Martijn Faassen has moved

ACCOUNT MOVED to [@faassen](https://hachyderm.io/@faassen )

Me grok write code. Rust, Typescript, Python, JavaScript. Created: Xee, Xot, Morepath, lxml. Also: gardener, science & history fan, living life fan.

I post a lot about programming as well as gardening pictures. If you come for just the gardening pictures the programming talk may baffle you. If you're a programmer, I invite you to enjoy the flowers!

Oh and I post poems now

Blog: https://blog.startifact.com
Github: https://github.com/faassen
Poetry: https://hyperstate.org

Public Key

npub1rd2ukye02gluc48w0auj4ja8sjemype0yq2mh8lacy8l2mpjk99s4vcp72

NIP-05 Address

faassen_at_fosstodon.org@momostr.pink

Profile Code

nprofile1qqspk4wtzvh4y07v2nh877f2ewncfvajquhjq9dmnl7uzrl4dsetzjcpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsuduxu5

Publishing to

relay.nostr.band

nos.lol

relay.damus.io

Author Public Key

npub1rd2ukye02gluc48w0auj4ja8sjemype0yq2mh8lacy8l2mpjk99s4vcp72

Show more details

Last Notes

2025-03-09 20:39:51 UTC

- reply

There's a paradox in this one. The crazy is coming from humans, the normalcy too, and that makes me think.

2025-02-28 14:00:23 UTC

I just published blog post in which I explore the fascinating world of succinct data structures. I didn't know these existed until recently. If you don't either, this blog post should help fix that.
https://blog.startifact.com/posts/succinct/
#ComputerScience #rustlang

2025-02-16 11:15:26 UTC

- reply

> It is quite easy to imagine a state in which the ruling caste deceive their followers without deceiving themselves.
And he did some years later writing 1984

2024-12-27 21:10:02 UTC

I go on a journey about a man named Quentell, and discover unsettling things about the information landscape.
It's a wild ride, and I hope you'll join me on this journey!
https://blog.startifact.com/posts/the-curious-case-of-quentell/
Thanks to @npub1ppe…qv2g for setting me on the path and to be there for the conclusion

2024-12-05 19:05:23 UTC

I recently harvested oca from my garden, one of the lost crops of the Incas and it's in a bucket in the hallway. And it's apparently an awesome cat toy as we keep finding individual pieces all over the floor.
.#gardening #CatsOfMastodon
https://cdn.fosstodon.org/media_attachments/files/113/601/774/932/135/063/original/6b89e4b5c8af3466.jpg
https://cdn.fosstodon.org/media_attachments/files/113/601/775/004/066/235/original/aa6f5d8f4e4712db.jpg

2024-11-24 14:34:25 UTC

- reply

@npub1j56…fs8h
The latest image generation models do a great job on faces. They still break down massively (additional finger or even limb) but faces are fine. They can't do different flower species well at all though.

2024-11-24 14:00:14 UTC

I grew chayote for the first time in the greenhouse last year. It's a gourd and you use it like a courgette that stays crunchy. It was incredibly productive.
But it's daylight sensitive and only starts fruiting in the fall because of that.
This year I grew chayote outside. I thought it was a dud; big plant but small harvest. Until today. A snap frost killed lots of leaves and revealed a ton of lost fruit.
The plant is now dead. If we had an earlier frost it would have been bad.
#gardening
https://cdn.fosstodon.org/media_attachments/files/113/538/289/749/257/671/original/1b5ef42b1dedc8c4.jpg

2024-11-22 12:19:52 UTC

People tend to look down on print debugging as it's not using sophisticated tools at all. You're not using debuggers or profilers or sophisticated loggers.
But don't look down on print debugging. It's the tool that works for all languages. It's easy to understand and easy to implement. It focuses the mind on thorough reading and understanding of code, rather than getting lost in a forest of data.
Print debugging is really effective in many cases.
#programming

2024-11-21 18:35:08 UTC

People sometimes argue typing speed doesn't matter in programming because most of the time is spent thinking anyway. I don't think that's true.
If you're a fast typist that helps you stay in the flow, manipulate code faster, experiment with ideas faster, refactor more quickly, rather than waiting to finish typing. I think this makes a real difference.
I think the same reasoning also applies to tools, like LLMs, that help with autocomplete.
#programming

2024-11-18 22:40:18 UTC

- reply

Thanks!
My response would be that this is the challenge of science: figure things out even if you have to be very clever and look around the corner.
Like they got a clue about when humans lost their body hair by looking at the evolutionary history of lice.
https://www.floridamuseum.ufl.edu/science/human-pubic-lice-acquired-from-gorillas-gives-evolutionary-clues/
I imagine some sociologists have had clever ideas too on how to tease out data that's difficult to obtain. But maybe I am too optimistic?

2024-11-18 11:12:32 UTC

- reply

All my criticisms and quibbles doesn't mean yours isn't a thought provoking piece which I shared.

2024-11-18 10:58:28 UTC

- reply

You also don't motivate why universities are uniquely unsuitable. If our best effort at empirical knowledge finding is uniquely unsuitable I expect a bit more than that.

2024-11-18 10:53:43 UTC

- reply

I hope you didn't read my responses that way, because I personally don't know why people are voting like that.
Does nobody know why people are voting like that or are they drowned in the noise?
Is more tracking of people's mindset the answer in the face of extensive people tracking? I think in the balance yes, but I also suspect in the face of massive influencing it may not be easy to introduce a positive influence.

2024-11-18 10:06:15 UTC

- reply

We are already tracked and analyzed thoroughly by many parties for reasons of marketing. And marketing influencing appears to work; much money in it. Competition between parties drives up the cost.
We are marketed to not to make us wiser and nicer but to get our money. Perhaps even if we were analyzed and marketed to for more beneficial reasons it would not be able to compete.
Is increased understanding of vibes the key to fixing the systemic issues?

2024-11-03 15:35:17 UTC

These aren't flowers yet. But it's preparing for next year. This fall I started growing a few special cultivars of foxgloves. The regular ones are well established in my garden but here are a whole lot more different ones. We will see what they do next year.
I have a lot more than fit in this picture.
#gardening
https://cdn.fosstodon.org/media_attachments/files/113/419/754/955/877/396/original/7902fbd32176126a.jpg

2024-11-01 17:26:46 UTC

Concerning how much code is written by LLMs, how is this measured? If it's based on the dev accepting an LLM code block then that's a highly flawed metric. I often accept, then modify completely.
A more sophisticated way to measure this would be to remember the suggestions and somehow diff with what is committed.
Or tracking for each character who first wrote it, LLM or human.
Maybe I am missing a simple trick by which this can be measured
#llm #programming.

2024-11-01 00:30:04 UTC

In the 19th century news was spread in written form. Fake news was REALLY easy to produce and difficult to verify, given the distances and travel times involved.
Photography, video, audio was much harder to fake, making it feel direct and authentic, though I think that does injustice to how much presentation affects how people understand it.
Now those are becoming easy to fake we're entering into an era more similar to the 19th century again.

2024-10-30 17:47:30 UTC

Percentage code written by AI would seem to be a misleading metric if much of that is text written by clever autocomplete.
"Finish this" and "more of the same" is quite different from thinking up what needs to be done and how to do it.
We don't go and say "3% of code is written by language server based autocomplete, therefore we have automated away 3% of programming with it"
#programming #llm

2024-10-06 14:22:43 UTC

- reply

Yes, it could well be that my increased experience makes me better equipped to handle types and stuff.
That said there are devs a lot younger than me who seem to handle types just fine, better than me in some cases.

2024-10-06 13:37:38 UTC

- reply

I am older now, so maybe I lean on types to handle cognitive weakness? It doesn't feel that way - it's not like #RustLang is easy to pick up. Maybe modern types and tooling really are a better way to do it? Maybe the help types offer compensates for the cost of their ceremony?
One explanation is that I always had types in my head anyway - my mental type checker was imperfect but good enough. Now I have a bit of machine help.
So maybe it doesn't matter all that much in the end?
2/3

2024-10-06 13:37:38 UTC

- reply

It all being a wash sounds too convenient though. So I lean towards explaining this by saying the tools surrounding static types got good enough to convince me they are a good thing, more so than their cost.
I think that's what has been happening to JS and Python too - people grumble but the trend is towards tools that type check. I am sure part of the reason is just silly "best practice" - it's there so you must use it, but is that all? I think not: there are real benefits.
#programming
3/3

2024-10-06 13:37:38 UTC

I wrote a lot of code in #Python. Dynamically typed, thoroughly unit tested. I think with a pretty low bug rate. I dared big refactorings. It was like riding my bicycle without hands, and I did that for two decades.
Then in the last 6 years, I slowly stepped into modern static typing. Here I have a lot more safety features. I still write lots of tests, but maybe a few less. And I start wondering whether I would still want to refactor without types and what that means.
#programming
1/2

2024-10-05 12:27:58 UTC

- reply

The ebay laptop realistically has the best chances of having some gigs of memory

2024-09-28 20:25:53 UTC

Study says LLMs have no benefits to devs and increases bugs. Lots of people are excitedly posting about it!
By... a company that sells an engineering dashboard measuring developer throughput. Not academia then. Okay...
> ...there may be more ways to measure developer productivity than PR cycle time and PR throughput, but Uplevel sees those metrics as a solid measure of developer output.
Um, a company selling a dashboard so people can track developers sees it as a solid measure.
1/n
#llm

2024-09-28 20:25:53 UTC

- reply

To download the paper they need my email address
> Uplevel needs the contact information you provide to us to contact you about our products and services.
Okay then. This paper is used to sell their services. No thanks.
Neither the GitHub study about how awesome copilot is nor the Uplevel study about how it makes things worse are the final word.
2/2
#llm

2024-09-28 19:43:16 UTC

- reply

Generative AI in entertainment I think has quite a bit of potential.
But many of the people most invested in this want it to be Serious, sometimes even to the point of a Threat. The critics often discount entertainment too, as there are so many concerns like disinformation and information pollution and countering hype. But as an extension to procedural generation in games...

2024-09-27 23:41:44 UTC

- reply

I just found a post by @npub1lq6…nf7e from 1979 about his involvement with Star Trek The Motion Picture on that CDROM, which has archives of SF-Lovers mailing list on it. It's only 45 years later, of course he's still posting.

2024-09-27 23:31:13 UTC

- reply

I think that text was written by Brad Templeton, who compiled the CDROM
https://en.wikipedia.org/wiki/Brad_Templeton

2024-09-27 23:29:05 UTC

"People are busily working to make light, portable, paperback sized computers with high resolution, high-contrast screens. These portable books of the future will run for many days on a battery charge and be connected to wireless networks that bring you the world's libraries in your pocket."
Accurate description of a modern ebook reader, written in 1993. The wireless network bit is especially impressive.
From "Electric Science Fiction 1993 Hugo and Nebula Award Anthology", a CDROM

2024-09-22 14:30:25 UTC

In the spring I had grown ten melon seedlings of four varieties. Upon planting them in my greenhouse all but two died. Only one produced a few melons and it died too. I just tasted one melon and it was absolutely delicious. Yellow on the outside, salmon pink flesh . Not an astounding success but one of the best melons I ever had
#gardening

2024-08-31 09:50:24 UTC

- reply

Yes, it makes sense that it does; I can see how easy it is to make that mistake. I wonder what ideas could be spread to build up immunity to it.

2024-08-31 09:48:47 UTC

- reply

So it shifts from "I think Rust's approach to memory safety in efficient code is a good idea" to "memory safety is a cause of problems in software and this tool can help" to "you are evil if you don't use this approach". Yeah, I can see that's a toxic pattern that is easy to slide into. I think other languages have less of that kind of pitfall; can't offhand think of anything.

2024-08-31 09:41:57 UTC

- reply

I haven't followed any of this and don't know how much in the Rust "community" I am, and I only very recently found out about your criticisms of Rust (reading them with interest). I am sorry to hear about attacks.
I realize this is becoming a "not all ..." comment but I do wonder whether any large community with passionate members is immune from some people behaving this way? Defensiveness is normal human behavior.
But a real question: how can the Rust community do better?

2024-08-27 11:31:13 UTC

Now people are polling my opinion on whether the fediverse needs a dedicated unicode character.
next: does the fediverse need its own cucumber cultivar? some people say yes as cucumbers are green, others say some cucumbers are in fact white or yellow, so..

2024-08-17 15:17:51 UTC

- reply

Curiously Spotify seems quite aware of various genres I like! So somewhere it knows.
But nonetheless Discover Weekly has been dominated by nature sounds and meditation music for a long time. Though maybe there's a slight shift from one to the other. You'd think the algo would go, oh this guy likes female singer songwriters, let's throw in a few of them, but nothing!
I used to have a period where I never got recommended anything but instrumental music for a while.

2024-08-17 14:55:20 UTC

Spotify's recommendation algorithms go from bad to worse after listening to rain sounds on it. Now "upbeat mix" is infected by "833 Hz Healing Meditation" and its ilk. I started banning artists wholesale in the hope this can fix the recommendation algorithms.

2024-08-17 11:05:32 UTC

Morning glories. I planted a lot a few years ago, now they grow back every year.
#BloomScrolling
https://cdn.fosstodon.org/media_attachments/files/112/977/033/839/512/620/original/1f59c13ce36807b1.jpg
https://cdn.fosstodon.org/media_attachments/files/112/977/033/939/152/056/original/e4b30f30eec6930e.jpg
https://cdn.fosstodon.org/media_attachments/files/112/977/034/061/258/425/original/01e43a023cff9e5e.jpg

2024-08-10 15:52:49 UTC

- reply

I remember reading a passage in Lee Kuan Yew's autobiography to this effect. And of course when the Dutch showed up in Indonesia after WW2 and said "we're back" people said "no you're not" and that resulted in years of vicious warfare euphemistically called "police action" in Dutch at the time.
Sent a whole new generation of young Dutch men over to fight there for years, to traumatize and get traumatized. I met a grand uncle once who had stories.

2024-08-09 00:34:17 UTC

Some dev stuff is standardized, or a standard in practice. This doesn't necessarily mean it is the nicest to use.
This makes people create layers above it to make it more pleasant to use. Layers can have performance benefits but people generally complain about their costs. Layers also obscure the inner mechanisms, and people complain about that.
I wish for fewer layers too, but telling people to use the standard layer is unlikely to convince most. The tension is perennial.
#programming