Loading…
19
Finally admitted my prompt engineering was 80% lazy shortcuts
Spent 6 months writing one-liner prompts, blaming the model when outputs came out flat. Caught myself in a meeting yesterday explaining a failure to a client, and realized the actual problem was me missing context and examples. Who else found that adding a single reference case fixed their results overnight?
2 comments
Log in to join the discussion
Log In2 Comments
aaronsullivan1mo ago
Ugh, same here! I spent weeks feeding my model these bare-bones prompts and then wondering why it sounded like a robot wrote a ransom note. Throwing in one solid example of what I actually wanted made the next output look like a completely different tool. It's honestly embarrassing how long it took me to figure out that the model wasn't the lazy one.
2
kai_webb911mo ago
The model wasn't the lazy one" hits way too close to home. Same thing happened to me, I blamed the tool for weeks when my prompts were pure trash. One example flipped everything, like night and day.
3