Lately Iāve realised how much I use and depend on AI tools for basic daily tasks. Not so long ago, I could read articles, study material or novels for hours on end and not lose focus, I could sit down, manually scribble my ideas with pen and paper and write creative content without second guessing every single word I was writing. I feel like the hobbies I enjoyed so much had become a task that needed to be perfected, polished, and needed an AI agentās approval at every step.
Itās ironic to think, we hit approve every time an AI agent is getting a task done for us, whereas in reality, we have stopped doing anything without its approval or suggestions. Weāve stopped trusting that our brain can learn, store and retrieve massive amounts of information and knowledge. Weāve stopped believing that we can navigate simple mundane scenarios or solve problems without AI.
āWhat you donāt use, you loseā, a phrase thatās been haunting me lately. How often have you caught yourself coming up with a brilliant idea but you immediately had to run it past an LLM to build the idea further or to see if it made sense, that abrupt stop after we get an amazing idea is preventing us from learning, creating and evolving further. Iāve been there too and found myself losing my flair and ability to think creatively. I wanted to change that.
Iām an architect turned data scientist but Iām also a creative individual who loves a good challenge. I absolutely love writing, programming and building things, so for starters I built myself a Python script that generates writing prompts for my 30-day writing challenge.
Hereās how it works:
A list of 60 topics Iām learning as a data scientist, each tagged with a subject and a type (explain, build, compare, debug)
A list of writing rules: two that every post follows, plus two more picked at random to suit the topicās type
Each run picks a topic I havenāt used yet and keeps a record so nothing repeats.
Then a simple āshow_promptā function displays the final prompt like in the following example.

Example prompt generated by the author
Here are some things I found challenging while building the script:
- How can I add variety to each consecutive prompt? Randomizing the topic and constraints meant that I had to randomly choose a topic, choose 2 constraints associated with the type of the topic selected (for example: if type was āExplainā, I would also need constraints associated with that type and not random constraints for the prompt to make sense). Then I had to combine the default constraints and current prompt constraints into one list and combine it with the selected topic.
- Was there a way to see a prompt but not use it up? Because each run uses up a prompt and saves its id in used, I had to use Pythonās argparse module, which lets me add options like --peek that allows me to only āpeekā at a prompt.
- What happens when all 60 prompts are used up? Is there a way to reset the generator? For this, I added an argparse option --reset, that allows me to reset the prompt generator to make all 60 prompts available again.
This was Day 0 of posting about a tool I built to help myself reach my true potential. Iāll be posting 1 post every day for the next 30 days based on a prompt I get, strictly adhering to the constraints, no skipping it if itās challenging and no AI to help me write or code.
In my own way, this is a little revolt against the complacency weāve willingly embraced in exchange for technological convenience.
Would you join this revolt?
Rupali-Gauravaram
/
dev-prompt-generator
A small Python script that picks my next data science writing topic, with constraints, for a 30-day challenge.
dev-prompt-generator
A small Python script that picks what I write about next. Each run draws one topic I haven't written about yet, from a list of everything I've been learning in data science, and adds a few writing constraints so every post is short and focused.
I'm using it for a 30-day writing challenge: one post a day, starting 3 Nov 2026. The posts live in posts/.
How it works
-
topics.json: 60 topics, each with anid,area(Python, SQL, Statistics, ML and more) andtype(explain, build, compare, debug). -
constraints.json: rules for every post (always), plus rules picked at random for the topic's type (by_type,per_drawof them). -
used.json: the ids already drawn, so a topic never comes up twice. It is created on the first run and kept out of git.
Run it
Run it from inside this folderā¦
Top comments (0)