Running Report Card - bespoke narrative reports from Strava data

Hi All, I’ve just had approval for a fun little app I’m working on.

Around a year ago I created the Running Report Card - a report that is generated from Strava (a popular running app) data:

It creates reports like this:

My goal is to make these reports bespoke, engaging and interesting. At the moment the logic is all procedural, and as a result it is quite a lot of work to get that bespoke feel.

Using GPT3 I can throw away all of this logic, replacing it with a prompt that gives a number of examples, and let GPT3 do its magic.

If you’re interested in the details, here’s an example prompt:

It really does a fantastic job, here’s an example completion:

In the last year Sara has been on a very respectable 79 runs, covering a total of 311 miles over 73 hours. Her longest running streak is 3 days, she's on a roll! If she keeps clocking up this mileage she'll have run the length of the UK in three years, what a versatile athlete. We'd classify Sara as a dedicated recreational runner, clocking up an impressive mileage. During these runs she has climbed 3,786 feet, that's the equivalent of climbing Snowdon three times (but with better weather)

If anyone has suggestions on better prompt design, or any other feedback, let me know. I’m looking to get the new GPT3-powered version live in the next week or two.

Thanks, Colin E.

2 Likes

HEY! that’s a neat idea mate!!!
So if I have
{
“Name”:“Bloggins”,
“Last Name”:“Bloggy”,
“Weekly Calls”:“100”,
“Weekly IMs”:“1000”,
“Files Worked On”:“300”,
“Emails Sent”:“5643”,
“Time spent on FB in hrs”:“42342345656564332”,
“Time spent actually working”:“423”
}
I could spit back a person’s weekly narrative … hrm… I like this. Creepy… but I like it :slight_smile: