The End(s) of Argument
Subscribe
Sign in
Home
Archive
About
Latest
Top
Discussions
Why You Should "Follow Up": Walls of Benin
An example of how an evidence-focused follow-up can make LLMs much smarter
Oct 15
•
Mike Caulfield
3
2
Get It In: Walls of Benin
How pushing the claim or artifact to the LLM for analysis can bring insights you weren't expecting
Oct 14
•
Mike Caulfield
8
1
Get it in, track it down, follow up
Three essential moves to using AI for verification and contextualization, with a bit of LLM-specific guidance
Oct 13
•
Mike Caulfield
11
5
Oh, Lord, Don't Let Me Be Misunderstood
Sometimes the best thing an LLM can do for you is misunderstand you
Oct 11
•
Mike Caulfield
6
2
September 2025
When wrong LLM answers get you to the right information
"Why do you say that?" can be just as instructive with an LLM as it is in real conversation. Also: I make you learn Stalnaker.
Sep 27
•
Mike Caulfield
17
14
AI can walk through all the doors at once
An explanation of why AI can be helpful with sourcing and contextualization and why it sometimes doesn't look that way in retrospect
Sep 23
•
Mike Caulfield
15
1
My Surprisingly Effective Follow-Ups File
I've been holding back on sharing this until I could present this more fully, but let's YOLO it.
Sep 19
•
Mike Caulfield
20
9
I Am Once Again Asking You To Use A Balanced, Evidence-Focused Follow-Up
Cutting through Russian disinformation with thirteen extra words
Sep 17
•
Mike Caulfield
14
LLMs Often Conflate Images In Bizarre Ways. Here's How to Work Around It.
A short demonstration
Sep 15
•
Mike Caulfield
8
6
Do students need to know how LLMs work, or to predict how they'll act?
A little venting about what educational explanations are for...
Sep 12
•
Mike Caulfield
19
9
Is "Are you sure?" a bad follow-up prompt?
A very minimal test with some fairly clear results. (The answer is yes, it's bad)
Sep 9
•
Mike Caulfield
10
Does successful LLM iteration require previous knowledge of the right answer?
Answer: No.
Sep 7
•
Mike Caulfield
15
6
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
