Conclusion

Personal

• When LLMs accelerate analysis vs. when they substitute for understanding you need to develop yourself
• Fatigue

Where to go next

• We wish we knew…

Exercises

1. Name one task from today where an LLM saved significant time
   • Estimate how long it would have taken without it
2. Name one task from today where you had to correct the LLM
   • Describe what you needed to know in order to spot the error
3. Find an energy cost estimate for a single LLM API call
   • Compare it to the cost of sending an email or running a SQL query
4. Identify one part of your analysis where you are not confident in the result
   • Write down exactly what you would do to check it
5. Find the official documentation page for the tool in the stack you feel least confident about
   • Identify one concept you want to learn next
6. Write a two-sentence policy for your own use of LLMs in research
   • What you will/will not use them for
   • What you will always verify
   • How you will keep yourself to this