Prompt:
writing blog posts folder - make a new dir called ask-ai-to-do and make a 9:16 ratio portrait image in the same style as our earlier "is this a good use of the tech...brain?" one (also in blog post folder good-use). this new image has "Ask AI How To Do Something" but the How is strikethrough. leonfurze.com lower left smaller font.
What I was trying to achieve: I wanted to create a reusable skill for generating styled portrait graphics with text overlays.
What happened: I designed a 9:16 portrait image with strikethrough text styling and packaged it as a global Claude Code skill so I could regenerate similar graphics repeatably.
Reflection: this made a nice, clean image... eventually. Was easier to just do it in Canva myself but I suppose the point is to persist and eventually develop a strong enough skill, even if it initially takes longer.
Prompt:
go to curric-leaders workshops and make high quality transcripts of any of the term 3 sessions that havent been transcribed yet
What I was trying to achieve: I wanted to create high-quality transcripts of the remaining Term 3 curriculum leader workshop sessions.
What happened: I generated transcripts for the workshop sessions and converted them to PDF and Word formats, including formatting fixes for tables and table-of-contents. The corpus now has properly formatted documents ready for reference.
Reflection: This consistently works well, using the Whisper Large model (not Whisper-MLX)
Prompt:
theres a new video doing the rounds of a robot sprinter beating usain blots speed - robots on a track and field track. i want you to find the video and put a well selected 5second clip of it into a GIF. lets make a tool of this in projects/tools/video-to-gif and wrap it in a skill once it works.
What I was trying to achieve: I wanted to find the viral robot sprinter video and extract a 5-second GIF clip, then package it as a reusable tool.
What happened: I sourced the robot race video, built a video-to-GIF extraction tool with smart frame selection and light/standard quality options, documented the process including off-by-one gotchas, and packaged it as a Claude Code skill. The output is a clean 5-second clip available in multiple sizes.
Reflection: worked well, after a few failed attempts getting the frame splitting correct. tested on two other gifs.
Prompt:
figure out a way to take a screenshot of the video at around the 40 minute mark where the two questions are up on the slide, and then neatly blur everyone except for me, Leon Furze. Capture it at a moment where I'm not eyes closed or mouth wide open or anythig ridiculous.
What I was trying to achieve: I wanted to capture a screenshot from my Zoom workshop video and blur all faces except mine for a public-shareable image.
What happened: I built an end-to-end skill using face detection to automatically identify faces, then selectively blur everyone except me while preserving the video moment. The tool extracts frames, runs detection, applies targeted blur, and is now packaged as a reusable Claude Code skill with Zoom layout reference docs.
Reflection: did exactly what it needed to do, including choosing a frame where I has my mouth closed
Prompt:
are you able to add a block to my teachable practicalaistrategies.com course homepage above the "new course VET" one that has: "Waitlists now open for Practical AI 6-Week Intensives" and then a short blurb and then two buttons that go to leonfurze.com/curric-leaders and wellbeing-leaders - youll need to use chrome like the other day when you nade a bunch of changes and im thinking it will need to be html or whatever, and youll QA it
What I was trying to achieve: I wanted to add a waitlist announcement block to my course homepage with dual CTA buttons and full QA verification.
What happened: I built an announcement block for the course homepage with buttons linking to my curriculum and wellbeing leadership programs. I QA'd it in preview mode, verified the layout and buttons, and published it live to the platform.
Reflection: increasingly impressed by the claude in chrome capabilities
Comments
No comments on this week yet.