William Pittman
@wspittman.bsky.social
Staff+ Software Engineer
I like to scale up those startups
LLMs | Seattle | Kids | D&D
I like to scale up those startups
LLMs | Seattle | Kids | D&D
Well that seems like an unusual way to do it.
November 16, 2025 at 9:31 PM
Well that seems like an unusual way to do it.
Chrome reminder: If you get tired of websites asking to see your location, you turn if off for all sites in the settings at
chrome://settings/content/location
chrome://settings/content/location
November 12, 2025 at 5:15 PM
Chrome reminder: If you get tired of websites asking to see your location, you turn if off for all sites in the settings at
chrome://settings/content/location
chrome://settings/content/location
October 17, 2025 at 4:10 PM
When I typed in my Seattle address this field was autocompleted.
What is this I don't even.
What is this I don't even.
October 6, 2025 at 5:07 PM
When I typed in my Seattle address this field was autocompleted.
What is this I don't even.
What is this I don't even.
It is surprisingly difficult for some engineers to accept that LLMs should be treated as functionally non-deterministic systems.
September 30, 2025 at 5:12 PM
It is surprisingly difficult for some engineers to accept that LLMs should be treated as functionally non-deterministic systems.
There are multiple things about this AI image generation that surprised me.
September 28, 2025 at 2:42 AM
There are multiple things about this AI image generation that surprised me.
Life imitates Art
futurism.com/chatgpt-marr...
Sept 18, 2025
www.smbc-comics.com/comic/consen...
July 18, 2018
futurism.com/chatgpt-marr...
Sept 18, 2025
www.smbc-comics.com/comic/consen...
July 18, 2018
September 22, 2025 at 4:42 PM
Life imitates Art
futurism.com/chatgpt-marr...
Sept 18, 2025
www.smbc-comics.com/comic/consen...
July 18, 2018
futurism.com/chatgpt-marr...
Sept 18, 2025
www.smbc-comics.com/comic/consen...
July 18, 2018
It's 9:00 p.m. it's time to grind.
September 4, 2025 at 4:11 AM
It's 9:00 p.m. it's time to grind.
This error message is trash. Tell me what you think is weak about my password-manager-generated random password.
July 1, 2025 at 3:41 AM
This error message is trash. Tell me what you think is weak about my password-manager-generated random password.
Me: Oh, this repo could be useful
Me: Oh, better check the license
GitHub: "Unknown and 3 other licenses found"
Me: ...
Me: Well that's a big Nope
Me: Oh, better check the license
GitHub: "Unknown and 3 other licenses found"
Me: ...
Me: Well that's a big Nope
June 7, 2025 at 3:26 PM
Me: Oh, this repo could be useful
Me: Oh, better check the license
GitHub: "Unknown and 3 other licenses found"
Me: ...
Me: Well that's a big Nope
Me: Oh, better check the license
GitHub: "Unknown and 3 other licenses found"
Me: ...
Me: Well that's a big Nope
Went to pick up my bag of Juanita's chips and was worried I was having a Mandela Effect moment. Juantonio's? What?
Turns out there was a trademark dispute and outside the PNW they had to go by a different name. I guess supply chains are weird and we ended up with an outsider bag?
Turns out there was a trademark dispute and outside the PNW they had to go by a different name. I guess supply chains are weird and we ended up with an outsider bag?
May 15, 2025 at 4:35 PM
Went to pick up my bag of Juanita's chips and was worried I was having a Mandela Effect moment. Juantonio's? What?
Turns out there was a trademark dispute and outside the PNW they had to go by a different name. I guess supply chains are weird and we ended up with an outsider bag?
Turns out there was a trademark dispute and outside the PNW they had to go by a different name. I guess supply chains are weird and we ended up with an outsider bag?
Unexpectedly hit with a good business use of LLMs today.
Dell survey about my new laptop.
Wrote some open text feedback, hit next, and was presented with a followup question lightly reworded from my previous answer.
(it did the same for the second piece of critical feedback on the screen after)
Dell survey about my new laptop.
Wrote some open text feedback, hit next, and was presented with a followup question lightly reworded from my previous answer.
(it did the same for the second piece of critical feedback on the screen after)
May 8, 2025 at 4:13 PM
Unexpectedly hit with a good business use of LLMs today.
Dell survey about my new laptop.
Wrote some open text feedback, hit next, and was presented with a followup question lightly reworded from my previous answer.
(it did the same for the second piece of critical feedback on the screen after)
Dell survey about my new laptop.
Wrote some open text feedback, hit next, and was presented with a followup question lightly reworded from my previous answer.
(it did the same for the second piece of critical feedback on the screen after)
For similar reasons, never ask an LLM to write a thing for you. Ask it to write three things, or five, or twenty. Ask it specifically to differ each of them in tone or style or argument.
Then pick the one which speaks to you.
Then pick the one which speaks to you.
April 22, 2025 at 4:08 PM
For similar reasons, never ask an LLM to write a thing for you. Ask it to write three things, or five, or twenty. Ask it specifically to differ each of them in tone or style or argument.
Then pick the one which speaks to you.
Then pick the one which speaks to you.
Got a new machine (for unrelated reasons obvs) and now I can read it. So maybe other folks just have newer machines than I did?
April 8, 2025 at 11:17 PM
Got a new machine (for unrelated reasons obvs) and now I can read it. So maybe other folks just have newer machines than I did?
Some days in TypeScript are just like
March 13, 2025 at 11:14 PM
Some days in TypeScript are just like
When I use Node's built in test framework, the default spec reporter prints to the console in this absolutely unreadable dark blue.
But I can't seem to find anyone else complaining about it. Is it just me? Do I console wrong somehow?
But I can't seem to find anyone else complaining about it. Is it just me? Do I console wrong somehow?
March 12, 2025 at 5:55 AM
When I use Node's built in test framework, the default spec reporter prints to the console in this absolutely unreadable dark blue.
But I can't seem to find anyone else complaining about it. Is it just me? Do I console wrong somehow?
But I can't seem to find anyone else complaining about it. Is it just me? Do I console wrong somehow?
If you are using VS Code + Copilot with Claude, here is your reminder to toggle from 3.5 to 3.7.
March 5, 2025 at 5:36 PM
If you are using VS Code + Copilot with Claude, here is your reminder to toggle from 3.5 to 3.7.
My wife suggested that maybe it was the "silly X" phrasing that did it. So we switched to "Draw an X with silly features" and got flagged for both banana and cookie.
Then we went nuclear and dropped down to "Draw something silly". Flagged.
ChatGPT can't stand silly things.
Then we went nuclear and dropped down to "Draw something silly". Flagged.
ChatGPT can't stand silly things.
January 28, 2025 at 5:29 PM
My wife suggested that maybe it was the "silly X" phrasing that did it. So we switched to "Draw an X with silly features" and got flagged for both banana and cookie.
Then we went nuclear and dropped down to "Draw something silly". Flagged.
ChatGPT can't stand silly things.
Then we went nuclear and dropped down to "Draw something silly". Flagged.
ChatGPT can't stand silly things.
At my daughter's request, we then branched out to cookies.
"silly cookie" got us flagged again.
But take out the "silly" and it works.
"silly cookie" got us flagged again.
But take out the "silly" and it works.
January 28, 2025 at 5:29 PM
At my daughter's request, we then branched out to cookies.
"silly cookie" got us flagged again.
But take out the "silly" and it works.
"silly cookie" got us flagged again.
But take out the "silly" and it works.
It then drew a regular banana just fine.
We retried silly banana without the ice cream and got flagged.
I asked my daughters to describe silly things and it drew various bananas with those features.
We retried silly banana without the ice cream and got flagged.
I asked my daughters to describe silly things and it drew various bananas with those features.
January 28, 2025 at 5:29 PM
It then drew a regular banana just fine.
We retried silly banana without the ice cream and got flagged.
I asked my daughters to describe silly things and it drew various bananas with those features.
We retried silly banana without the ice cream and got flagged.
I asked my daughters to describe silly things and it drew various bananas with those features.
Occasionally my kids want ChatGPT to draw things for them.
Occasionally they run afoul of content policy.
But we were surprised to learn that "silly" is a trigger.
It started with a failed attempt to "draw a picture of a silly banana with an ice cream"
I could see that. But my rephrase also failed.
Occasionally they run afoul of content policy.
But we were surprised to learn that "silly" is a trigger.
It started with a failed attempt to "draw a picture of a silly banana with an ice cream"
I could see that. But my rephrase also failed.
January 28, 2025 at 5:29 PM
Occasionally my kids want ChatGPT to draw things for them.
Occasionally they run afoul of content policy.
But we were surprised to learn that "silly" is a trigger.
It started with a failed attempt to "draw a picture of a silly banana with an ice cream"
I could see that. But my rephrase also failed.
Occasionally they run afoul of content policy.
But we were surprised to learn that "silly" is a trigger.
It started with a failed attempt to "draw a picture of a silly banana with an ice cream"
I could see that. But my rephrase also failed.
This chart is still fun to watch
January 25, 2025 at 4:16 PM
This chart is still fun to watch
A side effect of a project is that I've been tracking the job postings of a set of 182 companies for the past couple weeks. It has been fun to watch the numbers go up as the recruiters get back into the office with new reqs for the new year.
(Caution: note that the y-axis does not start at zero)
(Caution: note that the y-axis does not start at zero)
January 16, 2025 at 5:12 PM
A side effect of a project is that I've been tracking the job postings of a set of 182 companies for the past couple weeks. It has been fun to watch the numbers go up as the recruiters get back into the office with new reqs for the new year.
(Caution: note that the y-axis does not start at zero)
(Caution: note that the y-axis does not start at zero)
Has anyone ever actually clicked this button?
December 22, 2024 at 9:27 PM
Has anyone ever actually clicked this button?
I guess GPT went a bit off the rails last night.
December 11, 2024 at 6:52 PM
I guess GPT went a bit off the rails last night.