Jered Sutton
@j9n.dev
Bit plumber | ☕️ aficionado | My words are my own
My note-taking and journaling keep evolving. After testing countless tools, I realized success depends on sticking to your system—and refining it over time.
January 7, 2025 at 1:05 AM
My note-taking and journaling keep evolving. After testing countless tools, I realized success depends on sticking to your system—and refining it over time.
Phthalo green is a mind worm implanted by Bob Ross
January 3, 2025 at 2:12 AM
Phthalo green is a mind worm implanted by Bob Ross
One of the reasons I love running sound is that I get to listen to music, press buttons and generally feel like I’m flying a spaceship.
May 12, 2024 at 5:15 PM
One of the reasons I love running sound is that I get to listen to music, press buttons and generally feel like I’m flying a spaceship.
The world needs more callouts to great UX. This is a loudness meter for broadcast audio and it is a pleasure to look at and conveys information in an intuitive way.
May 12, 2024 at 4:58 PM
The world needs more callouts to great UX. This is a loudness meter for broadcast audio and it is a pleasure to look at and conveys information in an intuitive way.
Using physics to drain a kiddie pool. Syphons are so cool.
May 11, 2024 at 8:56 PM
Using physics to drain a kiddie pool. Syphons are so cool.
A new paper by researchers at Rutgers proposes embedding Large Language Models into an operating system with the OS kernel and the LLM kernel living side by side. This is similar to an idea I came across which proposes thinking of an LLM like a new type of CPU a "logic processing unit".
March 27, 2024 at 11:25 PM
A new paper by researchers at Rutgers proposes embedding Large Language Models into an operating system with the OS kernel and the LLM kernel living side by side. This is similar to an idea I came across which proposes thinking of an LLM like a new type of CPU a "logic processing unit".
Nice try Claude 3. But really, great work from the Anthropic team. The model feels very close to GPT4 in terms of capability. If you are in AWS and worried about data sovereignty this might be the model for you. At the moment, only sonnet is available in Bedrock, but I imagine that will change soon.
March 8, 2024 at 2:26 AM
Nice try Claude 3. But really, great work from the Anthropic team. The model feels very close to GPT4 in terms of capability. If you are in AWS and worried about data sovereignty this might be the model for you. At the moment, only sonnet is available in Bedrock, but I imagine that will change soon.
Mind your prompt lengths. A new paper explores the relationship between input length and performance of large language models. The study found that performance can begin to significantly degrade with as few as 3000 tokens.
March 1, 2024 at 12:20 AM
Mind your prompt lengths. A new paper explores the relationship between input length and performance of large language models. The study found that performance can begin to significantly degrade with as few as 3000 tokens.
The result of a discussion with the kids about the existence of bean and apple pie. DALL-E is a window into pocket dimensions.
February 26, 2024 at 12:00 PM
The result of a discussion with the kids about the existence of bean and apple pie. DALL-E is a window into pocket dimensions.
I know what you are thinking, this all looks like a schema, you should just define it. Maybe call it something like PromptJSON.
![{
"Scene": "Elegant portrait of a person in a Victorian setting",
"Subjects": {
"Main": ["person"],
"Secondary": ["vintage armchair", "ornate wallpaper"]
},
"Attributes": {
"Style": "classical painting",
"Mood": "sophisticated",
"ColorScheme": "rich, muted colors"
},
"Constraints": {
"Dimensions": "800x1000",
"AspectRatio": "4:5"
},
"Options": {
"PersonDetails": {
"Gender": "female",
"Age": "early 30s",
"Clothing": "Victorian dress",
"Expression": "thoughtful"
},
"Lighting": "soft, from the left"
}
}](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:omlwdlbtb34qlsq4pnz2eoh7/bafkreiczrwwtptzwueuvhqc4uolhjalcefzxuvmlpfcmgm63yvm3yqx6jq@jpeg)
February 23, 2024 at 11:44 AM
I know what you are thinking, this all looks like a schema, you should just define it. Maybe call it something like PromptJSON.
But what if we wanted to take inspiration from tools like Kustomize? We could call this hypothetical tool Promtize. It would merge a base YAPL file with an overlay allowing you to change subjects and keep scenes and themes.
![Scene: "a tranquil forest in winter"
Subjects:
- Main: "trees"
- Secondary: ["stream", "snow"]
Attributes:
Style: "photorealistic"
Additional:
Mood: "serene"
TimeOfDay: "dusk"](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:omlwdlbtb34qlsq4pnz2eoh7/bafkreig7rmzxwom5qlxhqk7r7ga7ir6xn6krixda6uoaz6jk7vjw6h5wym@jpeg)
February 23, 2024 at 11:40 AM
But what if we wanted to take inspiration from tools like Kustomize? We could call this hypothetical tool Promtize. It would merge a base YAPL file with an overlay allowing you to change subjects and keep scenes and themes.
Or maybe Toms Own Prompting Language (TOPL).
![[Scene]
description = "busy metropolitan street at night"
[Objects]
main = ["neon signs", "crowds of people"]
secondary = ["street vendors", "brightly lit skyscrapers"]
[Styles]
style = "hyper-realistic"
color_scheme = "vibrant neon colors"
time_of_day = "night"
[Parameters]
size = "1792x1024"
no_animals = true
[Options]
depth_of_field = "shallow"
perspective = "bird's-eye view"](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:omlwdlbtb34qlsq4pnz2eoh7/bafkreiao7wlyziisecb2atckpwnggtd7vl7xgvdc3q72bnatmlmfyjjcna@jpeg)
February 23, 2024 at 11:39 AM
Or maybe Toms Own Prompting Language (TOPL).
Why use an ambiguous language like english for prompting image generation models when you can use Yet Another Prompting Language (YAPL).
February 23, 2024 at 11:37 AM
Why use an ambiguous language like english for prompting image generation models when you can use Yet Another Prompting Language (YAPL).