Find the bottleneck in profiler output
Prompt that reads profiler or trace output, names the real bottleneck with evidence and suggests measurable improvements one at a time. Paste it into an AI tool of your choice, which then finds the bottleneck.
- Topic
- Coding & agents
- Type
- Prompt
- Level
- Advanced
- Time
- about 10 min
Results depend on the tool, the model and what you enter.
Result preview
Example: written by our editors, not real AI output.
Highest self time: format_price() · 41% · 12,400 calls per request Highest total time: load_orders() · 78% · of which 60% waiting for the database Likely cause: one query per order line (the line "SELECT … WHERE order_id = ?" appears 12,400 times). 1. Load order lines in one query. Measure: queries per request, p95 time. 2. Cache the currency format per request. Measure: self time of format_price(). Only a guess: the index on order_id is missing.
Fill in
What you type stays in this browser window. We store and send nothing.
- API endpoint for the order list
- nightly import job
- loading the dashboard
- data export to CSV
- search with filters
- under 300 ms at the 95th percentile
- job finished in under 10 minutes
- first view in under 2 seconds
- half of today's time
- no timeouts with 10,000 rows
Your prompt
You have edited the prompt. Changes in the fields above no longer overwrite it.
This is an instruction for an AI, not the result itself.
Where this prompt works
These AI tools are a good fit for this prompt:
How to use it
- Fill in the fields or tap a suggestion. Leave out names, addresses and other personal data.
- Copy the finished prompt into an AI tool of your choice.
- Measure before and after every single change under the same conditions and keep only changes with a clear gain.
Why does this work?
Separating self time from total time, and computing from waiting, stops you optimising the wrong function. One change at a time keeps cause and effect visible.
Tips
- Remove internal host names, keys and customer data from the output before you paste it.