How? Logically I don't get it.
(I'm genuinely confused by the "How?" question)
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
Reminds me of a project I was peripherally involved with many moons ago. The codename for the project was "Tardis" from Doctor Who. No problem there. But we ended up having to redo a significant portion of it later, and someone had the bright idea of changing the redo codename to "ReTardis". It was hilariously juvenile at the time, but I could see how, decades later as society gotten less tolerant of that kind of humor, the codename probably has become objectionable.
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
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.
They're my tools in my toolbox for my code. For most of the projects I've used them in, I've been the sole developer and the issues other folks have raised about naming schemes don't apply. I've shared them with colleagues, but if they use them they can call them whatever they'd like -- I'm not trying to say my way of doing things is better for everyone, but it is better for me. And maybe could be for someone else too.
(Sorry for both a three item list and a "not X, is Y" phrasing in my reply. Oh jeez -- and an em dash too. I'm working on moving my writing style away from what LLMs are throwing out there right now, but it's slow going.)
For new hires (or people in other orgs), shouldn't need long product descriptions trying to explain team lingo means.
It's not like a list of six LLM sub-agents is difficult to hand out, and there's even a public blog post detailing the names, specializations, and rationale for this in case you somehow forget and can't just /list-agents or whatever.
If they have non-descriptive human names, they should behave like people.
- Our payment system is down - Call George on the Phone and ask him to fix it..
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.
I just finished Lies Weeping, which is #12, I think. There’s 2 more on the way. I suspect they are already written.
> This isn’t whimsy; it’s how I remember who the work is actually for.
> These aren’t chatbots with personalities; they’re specialized configurations I invoke by name to focus my intent.
> That’s when I realized the naming wasn’t a quirk. It was a practice.
It is a quirk
> I’m not asking for a generic security scan. I’m saying that I need to look for what I missed.
You aren't asking for a generic security scan? It seems like you're asking for a generic security scan.
> I need to look for what I missed. 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.
> The names aren’t just labels. They’re invocations. They shape my intent before the work even starts.
They are just labels.
Unfortunately other topics are still catching me off guard, like the article about complex numbers posted today which I managed to get through a third of before realizing all the grating bits I was reading were because it was from an LLM.
To be fair, I certainly name my tools. But I didn't have to use AI to invent a whole bunch of "personalities" for them.
The LLM loves to torture concepts into statements with pithy veneers and three-item. Punctuated. Lists. “Pain points” as an example, really? All of these terms are just more specific than the ones they’re contrasted with, which don’t have much of a human element to them to begin with.
The irony of bemoaning this while AI-mimicking a team of people and getting a computer to write for you in its own voice…
As I've been asking Claude to "keep planning criticize ultrathink" very often and repeatedly, maybe I'll make a planning agent, one that helps me shepherd each plan well.
GZGavinZhao•1mo ago
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.