Improve Your AI Skills with Open Interpreter

Open Interpreter turns plain-language prompts into real, local actions—writing and running code, scraping the web, transforming media files, and editing images—so practice becomes less about “getting the answer” and more about directing an AI to complete tasks end to end. The workflow matters because it forces tighter instruction-writing: the model must navigate a working directory, create files, handle errors, and produce concrete outputs (Python scripts, text files, PDFs, sped-up videos, and edited images) on the user’s own machine.

The session starts with four self-made challenges designed to sharpen instruction skills. Challenge one focuses on file and text manipulation plus Python execution: Open Interpreter lists the current working directory, reads a text file (gtg do text), writes its contents into a new Python file (AGI dop/ai.py), then creates another script (count.py) that prints numbers up to 100. The result is verified by running the generated Python file in the terminal and confirming the countdown output.

Challenge two shifts to web work and document generation. Using the Verge (theverge.com) as the target site, Open Interpreter is instructed to scrape three H2 headlines, include a user agent, save the headlines and URLs to text files, then follow up by scraping the full content for the first headline’s URL. After extracting the article text into article_1.txt, it generates a summary and converts that summary into a PDF using a PDF library. When an error appears while trying to extract article URLs, the workflow includes a manual fix step (adjusting a specific line) and then rerunning the full script, producing the expected text and PDF outputs.

Challenge three demonstrates local video editing. Open Interpreter locates an MP4 file in the working directory and attempts to speed it up two times using a MoviePy-style approach. The run produces a shorter video, but the audio is lost—an outcome the creator treats as acceptable for the purposes of the challenge because the core timing reduction succeeds.

Challenge four covers image editing. Open Interpreter finds a PNG file, crops it to 50% size, then converts the cropped image to black and white, saving and reopening the result to confirm the transformation.

Setup is presented as straightforward: install via pip (pip install open interpreter), then run with a local model using LM Studio (interpreter --local). A YAML config is adjusted to disable safe mode (set to off) and to run offline (offline: true), trading some online features for a fully local workflow. The overall takeaway is practical: using Open Interpreter for task-based challenges helps people learn how to write prompts that reliably drive code execution and real file transformations, especially when working with a smaller local model such as “Mistral 7B mod” in LM Studio. The session also includes a sponsor plug for a HubSpot free ebook about using ChatGPT to streamline daily work.