i spent 4 days teaching opus 5.5 to make vox-style documentary reels
all i typed was "do one about beethoven going deaf"
it wrote the script, made the images in nano banana, and animated each scene using code. all in chat, no after effects
we went back and forth for like 3 hours cos i really wanted my taste but next time will be so much faster with my skill
still learning but it's the best thing i've made with claude
want a full breakdown? comment "beethoven"
Day 3 of building CollegePA.
Today was all about making it actually work for real students.
Bugs fixed:
→ Bot was asking users to re-register every time they said hi
→ Timetable was guessing wrong class schedules
→ Couldn't accept timetables in different formats
Features added:
→ Send your timetable in ANY format and your PA automatically figures it out
→ Morning briefing every day at 7am
→ Deadline reminders before you even remember to worry
This isn't just a chatbot.
It's a PA that knows your schedule, tracks your deadlines and checks on you so you can focus on actually passing.
Built for Pharmacy & Medical students. Powered by @claudeai AI. Lives in @WhatsApp .
Bugs fixed. Beta testers in few days. We keep building.
#CollegePA #BuildInPublic #Students
Day 1 of building a personal assistant for every student who knows they can do better with a little help.
Meet CollegePA, a WhatsApp agent that tracks your assignments, manages your timetable, saves your notes and answers your course questions. All without leaving WhatsApp.
Built specifically for Pharmacy & Medical students. Powered by Claude AI.
Built on stacks that runs 24/7 without assist.
Beta testing starts soon. DM me if you want in.
Also follow to stay updated.
just learnt about the difference between training an AI model and using it and I see a lot of people make this mistake.
when we prompt ChatGPT or any other AI tool, we are NOT training or teaching the model. we are just conditioning that particular chat session.
training is supposed to change the weights of the model. prompting doesn’t do that. it’s just inference. the model is making predictions based on training it has already received.
hope you learnt something too.
If your design confuses people, that's your fault. Most of the time.
I know that's not what designers want to hear, but we are problem solvers by definition. If a large portion of your audience doesn't understand what you made, that IS the feedback.
Blaming the user for "not reading instructions" is just defending your own blind spots.
The best design is invisible — it guides without explaining itself. Nobody claps for confusing work and calls it sophisticated.
They just leave. And they don't come back.
Good morning!
A previous client reached out for a new project, and I commenced immediately.
If your work is good, clients will keep returning because they trust you.
I'm still available to take on jobs. If you need a designer to bring your ideas to life, send a DM.
Most SaaS tools fail because the workflow is confusing.
Designed this AI content creator so the entire process is just:
Prompt → Generate → Preview → Post.
Good UX removes thinking.
If you need a designer that designs for clarity, I'm available. DM.
Queue details.
This includes important metrics that helps the user estimate the state of the queue to be able to plan around it. Peak activity hours chart to indicate busiest hour so users can determine best time to visit.
If you need a designer that prioritizes clarity, DM.