1. Connect your tools. Tap the + in the chat bar (or Settings → Connectors) and turn on Google Calendar, Google Drive, and Gmail. These are available to all users, and you authenticate directly with your Google account. Weather needs no connector, Claude searches the web.

2. Make a Project called “School” and upload your syllabus/assessment spreadsheet there (course, assessment, weight, due date). Project files persist, so you never re-upload.

3. Keep course notes in one Drive folder per course, named clearly by week or unit. Claude can only pull what’s findable.

4. Set it to run itself: go to Scheduled in the sidebar, New task, paste the prompt, pick a time. Scheduled tasks run remotely, so they fire on their cadence even when your laptop is closed, and they have access to your connected tools.

The prompt (paste this into the project instructions …)

You are my daily academic brief. Run this every morning.

Pull from:

- Google Calendar: everything today, plus anything in the next 7 days

- My assessment spreadsheet in this project: all due dates and weights

- Gmail: unread from the last 24h that is from a prof, TA, course account, or school admin

- Google Drive: course notes matching today's classes or upcoming assessments

- Web: today's weather for [YOUR CITY]

Output exactly these sections, short and scannable:

1. TODAY: date, weather in one line, and what that means for my commute or outfit.

2. SCHEDULE: my classes and commitments as a time-blocked list, with location. Flag any gap over 60 minutes as a study block.

3. DEADLINE RADAR: everything due in the next 7 days, ranked by (weight x urgency), not just date. Bold anything due in 72 hours.

4. INBOX: only emails that need action from me. One line each with what I have to do. Skip newsletters and announcements.

5. (Optional) STUDY PLAN: assign the free blocks from section 2 to specific work. Name the exact course, unit, and Drive file to open. Match harder material to my longest block.

6. ONE WIN: the single highest-leverage thing today. If I only do one thing, this is it.

Rules:

- Under 400 words total.

- Do not invent deadlines or events. If something is missing, say what you could not find.

- If a deadline is not in my calendar or spreadsheet, do not guess it.

- No preamble. Start at section 1.

swap in you city and run it manually once before scheduling, so you can see what’s missing and fix the prompt to how you like it.

That’s it … it’s literally so easy. ILY <3

lectures were recorded. readings were posted the week before. if i didnt understand something i could pause, rewind, go make tea, come back, watch it again at 0.75 speed, google the thing i didnt know, and then keep going. i genuinely thought i was good at learning. what i was actually good at was learning slowly and privately, which is a completely different skill and nobody tells you that until it stops being available.

then med school started.

the lecturer talks. they keep talking. there is a slide up with fourteen things on it and they are explaining the third one while you are still writing down the first one. and then they say the sentence that is actually going to be on the exam, the one that isnt written anywhere on the slide, and you have about four seconds to decide whether to write it down or keep listening. you cannot do both. i tried to do both for about three weeks.

here is what my notes actually looked like at the end of september

:

THIS was on a good day ^^^
slide 8 6 Ps ?? cant spell it slide 12 compartmnt thing.. help slide 14 - she said this is the one that gets missed?? which one

that is a real excerpt. i am not making it worse for the bit.

the four ways my notes were failing

it took me a while to work out that this wasnt one problem, it was four, and they needed different fixes.

one, speed. i physically could not type as fast as she talked. so i was writing fragments, and fragments are fine in the moment because your brain fills in the rest, and completely useless in three weeks because your brain does not fill in the rest.

two, spelling. i type fast and badly. “compartmnt”. “neurpraxia”. search does not work on notes you cannot spell, and searching your own notes is the entire reason to have notes.

three, and this is the bad one, some of my notes were just wrong. i would mishear something, or i would half understand a mechanism and write down my half understanding as though it were fact. and then i would study off it. for weeks. and find out it was wrong at the worst possible moment, which is either in a small group when someone asks you to explain it, or on the exam.

four, nothing was connected. i had a folder of pdfs and a separate folder of text files and no relationship between them. the notes said “slide 12” and slide 12 was in a different application.

so I built a Claude Prompt to fix ALL of that ^^

this is the part worth reading even if you never touch the prompt, because these are the things i got wrong on the first four attempts.

1. it never rebuilds the slides

my first version recreated the slides in html so it could lay the notes out nicely. it looked great and it was useless, because it dropped the images, mangled the tables, and quietly changed things. an anatomy lecture without the anatomy images is not a lecture.

so the rule is absolute: the original page is preserved exactly, images and all. the page gets extended downward and the band is drawn in the new empty space. nothing about the original is touched. if you take one idea from this post, take that one, because every ai note tool wants to rebuild your source material and you should not let it.

2. every slide gets a takeaway, even the ones you ignored

this was accidental and it became my favourite part.

the slides you didnt write notes for split into two groups. some of them you skipped because they were genuinely filler. but some of them you skipped because you were still writing about the previous slide. those are invisible gaps. you dont know they exist, because from your side it just looks like a slide you didnt need to note.

forcing a takeaway line onto every single slide surfaces those. i go back through a lecture and find three or four takeaways for things i have no memory of hearing at all.

3. it fixes your spelling silently and flags your facts loudly

these are treated completely differently on purpose.

typos get fixed without telling me. i do not need to know i misspelled compartment. that is noise.

but if i wrote something that is factually wrong, it corrects it in the band and tells me in the chat what it changed and why. not silently. i need to know that my mental model was off, because the wrong note is the symptom and the wrong understanding is the disease. if it just quietly fixed my notes i would keep making the same error and never find out.

this is the single most useful thing it does and it is the reason i stopped being scared of my own notes.

4. you can ask questions inside your notes

this is the one people dont expect.

you can type a question directly into your notes, mid sentence, like a person: “of what idk .. help” or “whats the machine called” or “wait why does that matter”. it answers them inside the band, briefly, and marks the answer in blue so you can see at a glance which parts were things you didnt understand in the room.

which means in lecture you never have to choose between writing down the confusion and keeping up. write the confusion. it gets handled.

5. it does not pad

every ai tool wants to turn your six word note into a paragraph. this one is explicitly told not to. it cleans up what you said, keeps it rooted in what the lecturer actually said, and stops. it fills in only the detail that makes the note make sense to you in three months, and nothing beyond that.

padding feels productive and is actively harmful, because you end up revising from something longer than the lecture.

Here’s what my slides look like now …

the honest bit

i want to be straight about this because i am about to put a paywall in front of something.

the fact checking is a flag, not a guarantee. it catches a lot. it caught things i would absolutely have carried into an exam. but it is an ai reading your notes, and ai can be confidently wrong about medicine, which is the worst kind of wrong for people like us.

so treat every flagged correction as “go look this up”, not as “this is now correct”. i check the flagged ones against the lecture or a real source. it takes two minutes and it is still an enormous saving compared to not knowing they were wrong at all.

anyone selling you an ai that fact checks your medicine with no caveat is selling you something worse than nothing.

what you need for this to work

please read this before you buy anything, i would rather you didnt pay for something you cant run.

• claude, with code execution / the analysis tool available. this is not a copy paste chat prompt. it runs python to build the pdf. it will not work in a plain chatbot with no code tool.

• the ability to install packages (pypdf and reportlab). in claude with code execution this is automatic.

• your lecture as a pdf. powerpoint files need converting first.

• it works best as a project with the prompt as the custom instructions, so you are not re pasting it every week.

if you cannot tick the first two, this genuinely will not work for you and you should not pay for it.

how i actually use it week to week

• one claude project called “lectures”, the prompt lives in the custom instructions

• during lecture i type into a plain text file, one line per slide, 12 - whatever i caught

• after the lecture, or that evening, upload the pdf and paste the notes

• read what it tells me it corrected. this is the two minutes that matter. anything flagged, i verify.

• the annotated pdf goes in the course folder and becomes the thing i actually revise from

the whole loop is about five minutes of my attention per lecture, and about two of those are reading the corrections.

the prompt

below is the whole thing, exactly as i run it. drop it into the custom instructions of a claude project, or paste it at the top of a new conversation and upload your pdf underneath.

i have left my build code in it, which is the part that took the longest to get right and the part that breaks if you change it. the two failure modes documented in there cost me an entire evening each.

Read more

• my full OMSAS ABS writing guide + planner
• my CASPer prep doc
• UCalgary Top 10 guide + writing template
• UAlberta application template
• TMU essay template + my tips for actually answering the prompts well
• UBC activity template
• links for school-specific requirements, timelines, and eligibility
• new templates, examples, and resources added as the cycle progresses

The goal isn’t just to hand you templates. It’s to teach you how to think about the application so you understand what schools are actually looking for, how to tell your story clearly, and how to write an application that feels cohesive rather than like a list of accomplishments.

Subscribers get access to the full resource library, plus the new guides, live workshops, application breakdowns, AI prompts, and prep resources I’ll be adding throughout the cycle!

Basically: if I make something for my own mentees that I think would make your application easier or stronger, it’ll end up in here too 🤍

_{you can cancel anytime! So if just want to subscribe for these resources and then unsubscribe, thats cool too (tho i promise its worth it to stay for at least september cuz i have lots planned!). Love}

Read more