A dictionary app cut its first AI result from 13.3s to 3.0s
UrLingo is a personal dictionary app where a person searches a word, the backend checks login, limits, and ps, OpenAI creates a structured dictionary entry, and the frontend shows the answer by it. The first useful output originally took more than 13 seconds to appear. OpenAI TTFT was 8,296 milliseconds, and the frontend received its first OpenAI chunk after 13,274 milliseconds.
The biggest waste was 1,088 hidden for a simple word definition. After and fixing the path, the latest results showed OpenAI TTFT p50/p95 at 1,247 milliseconds and 3,514 milliseconds. The first frontend OpenAI chunk fell to 3,038 milliseconds p50 and 4,873 milliseconds p95, with hidden reduced to 0.
Overall, TTFT p50 became 6.7 times faster, the first frontend chunk p50 became 4.4 times faster, the first frontend chunk p95 became 2.7 times faster, and priority tier was enabled on every run.
Key points
- The first useful frontend output dropped from 13,274 milliseconds to 3,038 milliseconds p50.
- OpenAI TTFT p50 dropped from 8,296 milliseconds to 1,247 milliseconds, a 6.7 times speedup.
- Hidden went from 1,088 to 0 for a simple dictionary response.
- The slower p95 case for the first frontend chunk improved to 4,873 milliseconds.
- Simple AI features should avoid unnecessary and measure the first visible response separately.