David Neuzerling's Avatar

David Neuzerling

@md.neuzerl.ing

Data scientist and R engineer. Also Python and Julia. Accidental econometrician. I mostly post about R, sometimes Georgism/YIMBYism and Warhammer. Melbourne, Australia.

2,621 Followers  |  448 Following  |  297 Posts  |  Joined: 14.07.2023  |  1.6926

Latest posts by md.neuzerl.ing on Bluesky

Every now and then Iโ€™ll think about the 15-year-old Melbourne kid who stole a tram, not to take a joyride but to deliver a tram service. I hope he got his dream job

02.08.2025 00:34 โ€” ๐Ÿ‘ 6    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

โ€œIโ€™m not a NIMBY butโ€ is always followed by the most generic NIMBY opinions

24.07.2025 05:01 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I asked a zoomer and they said โ€œdeleted or just killedโ€.

Iโ€™m asking if they still gib people

24.07.2025 03:48 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I should never have posted this. Iโ€™m just spreading the psychic damage

24.07.2025 03:46 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Got called grandpa for saying โ€œfragโ€, to mean killing someone in a video game. 1d4 points of psychic damage

24.07.2025 03:42 โ€” ๐Ÿ‘ 7    ๐Ÿ” 0    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 0

I am not smart enough to understand a script where a variable is redefined multiple times. Plz donโ€™t do this. My brain canโ€™t handle it โ˜น๏ธ

24.07.2025 01:27 โ€” ๐Ÿ‘ 7    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

The latest episode of Star Trek confirms that Wham! will still be culturally relevant 235 years from now

19.07.2025 00:48 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Itโ€™s somehow less authentic than any other social media, even Instagram

19.07.2025 00:35 โ€” ๐Ÿ‘ 3    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

โ€œChatGPT kills your critical thinking skillsโ€

Nah mate, 10 years in corporate work has already done that

15.07.2025 22:52 โ€” ๐Ÿ‘ 7    ๐Ÿ” 2    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Some LLM hints based on my experiences:

โ€ข Itโ€™s often better to chain a second โ€œrefinementโ€ prompt rather than trying to tweak an existing prompt. If you see common errors in LLM output, get the LLM to correct itself in a second round of processing.
โ€ข Put exceptions before general rules in a prompt.

15.07.2025 06:18 โ€” ๐Ÿ‘ 2    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Ten painted Tโ€™au Vespid Stingwing miniatures with orange-yellow bodies, bright red guns, and pale cream wings with brown shading. An additional Oversight Drone is present, painted in matching colours. All models are on round bases with sand, rocks, and tufts of green and brown grass.

Ten painted Tโ€™au Vespid Stingwing miniatures with orange-yellow bodies, bright red guns, and pale cream wings with brown shading. An additional Oversight Drone is present, painted in matching colours. All models are on round bases with sand, rocks, and tufts of green and brown grass.

10 vespids all done! These were airbrushed from orange, through gold, and finally a bright yellow. Finished off with a raw umber oil wash. Really happy with how these turned out.
#warhammer #warhammer40k

14.07.2025 06:34 โ€” ๐Ÿ‘ 4    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I just want a robot to do my laundry for me why canโ€™t the tech bros work on that instead of wellness optimisation subscriptions

12.07.2025 12:54 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Wrangling LLM output with LangChain The toughest prediction any data scientist makes is deciding which tools are worth learning. The explosion of generative AI only makes this harder. My prediction: LangChain is here to stay or, at leas...

Itโ€™s nothing impressive. Just some basic LangChain code, and some oil-painted Skaven: mdneuzerling.com/post/wrangli...

12.07.2025 06:43 โ€” ๐Ÿ‘ 1    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I didnโ€™t want to use AI slop for the headers of my tech blog posts, but then I realised that itโ€™s a personal blog that I have 100% ownership over, so I just started using pictures of my recently painted Warhammer models instead. Theyโ€™re completely irrelevant to the content. No one can stop me

12.07.2025 05:56 โ€” ๐Ÿ‘ 23    ๐Ÿ” 1    ๐Ÿ’ฌ 2    ๐Ÿ“Œ 0

If youโ€™re submitting an AI-generated resume, at least make sure your LinkedIn URL isnโ€™t a ChatGPT redirect link ๐Ÿ™„

06.07.2025 23:37 โ€” ๐Ÿ‘ 5    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Screenshot of job posting by Unilever for โ€œIce Cream Head of Artificial Intelligenceโ€

Screenshot of job posting by Unilever for โ€œIce Cream Head of Artificial Intelligenceโ€

New AI job title just dropped

19.06.2025 02:11 โ€” ๐Ÿ‘ 3    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 1

You canโ€™t think of any drudgery that you wouldnโ€™t rather outsource to LLMs?

I used it earlier to fix up some mismatched parentheses, for example.

17.06.2025 09:28 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

We are all Murderbot

14.06.2025 12:11 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Post image

Iโ€™m loving the Murderbot adaption, but what I really want is 19 seasons of The Rise and Fall of Sanctuary Moon. Thereโ€™s no where near enough trashy sci fi soap opera

14.06.2025 08:12 โ€” ๐Ÿ‘ 72    ๐Ÿ” 9    ๐Ÿ’ฌ 6    ๐Ÿ“Œ 1

The absolute high of fixing a bug in a programming language you donโ€™t know

10.06.2025 01:15 โ€” ๐Ÿ‘ 24    ๐Ÿ” 1    ๐Ÿ’ฌ 2    ๐Ÿ“Œ 0

Feel like I canโ€™t use emdashes anymore because of the association with AI

04.06.2025 05:24 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Cannot emphasise enough just how useful format-on-save tools like black and ruff are.

Being able to completely stop thinking about code formatting and yet get 100% consistent formatting across a team of arbitrary size? Incredible

04.06.2025 01:59 โ€” ๐Ÿ‘ 12    ๐Ÿ” 2    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Wrangling LLM output with LangChain The toughest prediction any data scientist makes is deciding which tools are worth learning. The explosion of generative AI only makes this harder. My prediction: LangChain is here to stay or, at leas...

New blog post! I use GenAI and Langchain to explore kanban card creation and unit testing.

02.06.2025 21:29 โ€” ๐Ÿ‘ 4    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Markdown should be universally accepted in every text input field

31.05.2025 22:01 โ€” ๐Ÿ‘ 30    ๐Ÿ” 3    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Once you get past the horrendous installation/system management, Python is a genuine joy to use.

Iโ€™m a big fan of black/ruff. Thereโ€™s value in making formatting a zero-thought process.

I like the type hints. I like that theyโ€™re strictly opt-in, but there when I need them.

25.05.2025 05:41 โ€” ๐Ÿ‘ 4    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
A screenshot of PlayStation trophies for Clair Obscur: Expedition 33, with 100% obtained

A screenshot of PlayStation trophies for Clair Obscur: Expedition 33, with 100% obtained

Iโ€™ve never been so obsessed with a game that I actually finished all of the achievements. This is a first.

Iโ€™m going to go door-to-door spreading the word of Clair Obscur. Go play this game

25.05.2025 02:21 โ€” ๐Ÿ‘ 7    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Disco Elysium solved this by having no combat.

Dark Souls solved this by having no story.

15.05.2025 22:59 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

All single-player video games should have difficulty options, not because games are too hard, but because no game worth finishing has a combat system interesting enough to outlast its story. At some point I want to click the button that says โ€œreduce enemy health to 10% so I can wrap this up nowโ€

15.05.2025 22:54 โ€” ๐Ÿ‘ 3    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Never gonna forget when someone came onto my GitHub and abused me in an issue because they thought I had misspelt โ€œfalseโ€ (I had written ifelse)

13.05.2025 00:02 โ€” ๐Ÿ‘ 24    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Clair Obscur is the best JRPG (FRPG?) since Final Fantasy X. Phenomenal art, music, characters, and story. Voice actors include Charlie Cox, Jennifer English, Andy Serkis, and Ben Starr. Donโ€™t sleep on this.

The game is aggressively French. I defeated a mime and it dropped a beret and a baguette.

08.05.2025 02:07 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

@md.neuzerl.ing is following 20 prominent accounts