Sign in

Sam Minnée

@sminn.ee
609 followers 477 following 1.5K posts

Techie, dad, Wellingtonian. Building with AI, trying to do so thoughtfully. Loves nuanced discussion of complex topics, and toilet humour. Founded Silverstripe in a past life, now CTO for tellfrankie.com and askastro.ai

PostsRepliesMedia
Sam Minnée @sminn.ee · 20/03/2026
Working from home has been a game changer for school-age parenting. Kids can largely mind themselves for a few hours at this age, but you need to be in the same building.
021
Sam Minnée @sminn.ee · 20/03/2026
Prescient of the National govt to tear down all the polices encouraging EV uptake, eh?
030
Sam Minnée @sminn.ee · 18/03/2026
Maybe I could be a lobster…
“Scientists say lobsters feel no pain” guardian article
000
Reposted by Sam Minnée
SE Gyges @segyges.bsky.social · 16/03/2026
the actual reason to run it when you're using an llm is that it throws so many compile-time errors that it is very difficult to write bad code in it
3251
Reposted by Sam Minnée
maia arson crimew 🏴 @crimew.gay · 14/03/2026
guy who does buddhism slightly wrong and ends up in nevada after he dies
7959111653
Sam Minnée @sminn.ee · 14/03/2026
This post reminded me of being 8 years old, reconfiguring our IBM PS/2 Model 30 every few weeks, to the confusion and frustration of my parents, trying to work out what you could program with DOS batch files.
010
Reposted by Sam Minnée
Sameer Vasta @sameervasta.bsky.social · 13/03/2026
This is the best product review I’ve read: samhenri.gold/blog/2026031... It doesn’t matter if you’re interested in the MacBook Neo or not; this is the kind of essay that makes you think about the potential of technology and the joys of exploration.
samhenri.gold
“This Is Not The Computer For You” · Sam Henri Gold
Sam Henri Gold is a product design engineer building playful, useful software.
341826630
Sam Minnée @sminn.ee · 13/03/2026
LLMs aren’t fancy autocomplete—Chris Luxon is.
140
Sam Minnée @sminn.ee · 12/03/2026
When I use auto-steer, am I vibedriving?
000
Sam Minnée @sminn.ee · 09/03/2026
Weird things happened today. I got up and my laptop had an alarm set for 2am. Then this evening my HomePod speaker randomly said “is this Sam?” or “is it Sam?” Bit creeped out tbh
220
Sam Minnée @sminn.ee · 09/03/2026
Say it again: Alignment ~= Reliability!
020
Sam Minnée @sminn.ee · 08/03/2026
Product idea: slack, but instead of frantic back & forth distracting the rest of your team, you’re chattering with an AI to work out what you’re really trying to say. The rest of the team reads the more thoughtful responses that drop out the end, designed for async consumption.
110
Sam Minnée @sminn.ee · 06/03/2026
I haven’t tried this yet but looks really interesting. Tools that make us better at the shifting skills we need to master in this brave new world are a win!
170
Reposted by Sam Minnée
Sam Minnée @sminn.ee · 04/03/2026
Maybe frustration is the human emotion for identifying problems with our agent harness.
042
Sam Minnée @sminn.ee · 28/02/2026
I’ve heard of a dead pixel, but never a dead ant! This ant has managed to get underneath the outer layer of glass on our TV, and die there. FML. 😬
130
Sam Minnée @sminn.ee · 25/02/2026
@keithng.bsky.social this post about how an economic analyst uses CLI coding agents reminded me of some of the work you do; thought you might find it interesting. newsletter.mikekonczal.com/p/three-ways...
newsletter.mikekonczal.com
Three Ways Terminal AI Has Changed How I Work (And Whether It's Coming for My Job)
How terminal AI compresses setup, robustness checks, and iteration without replacing judgment, and whether Olivia Rodrigo caused the inflation wave.
020
Reposted by Sam Minnée
Sam Rose @samwho.dev · 19/02/2026
What do LLMs see? I wrote a lil' tool that extracts the attention matrices out of open models and creates this typing visual, with each token's opacity changing according to its average attention score as the prompt progresses. Dimmer words are considered less important to the model.
1726444
Sam Minnée @sminn.ee · 18/02/2026
Another trick - set your e2e tests to enable this, and then you don’t need to wait for animations transitions, speeding your tests up bit. Faster tests, better accessibility – a win-win!
130
Sam Minnée @sminn.ee · 18/02/2026
I’m out of Claude Code tokens til 9pm. It’s 7.53pm. I don’t have a problem. I don’t have a problem.
280
Sam Minnée @sminn.ee · 05/02/2026
Anthropic have put their toe into agent orchestration. Something missing from this, and I gas town too, is input multiplexing: if I’m running a lot of agents and I’m not a cowboy, there are a lot of steps to approve and outputs to review. Bouncing between Claude windows is confusing & tedious.
230
Sam Minnée @sminn.ee · 27/01/2026
As someone a world away from the US implosion, this was one of the few articles about the tragedy I was moved to read. The part about how a mother sees their child hit hard, thinking of my late brother… Now I need to wash my face. 😭 Thanks, Kristen, for sharing.
020
Sam Minnée @sminn.ee · 27/01/2026
Robots are like 2018-2023 LLMs right now, a steady stream of technically impressive but still not practically useful updates.
140
Sam Minnée @sminn.ee · 20/01/2026
Live tweets of a bus driver essentially kidnapping his passengers. Insane stuff.
100
Sam Minnée @sminn.ee · 18/01/2026
TIL most all paracetamol dosages in NZ (which is the same as of acetaminophen) are the same as tylenol “extra strength” - I’ve never seen 325mg paracetamol.
040
Reposted by Sam Minnée
Ben Recht @beenwrekt.bsky.social · 17/01/2026
Claude Code UX
1433
Reposted by Sam Minnée
Mrs. Detective PikaBOO, Esq. @clapifyoulikeme.favrd.social · 16/01/2026
“Oughtposting vs isposting” is immediately load bearing for me
4552128
Sam Minnée @sminn.ee · 01/01/2026
When people say that GPT-4o is their friend.
130
Reposted by Sam Minnée
Citizen Platano 🇵🇷 @daniloc.xyz · 30/12/2025
I think about this a lot
massive boulder rolling downhill on a grassy slope, chasing several fleeing figures.

The boulder is labeled with:
	∙	improving model capabilities
	∙	emerging tools and practices
	∙	serious actual issues exacerbated by LLMs
	∙	increasing viability of local inference
	∙	declining training and inference costs
	∙	products impossible before LLMs
	∙	multiplying power of plutocrats
	∙	new digital abuse vectors
	∙	using fossil fuels to sidestep decrepit energy grid

The fleeing figures shout:
	∙	It doesn’t work!
	∙	It can’t search the web!
	∙	It produces bad code!
	∙	I miss my ZIRP labor power and need something to blame
	∙	it’s not inevitable
	∙	it’s a bubble
	∙	*it’s not inevitable*
	∙	shame will solve it
	∙	water! WATER!
Caption reads: “bringing a stale mozzarella stick to a paradigm shift.“​​​​​​​​​​​​​​​
1172
Sam Minnée @sminn.ee · 08/12/2025
This speech transaction from Cory Doctorow is a great set of skeptical but nuanced takes about AI. It’s more pro-worker than anti-AI, and it’s got me thinking about a few things… 🧵
120
Sam Minnée @sminn.ee · 24/11/2025
I’m moving from Fusion to FreeCAD and these tutorials have been great. Any other 3D printing modellers out there? What are you using?
141
Sam Minnée @sminn.ee · 21/11/2025
Personally I think I’d get more edification out of reading Jeffrey Goldblum’s emails, how about you?
110
Sam Minnée @sminn.ee · 13/11/2025
Lady at the bottle store asking for non-alcoholic pre-mix drinks. So… soda?
120
Sam Minnée @sminn.ee · 08/11/2025
IYKYK
020
Sam Minnée @sminn.ee · 03/11/2025
What do you call a pink ninja? A flaminja.
210
Reposted by Sam Minnée
werewolf ver mitzvah 🗿 @ver.ooo · 02/11/2025
karma police, label this account he talks in memes, he's replying to the news he's like a teenage llm
1113
Sam Minnée @sminn.ee · 25/10/2025
I just watched A House of Dynamite on Netflix and I think they accidentally deleted the last 10 minutes of the tape.
000
Sam Minnée @sminn.ee · 21/10/2025
Great little write-up of a handful of data storage techniques, combined to make a key store, mirroring the algorithms used in production ones.
010
Sam Minnée @sminn.ee · 17/10/2025
“Cautious optimist” favouring regulation while being optimistic about the tech’s prospects seems about right for me.
My AI-dentity is the cautious optimist
000
Reposted by Sam Minnée
tanadrin @tanadrin.bsky.social · 10/10/2025
all scientists can be classified according to just two variables: whether they're more mad physics or mad biology, and whether their hubris is in calling up what cannot be put down, or seeking to understand what no man was meant to know
we were definitely not meant to know about that worm with a hundred buttholes, for instance
812534
Sam Minnée @sminn.ee · 26/09/2025
Sure do!
010
Sam Minnée @sminn.ee · 26/09/2025
The real “the rapture is coming” story of the moment is AI data-centre spending. A little more grounded in reality, but equally a product of religious fervour
000
Sam Minnée @sminn.ee · 25/09/2025
Yes you can, all you need is a dry mouth, some cellphone and a microphone.
010
Sam Minnée @sminn.ee · 25/09/2025
Chat, can you use a vegan’s poo as fertiliser?
100
Reposted by Sam Minnée
Sam Minnée @sminn.ee · 24/09/2025
Logically, all nouns starting with a vowel should be amended to start with a nen.
111
Sam Minnée @sminn.ee · 21/09/2025
I don’t know why my wife is so upset. I wanted frozen fish sauce cubes and our fridge has an ice maker, what’s the problem?
120
Sam Minnée @sminn.ee · 19/09/2025
All your favourite Disney+ shows are also available on thepiratebay.org
bay.org
020
Sam Minnée @sminn.ee · 10/09/2025
Product idea: ChatGPT but instead of an LLM the responses are generated by strangers on the internet. Chat, is this a good idea?
120
Sam Minnée @sminn.ee · 06/09/2025
Any #ClaudeCode users on my feed? I’ve been working on an MCP this weekend to provide a range of refactoring operations, notably including move function, rename file & folder, which I’ve not seen other LSP/MCP packages do. Anyone interested in this? Why / why not? Should I open-source it?
110