube tutorial series,\u201d he said.<\/p>\n\n\n\n
\u201cYou need to impress it to get better results. This is less necessary for building iOS apps and more necessary for VibeOS. It\u2019s also not a bulletproof solution, but it worked well.\u201d<\/p>\n\n\n\n
He said the main motivational method was being nice to the assistant and having it read a bunch of irrelevant code or the session logs at the start of the session to understand how cool the project is.<\/p>\n\n\n\n
\u201cThen you tell it, \u2018See, you built all this, you can make a browser, an OS is harder than a browser,\u2019\u201d he explained.<\/p>\n\n\n\n
\u201cThe irony is when you vibe-code it, the browser is harder than the OS because you need to parse one million things, and that\u2019s why the VibeOS browser sucks, but the technique is useful.\u201d<\/p>\n\n\n\n
Our experience with the browser confirmed that it does, indeed, suck. Although it could download and display plaintext webpages, any modern website would cause the whole OS to crash.<\/p>\n\n\n\n
Using Gemini to explain bugs to Claude<\/h2>\n\n\n\nVibeOS browser and system monitor<\/figcaption><\/figure>\n\n\n\nAnother technique \u015eenol used was to use Google\u2019s Gemini large language model to help Claude understand and fix bugs.<\/p>\n\n\n\n
\u201cWhen debugging, looking at multiple files that are potentially bigger than Claude\u2019s context window is necessary,\u201d he explained.<\/p>\n\n\n\n
\u201cFor this, I repomixed a bunch of files in subfolders and asked Gemini to find out the reason why the code was not working, and explain it in English for Claude to fix.\u201d<\/p>\n\n\n\n
Another technique \u015eenol used was to use Google\u2019s Gemini large language model to help Claude understand and fix bugs.<\/p>\n\n\n\n
\u201cWhen debugging, looking at multiple files that are potentially bigger than Claude\u2019s context window is necessary,\u201d he explained.<\/p>\n\n\n\n
\u201cFor this, I repomixed a bunch of files in subfolders and asked Gemini to find out the reason why the code was not working, and explain it in English for Claude to fix.\u201d<\/p>\n\n\n\n
![\"\"](\"https:\/\/mybroadband.co.za\/news\/wp-content\/uploads\/2026\/01\/VibeOS-DOOM.jpg\")
![\"\"](\"https:\/\/mybroadband.co.za\/news\/wp-content\/uploads\/2026\/01\/VibeOS-terminal.jpg\")