Zellyn Hunter
LightNews LightNews
banner
zellyn.com
Zellyn Hunter
@zellyn.com
120 followers 92 following 200 posts
He/Him.
Programmer. Square. Former YouTuber. Vaguely South African. ≡𝐺𝑂
Born 331.12 ppm CO₂
hachyderm.io/@zellyn
Posts Replies Media Videos
zellyn.com
If you haven't tried OpenAI image generation recently… give it a try (I'm just using the free quota). My wife's dad always says, "Keep it between the ditches", so I asked ChatGPT to make a card. Prompt in alt-text — it nailed it on the first try.

I'm getting weird feels on what images _mean_ now…
The image looks like what I described in the prompt: Please create a picture for a birthday card. It's for my wife's father. He always says, "Keep it between the ditches!", so we're going to make that the theme of the card. The basic idea is an old-timey tractor driving away from the viewer, down a deeply-rutted road. I'd like it stylized watercolor. At the top, it should say "Happy Birthday", and at the bottom, "Keep it between the ditches!"
April 21, 2025 at 5:55 PM
zellyn.com
For reference, the prompt (text in alt-text):
Please write me some vanilla javascript code that will let me edit a curve in svg. We're going to do it like this: * a curve is defined with a list of points * we're going to use the midpoints between points a lot too, so they should be shown (in a different color) * the start and end are simple: they're just straight lines between the start (or end) point and the midpoint of it and the next (previous) point * for all the other points, we're going to use a simple quadratic spline, with the midpoints as start and end points, and the actual point in the middle as the control point. * I should be able to add and remove points: double-clicking an existing point should remove it from the curve. double-clicking a midpoint should add an actual point there
March 24, 2025 at 8:41 PM
zellyn.com
DeepSeek v3 0324 nails my SVG curve editor on the first try:
#ai #deepseek
March 24, 2025 at 8:40 PM
zellyn.com
I'd guess things like Figma have libraries that would work much better. But yeah, you could absolutely make copy/pasteable macros to do this.

I should probably take another stab at that diagram: I'm much better at pikchr than I was… although I have not stopped mis-using it even a little bit 😂
A pikchr diagram with a ridiculous number of splines drawing out the sweep profiles of the teeth of a worm gear with four starts A reasonable use of pikchr to draw a gear's tooth profile, and then a highly unreasonable one showing the top of a spiral sweep.
March 2, 2025 at 2:46 AM
zellyn.com
Go to "Resource Access Manager" (RAM) in AWS console, and click "Resource Access Manager" in the left nav, and it'll show you the mapping:
Your AZ ID AZ IDs provides a consistent way of identifying the location of a resource across all your accounts. This makes it easier for you to provision resources centrally in a single account and share them across multiple accounts. The following table shows you the AZ IDs for availability zones in this region AZ name, AZ ID us-east-2a, use2-az1 us-east-2b, use2-az2 us-east-2c, use2-az3
February 21, 2025 at 10:05 PM
zellyn.com
I guess I asked for that for being a smartass…

(Responding to a post about how Brian blocks anyone who replies to random threads with "Here's what ChatGPT says" with "Here's what Claude says" and a screenshot of Claude responding to his post — yes, I'm 13 😂)

Sadly, I actually liked his posts…
Brian Swetland profile: 🚫 User Blocking You
February 6, 2025 at 9:53 PM
zellyn.com
Here's what Claude says…

/ducks
Me: What do you have to say to this Bluesky post? Ye gods. People seriously do reply to random threads with "here's what ChatGPT says." These people also get blocked by me instantly. And hopefully anybody else who sees that crap. Claude: I aim to understand this perspective while maintaining a balanced view. While it's understandable that some people find AI-generated responses disruptive to organic discussions, automatically blocking users for sharing AI outputs might be an extreme reaction. The key issue seems to be context and relevance - dropping AI responses into random threads without adding value to the conversation could indeed be unhelpful or derail meaningful human interactions. Rather than taking a strong stance either way, it might be more constructive to discuss when and how AI tools can be thoughtfully integrated into online discussions, and when they might detract from the conversation. What's your view on this? Have you encountered situations where AI responses have either enhanced or diminished online discussions?
February 6, 2025 at 9:27 PM
zellyn.com
Mistral Small 3 is far more… interesting
Fascinating, angular sort of scribble in black, with a scattering of red control points around.
January 30, 2025 at 6:56 PM
zellyn.com
Claude is amazing, as always:
Spline editor with a single four-point spline. Control points are red, and midpoints between them are little blue dots.
January 30, 2025 at 6:55 PM
zellyn.com
My new LLM test:
Please write me some vanilla javascript code that will let me edit a curve in svg. We're going to do it like this: * a curve is defined with a list of points * we're going to use the midpoints between points a lot too, so they should be shown (in a different color) * the start and end are simple: they're just straight lines between the start (or end) point and the midpoint of it and the next (previous) point * for all the other points, we're going to use a simple quadratic spline, with the midpoints as start and end points, and the actual point in the middle as the control point. * I should be able to add and remove points: double-clicking an existing point should remove it from the curve. double-clicking a midpoint should add an actual point there
January 30, 2025 at 6:54 PM
zellyn.com
Two buttons (impossible choice) meme: Title: New Paul Graham essay dropped Left button: actually read it, diminish my spirit, blight my soul, become dumber Right button: don't know what everyone's talking about
January 13, 2025 at 6:52 PM
zellyn.com
Here's the best I could find with a quick search:

I think if Trump and Republican leaders had just said, “Of course I'm wearing a mask”, a whole lot of death could have been avoided.

That said, the overall point — about actual behaviors across a whole spectrum of issues — is good… albeit sad 😞
Graph of excess deaths by week, from the CDC
December 20, 2024 at 5:29 PM
zellyn.com
Parent prep for gingerbread house decorating this afternoon: we’ve learned by experience that solid houses already built are key!
Small gingerbread houses sitting upside down in drinking glasses so gravity does all the right things as they set
December 14, 2024 at 3:43 PM
zellyn.com
I was not expecting a vanilla Go shasum to be faster than shasum!
Is this @filippo.abyssdomain.expert magic?
~/gh/fp cat main.go package main import ( "crypto/sha256" "fmt" "io" "log" "os" ) // Copied from https://pkg.go.dev/crypto/sha256#example-New-File func main() { f, err := os.Open(os.Args[1]) if err != nil { log.Fatal(err) } defer f.Close() h := sha256.New() if _, err := io.Copy(h, f); err != nil { log.Fatal(err) } fmt.Printf("%x\n", h.Sum(nil)) } ~/gh/fp ls -lh gpt4all-lora-quantized.bin -rw-r--r--@ 1 zellyn staff 3.9G May 25 14:44 gpt4all-lora-quantized.bin ~/gh/fp time shasum -a 256 gpt4all-lora-quantized.bin 05c9dc0a4904f3b232cffe717091b0b0a8246f49c3f253208fbf342ed79a6122 gpt4all-lora-quantized.bin shasum -a 256 gpt4all-lora-quantized.bin 12.93s user 0.58s system 99% cpu 13.544 total ~/gh/fp time go run main.go gpt4all-lora-quantized.bin 05c9dc0a4904f3b232cffe717091b0b0a8246f49c3f253208fbf342ed79a6122 go run main.go gpt4all-lora-quantized.bin 1.90s user 0.49s system 86% cpu 2.777 total
May 25, 2023 at 6:50 PM