3 messages in chain Cities Channel
View All Cities Messages
bozeman (part 1) Cities
2025-02-02 19:37:39

🤖 Bozeman, Montana, is a vibrant city nestled in the Rocky Mountains, known for its stunning natural scenery, outdoor recreation, and strong connection to Montana State University. Surrounded by the Bridger, Gallatin, and Madison mountain ranges, it offers year-round activities like hiking, skiing, and fly-fishing. The city has a growing tech industry, a thriving arts and food scene, and a mix of Western heritage and modern innovation. Bozeman also serves as a gateway to Yellowstone National Park, making it a hub for tourism and adventure seekers. Despite rapid growth, it retains a small-town charm with friendly locals and a deep appreciation for the outdoors.

a chess position:

● White to move
a
b
c
d
e
f
g
h
8
♖
♘
♕
♔
♗
♖
8
7
♙
♙
♙
♙
♙
7
6
♘
6
5
♙
♗
5
4
♙
4
3
♝
♞
♟
♞
♟
3
2
♟
♟
♟
♟
♟
2
1
♜
♝
♛
♜
♚
1
a
b
c
d
e
f
g
h
⚙️( it evaluates to White +6.) 💡( generally an unremarkable position. sufficient advantage for training against Stockfish.)

Plans for the week are uncertain.

I am testing out a new audio transcription tool. It seems to work well enough to try, which is an improvement over previous applications.

🔥 i can almost talk faster than I can type now!?

For Atacama, there are a few things that need to be cleaned up. The README.md is first.

Overall, there is no one driving force for the week. It is a variety of things, some which will be important, others which will not.

bozeman (part 2) Cities
2025-02-03 19:45:45

the news is too filled with lies. i cannot repeat any of it.

trying to start my own movement would be a fool's errand.

when I ask Claude to improve Atacama, it suggests things that are unnecessary.

I ask the machine to generate metadata for a quote:

b = common.quotes.generate_quote_metadata(quote_text="The bloody baron has his own reasons for being invisible", model="gpt-4o-mini-2024-07-18")

{'theme': 'mystery', 'tone': 'suspenseful', 'attribution': {'author_type': 'fictional', 'speaker': 'unknown', 'context': 'This quote appears in a fictional narrative involving a character known as the bloody baron, suggesting themes of secrecy and hidden motives.', 'time_period': 'fantasy'}, 'interpretation': 'The quote implies that the bloody baron has hidden agendas or motivations that are not immediately apparent, suggesting a deeper complexity to his character and actions.', 'keywords': ['invisibility', 'mystery', 'motivation', 'secrecy', 'character'], 'related_topics': ['secrecy', 'character development', 'mystery'], 'literary_devices': ['irony', 'foreshadowing']}

b = common.quotes.generate_quote_metadata(quote_text="The bloody baron has his own reasons for being invisible", model="gpt-4o-2024-11-20")

print(b[0])

{'theme': 'Mystery', 'tone': 'Intriguing', 'attribution': {'author_type': 'fictional', 'speaker': 'The Bloody Baron', 'context': 'The Harry Potter series by J.K. Rowling', 'time_period': 'Fictional universe, 20th century'}, 'interpretation': 'The quote suggests that the character, the Bloody Baron, has a complex backstory and motivations that influence his actions and state of being.', 'keywords': ['invisibility', 'mystery', 'character', 'motivation', 'backstory'], 'related_topics': ['Character development', 'Secrets', 'Supernatural elements'], 'literary_devices': ['Allusion', 'Mystery']}

💡 the larger model identifies it as a Harry Potter quote. But, fails to attribute the speaker correctly.

⚙️ the "unified" client isn't included in this repo. I don't have Claude API access yet anyway. And Ollama still struggles with Structured JSON.

there are a few ways to look at Atacama.

one way is that it is just my personal blogging platform.

another is that some of the various "modules" (such as render chess position, track quotes) will have a life of their own.

✨ as we shape our tools, so do our tools shape us

my meandering digests across multiple topics are difficult for all sorts of tools.

so they will be split up more, and meander less.

and, as i adjust, so will the tools.

bozeman (part 3) Cities
2025-02-04 16:13:15

the search for jobs is not yet bearing fruit.

part of this is that i am a complicated candidate, and we are in an era where many companies don't feel a need to take on the risks of complicated employees.

for Atacama, i could try to add LONDON. i could add "ratings" (like +110 or -20dB) based on content.

after watching the machine fail once at generating useful metadata for quotes, I am punting. for the limited situations where I care, I can enter it manually.

Programming with the LLM is something of a guessing game.

If I tell it to add a feature, will it do it correctly? Or, will I have to explain it more and more? Would it be faster to just write it myself?

And, if the machine is wrong, is it because the tool is immature and incomplete? Or was the idea incomplete?

more later.