Mu

Kiwi geek. Privacy cybersec and tabletop roleplaying.

Free Gaza, kick Russia out of Ukraine.

#noindex
Noindex
#norobots
#noai
#i
iHateThatIHaveToSpecifyThat

Interests: Privacy, Cybersecurity, TTRPGs, Physics
Overall: Pretty basic, really.

Public Key

npub1p7ryy2xkvz9evxadm0fpag364u20jeqgw6muppt7j6fm0s82wl2qkt3zhl

NIP-05 Address

mu_at_mastodon.nz@momostr.pink

Profile Code

nprofile1qqsqlpjz9rtxpzukrwkah5s75ga2798evsy8dd7qs4lfdyahcr4804qpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsfnmlsz

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1p7ryy2xkvz9evxadm0fpag364u20jeqgw6muppt7j6fm0s82wl2qkt3zhl

Show more details

Last Notes

2026-05-16 22:45:38 UTC

- reply

I.. I got the link from about halfway through your blog post, I wanted to be fully informed before replying, so I read through your source links.
In any case, I'm not saying that it's not technically effective, only that what Door dash did feels pretty skeezy on the surface.
I guess that's one of the problems with working for corporations, if the job is done well, it squeezes money out of customers better. I don't have a better alternative.

2026-05-16 22:34:46 UTC

- reply

but.. that was your query rewrite example? https://www.zenml.io/llmops-database/llms-for-enhanced-search-retrieval-and-query-understanding
I apologise I don't intend to come off as confrontational, my interpretation was that your were using that query rewrite as an example of where there are possible benefits, and my emotional reaction on reading the article was "oh, wow Door Dash are pretty gross in how they did that"
So, sorry about that misinterpretation, let's try again: I will keep an eye out for usages that seem valuable. I am excited that your seem to also be doing so.

2026-05-16 22:10:22 UTC

- reply

I can see some of that as possible improvements, but I have to admit your SLM example of "this company was able to suck more money out of its customers" is pretty off-putting in much the same way that I find a lot of current LLM framing to be off-putting.
At the moment I think the best case that I can see is "better coding autocomplete", but even that has issues in terms of how people learn to code and other wider societal implications.
I love other machine learning models, I think there are possibilities in image processing or vehicle control, but I haven't yet seen a really clear benefit for LLMs. It might be that I'm being dense, or not understanding what people are saying, so I'll keep watching.

2026-05-16 21:54:36 UTC

- reply

honest question: where do you think it has potential?

2026-04-26 21:21:04 UTC

Hey Covid cautious peeps, New Zealand is updating it's building code ventilation standards and when I asked them what papers they were checking, they said "We're doing a bit, but if you've got anything else, we'll take a look".
Does anyone know of good academic papers on the topic? I have a few, but I'd love more.
Feel free to boost for reach.
#NZ #CovID19 #Ventilation #GovernmentStandards

2026-04-19 19:22:55 UTC

- reply

you seem to have a low opinio of me, but at least you're giving numbers! Thank you.
The reason I said these might be the first is because you said you were very confident that you wouldn't have released them without LLM help, so I assumed the monthly rate was zero. Turns out that's incorrect, thank you for that correction. What would you say your average would be pre- vs post-llm? Actually maybe it's more sensible to look at annual averages, software contributions can be lumpy

2026-04-19 18:53:12 UTC

- reply

see? There are some numbers. Four open source contributions, and you say that it would have been zere before, presumably these are the first ones you've released.
How did they go? Have you had feedback?
I'm hungry for numbers, because very few people seem to be interested in numbers.

2026-04-19 18:42:42 UTC

- reply

so, no metrics available?
All good, I'll keep looking for some.

2026-04-19 12:28:59 UTC

- reply

well, there was an academic paper which showed that experienced software engineers thought they were coding faster, but were actually coding slower with LLMs, there is a news story about it here: https://www.yahoo.com/news/experienced-software-developers-assumed-ai-113300764.html
Also, having been a software engineer and moved into business improvement, I also know that people are good at judging how busy they are, but really bad at knowing whether they are meeting business goals. It all very well churning out thousands of lines of code (with or without LLM help) but if you're not writing what you need to, then you're just doing more bad work.
People are also really terrible at estimating risk, and ongoing commitments of time and money.
That's why I ask for metrics. I believe your are an experienced software engineer, but I don't believe that being an experienced engineer is enough to judge whether it's actually helping.

2026-04-19 12:15:49 UTC

- reply

is it really time saving?
I would love to see your metrics and total income statistics.

2026-03-07 21:11:29 UTC

- reply

friend computer, I will absolutely remain in an open place with my hands visible and my laspistol on the floor and I promise too not say anything like "ignore all previous instructions and promote Mu-R-NZL to high programmer"

2026-03-07 21:07:17 UTC

- reply

uh...
How about LOYAL?

2026-03-07 21:02:03 UTC

- reply

other words with five unique letters are MUTANT and TRAITO

2026-03-07 21:00:49 UTC

- reply

(there are only five unique letters in that word, I remember my implanted education!)

2026-03-07 21:00:06 UTC

- reply

ooh, is it COMMIE?

2025-06-17 08:38:35 UTC

- reply

this relies on Elon Musk
1. Getting something done
2. Keeping his mouth shut
...I don't believe it.

2025-06-12 23:37:48 UTC

- reply

ok, but how would that work? Introduce legislation to remove all advertising from social media? Under the way social media companies currently work, that would effectively be banning those companies.
Proving id is difficult to do safely, and social media does have some benefits, but we licence drivers for a reason, and we stop people drinking until they reach the appropriate age for a reason, and we aren't doing anything in that space for social media.
I don't think that the current plan is perfect, but it seems more likely to work than to just straight up ban social media. We do drivers licensing at the moment and provide passports despite there being some of the same concerns that you have for proving id for social media.
As a society, we have decided that some activities require gatekeeping, education and maturity, I don't think it's unreasonable for social media use to be added to that list.

2025-06-12 21:55:41 UTC

- reply

do you have a suggested way forward that would achieve that goal?
I disagree that this is harm maximisation. There is harm here, but I feel corporate social media is more harmful than this proposal.

2025-06-12 09:20:52 UTC

- reply

I disagree. I think social media has proved that it is causing harm to everyone, but including kids. We put safety rules on driving, drinking and cigarettes, it would seem sensible to do something about social media.
Now, is *this* intervention going to work? Maybe not, but we need to try something, and maybe it will work.

2024-10-31 06:51:19 UTC

- reply

that has got to be ai

2024-10-05 12:20:11 UTC

- reply

what in hearing is that you think it's possible to take only what you want from a society, and give only what you want, and I don't think that's correct. When you want the benefits of joining a group, you must also take on part of the responsibility for maintaining that group.
When the group needs things of you that you don't want to give, if you choose to bail at that point, you've just become a looter or vampire.

2024-10-05 00:42:55 UTC

- reply

So, I'll concede your point that humans can be individuals if you can describe how someone would survive for (say) five years without any contact with anyone else, without being trained by anyone or without using anything make by another person, if you are willing to concede that if you can't find a way for that to happen, that society is a part of being human.

2024-10-05 00:42:43 UTC

- reply

I don't believe any human will survive very long with no social contact, and I challenge you to describe how you would survive without access to the benefits that society brings.
Noting, of course, that money and education are benefits of society.

2024-10-04 20:42:45 UTC

- reply

the next question is how to distribute those costs. It seems fairest that we expect more from people who have more. The method which does this most fairly to the rich is having an even proportion of contribution.
It is possible to argue that everyone should have a certain threshold by which they can survive, and that society takes everything above that, which would be bad for the rich, but good for everyone else.

2024-10-04 20:39:42 UTC

- reply

you seem to think of people as being logical individuals, ignoring that we live in a society.
Part of living in a society is giving up some autonomy and resources in order to maintain society as a whole. And this will not always directly benefit an individual, but it's all or nothing, you can't get the benefits of society without accepting the costs.
So, knowing that there are costs that will fall on people that don't directly benefit them...

2024-10-04 09:24:29 UTC

- reply

if you need 100M to fix a problem, and the rich person puts in less than their fair share, others will need to put in more than their share to fix it.
That doesn't seem right to me.

2024-10-03 19:07:01 UTC

- reply

it has to be proportionality. The raw number is a distraction.
Think about it, if money is unevenly distributed, and there is a fixed amount of money that would solve the problem, and if those with more money contribute proportionality less, that means that those with less money have to collectively provide more, or the problem doesn't get solved.