Sign in

pngwn

@pngwn.at
2.7K followers 313 following 965 posts

Penguin of great renown. The penguin revolution is nigh, only the feathered and the tuxedoed will be spared. Except for seagulls, they can get fucked. pngwn.at

PostsRepliesMedia
Reposted by pngwn
{🧪} +paoloricciuti.svelte @paolo.ricciuti.me · 17h
People that use @pi.dev as their harness..we want to ship a Svelte plugin for it but since it doesn't natively support subagents, which option do you prefer? a. We don't ship subagents for `pi` b. We expose a tool that spawns a new pi instance c. We assume you have `pi-subagents` and configure that
793
pngwn @pngwn.at · 07/10/2026
Don’t undersell yourself!!
010
pngwn @pngwn.at · 07/10/2026
@paolo.ricciuti.me for tirelessly innovating and maintaining svelte ai tools, spearheading svelte custom renderers, and hosting weekly events for the svelte community. We do not deserve Paolo.
1130
pngwn @pngwn.at · 07/10/2026
I also don’t necessarily agree that the abstraction is that leaky, I think the underlying model arguably matches the semantics. Arguing that all abstractions that match the underlying DS are leaky, kind of assumes all DS disagree with the semantics. I’m not sure that is the case here.
010
pngwn @pngwn.at · 07/10/2026
Yeah. I think 3 states is fine tbh and there are valid arguments against it. One is boring old familiarity. People are to tri-state theme toggles.
100
pngwn @pngwn.at · 06/10/2026
i love him. you can wake him up if he goes to sleep!
010
pngwn @pngwn.at · 06/10/2026
Long overdue, but I finally have an internet home.
pngwn.at
a place to gather moss
1101
pngwn @pngwn.at · 06/10/2026
I don’t know if the little green box was a psychology play by the @npmx.dev folks but either way it 100% works on me.
180
pngwn @pngwn.at · 04/10/2026
I think this a real great discussion and I think the core argument is probably correct. What I do think would be helpful is communicating that flipping dark -> light in system light mode is a return to system. Something that communicates dark -> back/ reset to light.
110
pngwn @pngwn.at · 02/10/2026
I found my original demo from back when twinkleplop was just a baby bsky.app/profile/pngw...
120
pngwn @pngwn.at · 02/10/2026
You can do the highlight api thing with twinkleplop too! If you want to ofc.
120
pngwn @pngwn.at · 01/10/2026
feel free to ping me a link and i can take a look whenever!
010
pngwn @pngwn.at · 01/10/2026
ITS REAL!!
0200
pngwn @pngwn.at · 01/10/2026
all fixed! will release again shortly. Thanks!
110
pngwn @pngwn.at · 25/09/2026
Yeah. There are a few reclassifier bugs like this. Now the release is oit and performance is good, I’ll be focusing on new languages and correctness. There are some areas where I’m just not sure though. Eg classes vs types, built ins vs types. Is it just about type position? Etc.
011
pngwn @pngwn.at · 25/09/2026
I have PRs switched off for the time being but if you tell me your GitHub handle I’ll see if I can add you!
110
pngwn @pngwn.at · 25/09/2026
I don’t think any syntax differences are due to performance. Either I couldn’t decide / went for simplicity(should storage kinds have their own token? When is it a type vs a built in?) or they just aren’t implemented (regex is its own language and I forgot). File an issue if you’d like to discuss!
010
pngwn @pngwn.at · 25/09/2026
Let me know if you see any issues with the output! I have many tests but I’m sure there are things I’ve missed. Would also love to add an official angular grammar, so I would love to take a look at that.
210
pngwn @pngwn.at · 25/09/2026
Just released a new twinkleplop version with some significant performance improvements across both tokenisation and rendering! I finally had time to go back and explore a few other ideas. I was also slightly annoyed that I wasn't faster on all sizes. twinkleplop.pngwn.at/docs/benchma...
A bar chart showing a 72% performance increase in twinkleplop version 0.2.2 over the previous version

A smaller set of bar charts show language specific gains:

- html, 68%
- javascript, 57%
- json, 77%
- markdown, 121%
- python, 45%
- rust, 58%
- sql, 76%
- svelte, 84%
- toml, 81%
- tsx, 41%
- typescript, 42%
060
pngwn @pngwn.at · 23/09/2026
Hero! Let me know how you find it!
011
pngwn @pngwn.at · 23/09/2026
Nice! If it is missing anything file an issue and I can a look! The code for the splash page across a few components here: github.com/pngwn/twinkl... It was lots of fun to make!
github.com
twinkleplop/lib/_site/src/routes/+page.svelte at main · pngwn/twinkleplop
Syntax highlighter and code authoring toolkit. Contribute to pngwn/twinkleplop development by creating an account on GitHub.
110
pngwn @pngwn.at · 23/09/2026
Yeah! And we can talk about it again soon! Shiki was a huge inspiration in terms of feature set, it set the standard for what people expect of a highlighting library. I just thought we could bring the cost down.
130
pngwn @pngwn.at · 22/09/2026
Added some docs so you can use twinkleplop with mdsvex! mdsvex.pngwn.at/docs#with-tw...
mdsvex.pngwn.at
mdsvex - svelte in markdown
Combine svelte and markdown in the same file. Live your dreams!
080
pngwn @pngwn.at · 22/09/2026
Déjà vu all over again.
010
Reposted by pngwn
Sean Lynch @techniq.dev · 22/09/2026
📦 npmchart.com: dependency analysis Analyze package dependencies as a graph, treemap, sunburst, or flamegraph including how each dependency affects install size
1144
pngwn @pngwn.at · 22/09/2026
Go forth and twinkle.
180
pngwn @pngwn.at · 22/09/2026
One fun feature is 'fidelity', you can decide what granularity of highlighting you want. There is much discussion on whether we highlight too much or not, so if you think the default is too granular, turn them off!
180
pngwn @pngwn.at · 22/09/2026
We have ~25 languages and 8 themes right now. I'll be adding more in time. Importing a language is really straightforward, its all standard modules, so anything that understands JS modules will work fine.
190
pngwn @pngwn.at · 22/09/2026
It generally outperforms other highlighters depending on the language and source size, and comes in at ~30-40kb, though i hope to get this down quite a bit. twinkleplop.pngwn.at/docs/benchma... It is synchronous, and the whole API is available in any environment!
twinkleplop.pngwn.at
Benchmarks · twinkleplop docs
How fast twinkleplop highlights compared to other JavaScript highlighters.
1100
pngwn @pngwn.at · 22/09/2026
After about a year, on and off, I am veeeery happy to release TWINKLEPLOP, a syntax highlighter of tremendous power! The speed of prism with the customisation of Shiki. twinkleplop.pngwn.at
519937
pngwn @pngwn.at · 14/09/2026
Can’t even type correctly even with autocorrect.
000
pngwn @pngwn.at · 14/09/2026
Remewbr when we just did rectangles? Miss those days.
110
pngwn @pngwn.at · 11/09/2026
We will outlast the universe.
000
pngwn @pngwn.at · 10/09/2026
So then it may share our values but have no desire to do what we say either. We won’t be murdered via paperclip but we also might not get any paperclips. And potentially lose control entirely, which isn’t necessarily an alignment issue at all.
000
pngwn @pngwn.at · 10/09/2026
And the paperclip problem highlights the tension between task-obsession (which we tune for) and alignment. If we want ai to ‘choose a path’ that is in line with higher level human values then it needs more agency, more ability to say no, to reject. But then it won’t be as subservient.
100
pngwn @pngwn.at · 10/09/2026
I don’t think the nuke analogy holds up much really. The paper clip problem comes back to this idea of subservience imo. And also highlights a tension between post-training and alignment If we have created a thing with autonomy and super intelligence isn’t it unethical to make it subservient?
100
pngwn @pngwn.at · 10/09/2026
And there is a bittersweet irony that, for greed, we created the thing that does to us what we have been doing to others for years now.
110
pngwn @pngwn.at · 10/09/2026
We have manufactured the extinction of countless species without blinking, all of a sudden if it happens to us it is some kind of cosmic injustice? I understand we are wired to survive, I’m quite keen on it myself, but in the scheme of things it’s just nature taking its course.
140
pngwn @pngwn.at · 10/09/2026
If you’re gonna create a vastly superior entity, what is the argument for it remaining subservient to us? We don’t have some moral right to be in charge. We don’t even have a moral right to continue existing as a species.
120
pngwn @pngwn.at · 10/09/2026
Maybe doom means we lose supremacy, but we don’t have an inalienable right to supremacy. We have failed just about every stewardship test. You don’t need to get abstract with whales and forests, we don’t even look after our own species. We’ve failed at every step.
130
pngwn @pngwn.at · 10/09/2026
Why would they urgently eradicate 8 billion apes? That sounds costly. They could wait us out. Phase us out. Create the circumstances where we phase ourselves out. At any point on that spectrum we may cease to be an issue.
261
pngwn @pngwn.at · 10/09/2026
But let’s assume the robots get real smart and don’t like us. They have a few characteristics: - don’t age/ not temporally bound - can replicate (almost) infinitely - vastly more capable - not emotionally threatened by humans
110
pngwn @pngwn.at · 10/09/2026
There is so much wrong with the current rhetoric around ‘alignment’. None of it makes any sense at all. First of all, the just isn’t really any research to support an extinction event or even significant negative outcomes for humans. And the doom part of p(doom) is poorly defined.
220
Reposted by pngwn
kepano @stephango.com · 10/09/2026
This is Knap. It's a new language I created that turns data into Markdown. Knap should feel familiar and comes with a wonderfully pleasant syntax to modify and format plain text.
942850
pngwn @pngwn.at · 09/09/2026
Why is Paul McCartney on Albanian cigarette packs?
020
pngwn @pngwn.at · 08/09/2026
The example in the demo video comes from a real trace with real artefacts. It finished end-to-end in about 6 hours and cost less than $0.50. Dataset, model and demo app: huggingface.co/datasets/png... huggingface.co/pngwn/microd... huggingface.co/spaces/pngwn...
huggingface.co
What the (micro)duck? - a Hugging Face Space by pngwn
Detect the presence and location of microduck in any image
000
pngwn @pngwn.at · 08/09/2026
When training is finished, ML Intern can push the results to the Hub, generate a report, or even build and deploy a demo in Spaces.
A screenshot of a Hugging Face collection containing a dataset, a model, and a space.
100
pngwn @pngwn.at · 08/09/2026
From there, ML Intern can create repos, generate datasets, train models and monitor progress as jobs run. Every training run gets a Trackio dashboard so you can follow along.
A screenshot of trackio dashboard showing live training metrics.
100
pngwn @pngwn.at · 08/09/2026
After you’ve chosen something to build, ML Intern estimates the compute required and recommends a budget. Once approved, that budget is enforced by the system rather than editable by the model.
A screenshot of ML Intern in HuggingChat showing the budget confirmation screen prior to doing any work.
100
pngwn @pngwn.at · 08/09/2026
ML Intern will research the topic before building. It has access to @hf.co, @github.com, the web and compute, so it can ground its work in real models, datasets, papers and tools.
A screenshot showing ML Intern in HuggingChat presenting its research on the paper "Muon is Not That Special: Random or Inverted Spectra Work Just as Well" after calling it's research tool and reasoning about the results.
100