Tom Slick, Hedy Lamarr the Normandy and Other Things: Lab Book for 2026-06-28/29

 I haven't mentioned my portfolio site here before, but I did manage to fix page view tracking on it today, so that's kind of nice. It was the final project for a digital portfolio class I took at City College San Francisco and highly recommend.

History Research Recap

I'm offloading a significant amount of research work for my history of physics book, The Gladych Files, onto an orchestrated platform of LLM agents in Gas Town. The bulk of the work is to research passengers on trips the main characters of the book took from the 1930s to the 1950s. One of the main subjects of the book is Tom Slick. While returning from a trip to attempt to spot the Lochness Monster while he was a student at Yale—seriously, I love this book—he was aboard a ship, the Normandie, with Hedy Lamarr. The ship also had over 1,000 other passengers including the grandchildren of Henri Matisse. My  agentic AI research team comes to task because it's not a small project to research each of a thousand passengers for the historical significance.

While doing this I'm getting to develope and evolve various LLM strategies including MCPs, eval frameworks and protocols, and token budgeting.

The Hedy Lamarr Mystery

I'm starting work on the mystery  of Hedy Lamarr aboard the Normandy today. I've researched all the passengers aboard the ship, and Hedy Lamarr, whose name at birth was Hedwig Eva Maria Kiesler, is not listed as a passenger. There's a ham radio bent to this as well. Hedy invented and patented frequency hopping spread spectrum radio.

Hedy Lamarr

We know that she was a passenger based on several first hand accounts—including her own—that it was while on ship that she negotiated her hollywood deal with Louis B. Mayer. The lore has it that she was posing as the governess of Grisha Goluboff, a 14 year-old violin virtuoso from San Francisco at the suggestion of Goluboff's manager Isadore Nobel. Although, according to this article, we may not be able to find her because she didn't have a ticket at all.

Grisha on the Normandie

The first step was to pull all the travel manifest data into a database. There were 1,690 passengers on the ship, not counting the remaining crew pages I haven't analyzed yet. I didn't want to spend token assembling analysis results into a database, so I was careful to ask the agent I was working with in Gas Town to 

• first create a schema for the database that we agreed on
• then read the column headings for every table on every manifest analysis page for this particular ship passage. The agents doing OCR and passenger investigations didn't have to stick to column names, so I asked for a json structure mapping varied column names to their corresponding solumn names in the db schema
• armed with this information, I asked the agent to write a python script to scan all 51 pages of findings and assemble the database.

This paid off as I used very few of my Claude tokens on this step.

I installed the llm agent plugin for datasette. I had a bit of a time figuring out the command line to pass in the model I wanted to use. What finally worked was 

datasette --root normandie_1937_1688.db --plugins-dir=plugins  -s plugins.datasette-llm.default_model gpt-5.5

I then, went to the address http://127.0.0.1:8001/-/agent/ to use the agent plugin with my newly created data.

Usage

datasette-agent was handy to have. I know the SQL to do the queries, but didn't have to think about it. Unfortunately, for me, I accidentallly set the model to GPT-5.5 which led to a lot of summaries and subsequently, me asking where the rows I'd seen whiz by on the screen went. I'll be using an older model next time. GPT-5.5 also leaned towards only selecting, or highlighting, passengers from Austria when I started the session. Again with using a 'less' capable model next time.

Results

Not much of any use has turned up yet. Using the database, I narrowed down the field to single women who boarded the ship in England who were younger than 32.

Possible Hedy Lamarr aliases: single women age 16 - 31

Removing the single filter expands the pool of possible passengers and crew a little bit.

Adding married women in the same range

Usage

I had datasette-agent set up to use my API token rather than my codex subscription. The prompt was fairly lengthy, and I made a mistake of setting the default model to gpt-5.5 when it should have been 5.4 mini. Still, the cost was manageable by keeping track of my expenses as I went.