I'd be very curious to see what sort of code / prompting goes in to these agents, and what sort of results you see from them - is the name just a personal reminder, or do the LLM subagents incorporate these philosophies? What sort of behavioral changes do you see from this method?
pooper 2 hours ago [-]
I appreciate the writer actually taking the time to explain why `george`. I have worked in some projects where some thing-a-majing or another is called `valhalla` or `thor` or something or another but there is no documentation as to why it is called that and the people who were responsible for naming them so have already ridden into the sunset. If I ever meet him, I "just want to talk" to this CTO who named US East region 2 as "eu2".
zdragnar 47 minutes ago [-]
The problem is that, in any organization past a few people, someone will eventually wonder if they were the inspiration for a particular name, and not in a good way, or someone might introduce politics or something else divisive.
It's better to have arbitrary names that are memorable in some way but not common enough to be associated with someone living within recent memory.
IMHO, YMMV, yada yada
striking 1 hours ago [-]
It's the sequel to EU. EU 2.
AtheistOfFail 2 hours ago [-]
> I "just want to talk" to this CTO who named US East region 2 as "eu2".
How? Logically I don't get it.
silisili 2 hours ago [-]
Not OP... all I could come up with is they didn't remember US east vs east US, so landed on EU2 meaning 'east US 2.'
fvcffcdddfcfff 2 hours ago [-]
[dead]
hdjrudni 36 minutes ago [-]
How does one person talk to another person who named a thing? Well, you can either meet in person and use your mouths, or you can pick up a phone....
(I'm genuinely confused by the "How?" question)
tomjakubowski 18 minutes ago [-]
Eastern US 2
jpollock 2 hours ago [-]
Consider multinational orgs - "EU2", and collisions with English when speaking "you too".
1 hours ago [-]
hoppp 35 minutes ago [-]
George, Ray, Agatha... Ok
As long as you are the only one managing these systems.
But the moment you involve other people, this is the worst naming possible.
handoflixue 12 minutes ago [-]
If a new hire can handle "Talk to George, he's our security guy" then "Talk to George, it's our security LLM" shouldn't be that much harder?
ryze20245 2 hours ago [-]
> I added Helen to my roster just this week.
decotz 1 hours ago [-]
yeah. cringe.
I debate with myself is this Aghata can truly be trusted to be
> I need to find the secret traveling farther than it should, the data leaking where it shouldn’t, the assumption I made that an attacker won’t make. I need to be paranoid on behalf of the users whose data and trust I’m protecting.
at the end of the day its still an llm. but hey, I want to call Claude _Claudius_ all the time but I don't cause it'll shut me down real quick
orliesaurus 38 minutes ago [-]
It's a good way to name your agents, who do they help/work for... Smart move
GZGavinZhao 2 hours ago [-]
As the end of the article says, to the author this is more of a "ritual".
I don't know how effective it is, but I can't imagine this would undermine the quality of the output, so if it adds a little bit of humor and human-ness to my workflow, I'm happy to try it out.
fragmede 1 hours ago [-]
There are only two unsolved problems in computer science. Cache invalidation and naming things. And off-by-one errors.
ChrisMarshallNY 27 minutes ago [-]
Dehumanize humans, and humanize non-humans.
Makes sense to me…
But seriously, naming things is always a sticky wicket.
I tend to name my various devices as characters from Glen Cook’s The Black Company.
My iPhone is Thai Dei, my iPad is Soulcatcher, my Watch is Goblin, and my Mac is Mogaba. It helps me to keep them distinct from my simulators.
If I wanted really crazy names, I’d use Garret P.I. As a source.
sverhagen 1 hours ago [-]
> This isn’t whimsy
Uh, yes it is? It's just whimsy with an explanation. Long live descriptive, preferably short, names.
handoflixue 7 minutes ago [-]
It's also a reminder - we're not just here for the surface concept of X, we're here for the deeper philosophical reasons of Y and Z. The goal isn't to check off a "disability accessible" checkbox, it isn't even to "think how disabled people might use this" - it's to be actually accessible to all the actual people with actual disabilities.
Trust me, there are a a LOT of people who need this reminder.
I'd expect the difference in prompts produces significantly different LLM outputs, too - tell an LLM to check boxes and it won't show much initiative, but give it a philosophy and it will often suggest ideas you missed.
Vpsteroski 55 minutes ago [-]
George? IDK
Rendered at 03:05:52 GMT+0000 (Coordinated Universal Time) with Vercel.
It's better to have arbitrary names that are memorable in some way but not common enough to be associated with someone living within recent memory.
IMHO, YMMV, yada yada
How? Logically I don't get it.
(I'm genuinely confused by the "How?" question)
at the end of the day its still an llm. but hey, I want to call Claude _Claudius_ all the time but I don't cause it'll shut me down real quick
I don't know how effective it is, but I can't imagine this would undermine the quality of the output, so if it adds a little bit of humor and human-ness to my workflow, I'm happy to try it out.
Makes sense to me…
But seriously, naming things is always a sticky wicket.
I tend to name my various devices as characters from Glen Cook’s The Black Company.
My iPhone is Thai Dei, my iPad is Soulcatcher, my Watch is Goblin, and my Mac is Mogaba. It helps me to keep them distinct from my simulators.
If I wanted really crazy names, I’d use Garret P.I. As a source.
Uh, yes it is? It's just whimsy with an explanation. Long live descriptive, preferably short, names.
Trust me, there are a a LOT of people who need this reminder.
I'd expect the difference in prompts produces significantly different LLM outputs, too - tell an LLM to check boxes and it won't show much initiative, but give it a philosophy and it will often suggest ideas you missed.