Mike Caulfield
Saturday, October 10
Don’t throw away your rationales!
Or how I built 46,000 description cards that still reflect my months-long journey
mikecaulfield.substack.com
Friday, October 9
In the world of LLMs, the description is the hyperlink
The Memex has arrived, for anyone with the sense to see it.
mikecaulfield.substack.com
Sunday, October 4
Yes, AI can watch a dramatic film and fully explain its motives, relationships, and mechanisms.
The gap between what people think AI’s capabilities are and what they actually are continues to widen.
mikecaulfield.substack.com
Wednesday, September 30
Fact-checking a film summary by having your computer watch the film
My fact-checking AI agent now can stream public domain films, make contact sheets, pair that with transcribed dialogue from the audio stream, and fact-check the film summary.
mikecaulfield.substack.com
Monday, September 28
Yes, when LLMs don’t use tools they still suck at niche questions
And that’s important for nonobvious reasons
mikecaulfield.substack.com
Tuesday, September 22
A sample Jev-based classroom activity: Applying Northrop Frye’s mythoi to films of the 2000s
A step-by-step guide.
mikecaulfield.substack.com
Monday, September 21
Jev might be overhyped as a solo tool, but solo is not the point
I’ll have more to say about this later, but I do think Jev is going to have a non-trivial impact on how we use AI.
mikecaulfield.substack.com
Saturday, September 19
The “Experts Excel and Novices Don’t” critique gets the AI educational situation backwards
We want technologies in classrooms that reward knowing things, it’s good actually
mikecaulfield.substack.com
Saturday, September 12
Prompts can be poetic
It’s worth thinking about an LLM not as something that writes, but something you write into.
mikecaulfield.substack.com
Thursday, September 10
Agents are portable but judgments are not
This is a bigger problem than people realize
mikecaulfield.substack.comCracking Style
Getting LLMs to understand film style has been exhausting, but I’m getting closer
mikecaulfield.substack.com
Sunday, September 6
AI as “fuzzy disintermediation”: some patterns and risks
Some thoughts I haven’t seen elsewhere
mikecaulfield.substack.com
Thursday, September 3
Yes, Google’s AI Overview (AIO) still hallucinates quite a bit
And AI Mode doesn’t and we should teach students that
mikecaulfield.substack.com
Tuesday, September 1
Neither the restaurants using AI slop photos nor the people critiquing them care about pictures
At enough intensity, AI snobs and AI slobs are just two sides of the same coin
mikecaulfield.substack.com
Monday, August 31
“Just reason it out”: A simple follow-up to get the LLM to fall back on semantic capabilities vs. search?
Most times we are begging the LLM to ground in search, but sometimes we need the opposite.
mikecaulfield.substack.com
Tuesday, August 25
Note on Masley’s “I think the data center backlash is mostly about data centers”
Three related points
mikecaulfield.substack.com
Thursday, July 30
My De-Claudifying Memory File to Ditch Claude's "Word Mosaics"
How I at least partially mitigate Claude Fable's ridiculous yoda-speak
mikecaulfield.substack.com
Tuesday, July 21
Blog Posts as LLM Repair Code
To Fix the Crappy Prompts Claude Wrote for You, Have Claude Read My Blog Post
mikecaulfield.substack.com