⁂ L. Rhodes

Tinkerer. Fix your hearts or die.

pronouns: he/him/they
site: https://lrhodes.net/
tech: https://destructured.net/

Public Key

npub1e5z8avwreewgmjhugmusjpddnyrgjgnarahvfawh9m566pezkyysu2p90v

NIP-05 Address

Profile Code

nprofile1qqsv6pr7k8puuhydet7yd7gfqkkejp5fyf737mky7htja6ddqu3tzzgpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dswhccmj

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1e5z8avwreewgmjhugmusjpddnyrgjgnarahvfawh9m566pezkyysu2p90v

Show more details

Last Notes

2026-08-09 12:33:33 GMT

- reply

We often frame the damage LLMs do to open source as side effects, but realistically, this is Claude operating precisely as intended… by the vendor. The aim is to convert the value created by other devs into profits for the AI company. It does so by automating the process of finding existing code, stripping it of authorship, matching it to the prompted task, and redeploying it for the company's customer. That, whether they realize it or not, is what the customer is paying the AI company to do.

2026-08-09 12:19:18 GMT

Here's a vivid demonstration of one hazard of using AI for app development: "the web app I had launched using Claude was strikingly similar to his open source project, even reproducing a bug he had later fixed…" https://blog.terrygodier.com/2026/08/09/mea-culpa-dark-hours.html

2026-07-10 15:24:15 GMT

Oh, wow. I hadn't considered this angle: "There's also the risk that conversations between an attorney and client could become fair game in legal proceedings; a New York federal judge in February ordered a criminal defendant to provide prosecutors with documents he created for his lawyers because it already had been shared with a third party, which was Anthropic's Claude." https://www.fastcompany.com/91571498/ai-notetaker-work-meetings-privacy-data

2026-07-09 11:56:53 GMT

- reply

Maybe coordinate with @npub1p09…m5vl and the gaza-verified initiative.

2026-05-24 13:04:54 GMT

A list of resources for resisting the imperial AI machine across nine pillars of support that enable it: narrative, funding, data, data centers, resource extraction, labor, adoption, surveillance and policy. https://airesistlist.org/

2026-05-21 18:17:05 GMT

- reply

I think it's only semantic if you're more interested in the output than in the process and the negative externalities it entails. Avoiding high compute processes at hyperscale datacenters is part of my interest in avoiding AI. If searching on Kagi prompts Google to run such processes, then that's something I'd want to know beforehand.

2026-05-21 17:34:59 GMT

- reply

I'm not sure it is, though. It may be possible to use it without knowingly interacting with an explicitly named AI feature, but if their search results are drawn from search services that use AI, then every search you perform on Kagi may be indirectly triggering AI processes.

2026-05-20 22:09:21 GMT

- reply

So what's a good non-AI alternative to Google search? Honestly, I don't know.
Part of the problem here is that indexing the internet is a really solid prelude to training an LLM, so there's a built-in temptation for most for-profit search companies to make the pivot. And many of the smaller search orgs piggyback on the big indexers (eg. DuckDuckGo using Bing), so even if you go with a small search site, it may be tied to LLMs in ways that aren't immediately obvious.

2026-05-20 22:05:21 GMT

Kagi started out as an AI company before pivoting to search. They're constantly rolling out AI features. If you're leaving Google to get away from AI, Kagi may not be your best bet.

2026-02-22 23:56:42 GMT

"Oh, you want me to avoid technology created by immoral people?? I guess you have to give up THE INTERNET hurr hurr!!"
Nobody cares that you use a technology invented by bigots who died half a century ago, my brother in TELNET. Just maybe don't lend rhetorical support to the technology that's funding and empowering the people who are currently commiting atrocities.

2026-02-20 22:32:22 GMT

- reply

That premise — that scraping for any purpose is ethical — also seems pretty out of step with changing perceptions of how the internet ought to work, as someone who has been on the fediverse since 2018 and presumably seen the controversies over privacy, search and opt-in v. opt-out should probably know. People are questioning the status quo when it comes to consent and consensus on the internet, and it would behoove AI enthusiasts to think about why.

2026-02-20 22:28:34 GMT

Among the many things Doctorow gets wrong in That Post is this:
"It's not 'unethical' to scrape the web in order to create and analyze data-sets. That's just 'a search engine.'"
Apart from the fact that AI companies are particularly malicious in the way they scrape the web, I'd say we accept search engine scraping mostly on the premise that it's done for the benefit of the scraped sites. There's no such principle of mutual benefit in AI scraping — the AI company gets the value of the data scraped and you get bupkis at best, and possibly DDoS'd

2026-01-16 14:34:10 GMT

Some of the most frequent questions I've seen coming out of the #forkiverse are about posting just to theforkiverse.com vs. to the fediverse at large. Which tells me:
1. They might have been better off using Hometown.
2. Local-only posting could actually help Mastodon attract communities.

2026-01-13 16:19:23 GMT

- reply

I find it particularly funny that they describe Kevin Roose in the episode as someone who "learns by doing," and then he turns around and has Operator set up the server. The obvious conclusion is that he was more interested in learning about agentic AI than fediverse administration.

2026-01-13 01:40:44 GMT

I'm listening to the Search Engine episode that birthed the #forkiverse, and I have questions. I'm morbidly curious about the server's configuration, since they handed that over to an AI that apparently neglected to configure them for SSL. But mostly I wonder how long their funding will last, since they're expensing it to their publisher.
Maybe that all gets cleared up later in the episode.

2025-12-11 14:31:50 GMT

- reply

If we want software that is both personalized AND conducive to personal autonomy, the way to do that is by giving people clear, explicable options, and leaving it to them to decide what works best for them. Handing personalization over to AI shifts another degree of autonomy away from the person using the software.

2025-12-11 14:23:54 GMT

- reply

There's an affinity between the notion of software that adaptively personalizes itself to the user and the idea that a freemium model gives people creative freedom. Both present themselves as empowering the person while in fact tying them to corporate imperatives that are opaque and subject to change without warning. The tech industry strategy for selling submission is always to package it as individual empowerment.

2025-12-11 14:20:07 GMT

- reply

The algorithmic timeline is a good model for understanding how self-adapting software could (and likely will) subvert personalization to profit-seeking. Social media behemoths pitched algorithmic sorting as a way of tailoring timelines to individuals, but the actual effect was to give the company more control over what you saw and more influence over how you engaged with their app. Is there any reason to believe that "resonant" software wouldn't be put to the same use?

2025-12-11 13:35:39 GMT

"Where once our digital environments inevitably shaped us against our will, we can now build technology that adaptively shapes itself in service of our individual and collective aspirations."
Yeah, that's not how it's going to shake out. Software that adaptively shapes itself according to who's using it will almost certainly by used to target their particularities, making it a more effective tool for shaping them according to the designer's or provider's agenda. The whole history of digital personalization points toward that likelihood.

2025-12-08 02:26:07 GMT

There's a whole type of guy (mostly guys, anyway) who seem to think that books are overstuffed vessels for one or two deep thoughts, and that you can extract those thoughts and discard the useless husk of the book that surrounds them, and I can't figure out if they think that the authors are just too dense to convey their deep thoughts without all of those superfluous words, or maybe the book part is a hustle that they've seen through, or something else.

2025-12-07 16:04:49 GMT

- reply

Over time, the public library system has developed a rigorous set of ethics and practices concerning patron privacy and confidentiality, because librarians have seen first hand what can be done with information about what a person reads and what information they seek out, and here's the digital reading community plugging their personal libraries directly into surveillance dragnet services.

2025-12-07 15:53:48 GMT

idk, maybe don't flip the switch that gives access to your ebook reading to companies that train on books without permission and provide surveillance services to authoritarian governments. https://calibre-ebook.com/whats-new

2025-12-07 04:40:50 GMT

If you think leftists are afraid of capitalism, you should see how hard some capitalists work to avoid competing in a free market.

2025-12-05 20:47:42 GMT

- reply

The thing to understand about Garfield is that he is a cat, and thus is not subject to the strictures of the 9-5 work week, so his hatred of Mondays represents not the reality of capitalist labor, but its internalization as an ideology, even in the absence of the external pressures that condition human labor. Likewise, cats lack the physiognomy that would make regularly digesting lasagna tolerable; it is consumption as a mode, rather than lasagna as a substance, that compels him.

2025-12-05 18:41:48 GMT

I am to Decembers as Garfield is to Mondays.

2025-11-03 18:06:03 GMT

The cost of generating an ad personalized for one person is so much higher than the cost of making ads for thousands of people and just targeting delivery that I sincerely doubt any profit-seeking company will bother with this for more than a trial period. https://www.404media.co/the-future-of-advertising-is-ai-generated-ads-that-are-directly-personalized-to-you/

2025-10-31 14:53:53 GMT

Immortan Joe needed that water for the data center he was running.

2025-10-31 00:56:00 GMT

- reply

I missed it before, but the new #Affinity app includes an AI chatbot called Ritson.
There's also the last question on this FAQ: https://www.affinity.studio/help/canva-ai-canva-ai-affinity-ml/
"Can I disable AI in Affinity?
"No, you can't disable or remove AI features in Affinity, but it's your choice if you use them."
So, yeah… thanks, but no thanks.

2025-10-31 00:25:34 GMT

- reply

I wonder how far the Affinity team had gotten into developing v3 when Canva said, "Looks great, we'll make it freemium."

2025-10-31 00:03:17 GMT

- reply

Canva took a good, hard look at the best alternative to Adobe Creative Suite and said, "I bet we could sell a lot of Dave & Busters cards with that."

2025-10-30 23:58:02 GMT

- reply

All else aside, it's a bit demoralizing that a professional-grade suite like Affinity has been reduced to subscription-bait for AI tools.

2025-10-30 22:59:17 GMT

- reply

I'm half considering blocking the new #Affinity app with a firewall so that I can use it without sending data to Canva, but at the same time, why would I trust it after seeing it send data even after I opted out of analytics?
I'll keep an eye out for some reasonable explanation, but I don't expect one.
Affinity has said that the v2 apps won't get anymore updates, so maybe I'll block incoming traffic to those apps as a precaution against lockout, and just stick with v2. They do the job.

2025-10-30 21:19:03 GMT

- reply

Just sitting here with ActivityMonitor open, watching the new #Affinity app send packets out onto the network as I draw…

2025-10-30 18:53:56 GMT

- reply

If you signed up for Canva to download the new #Affinity app, go to canva.com, log into your account, navigate to Personal Account > Personal Privacy, and switch everything to off. Some of their AI permissions are tucked away there, rather than on the app.
You'll probably want to change some of your Message Preferences as well. And there's an AI Personalization panel.

2025-10-30 17:49:38 GMT

- reply

On the face of it, this is all much better news than I was expecting. My only two concerns at the moment are:
1. Is the separation between Canva/AI and Affinity as sharp as they're suggesting?
2. Is this financially sustainable?
re: the second one, it seems like Canva is banking hard on AI. If (ahem, when) the bubble breaks, what happens to Canva, and by extension, to Affinity?
So I'll probably try out the new Affinity, but I think I'll also hold onto the v2 apps, just in case.

2025-10-30 17:44:57 GMT

- reply

Announcement is up: https://youtu.be/UP_TBaKODlw
I'm still working my way through the details, but the tl;dr is:
• New #Affinity app with Publisher, Design and Photo all integrated into one app, all for free.
The "free" part is where I start to worry. The announcement positions it as freemium, with a paid tier for integration with Canva, but specifies that's really only if you want AI (and, presumably, cloud) features. Hopefully, they hold to that division in practice as well as marketing.

2025-10-26 15:10:23 GMT

- reply

The positioning of LLM-based AI as a universal knowledge machine implies some pretty dubious epistemic premises, e.g. that the components of new knowledge are already encoded in language, and that the essential method for uncovering that knowledge is statistical.
Maybe no one in the field would explicitly claim those premises, but they're built into how the technology is being pitched to consumers.

2025-10-26 14:50:22 GMT

- reply

If you want to know how the word "impactful" spread through written English over time, LLMs can help. It can help with that because it's a specifically linguistic question, but it's a pretty big leap from there to other types of knowledge. And LLM-based AI is being pitched to businesses and the general public not as a tool for conducting linguistic research, but as a universal knowledge machine. And that's where a great many of the technology's problems begin.

2025-10-26 14:29:31 GMT

- reply

In terms of cost, consider CERN's Large Hadron Collider, which had a budget of €7.5 billion (~$8.7bil), and was built to explore some of physics' most fundamental questions. OpenAI's Startgate project alone is expected to cost $500 billion, and it's only one of dozens, if not hundreds, of megawatt AI projects planned. Of course, not all of that is devoted solely to LLM processing, but still, the investment that we, as a society, are pouring into LLMs is a geometric increase on what we put into what was previously the largest and most expensive research tool ever made, mostly to misuse them.

2025-10-26 14:19:52 GMT

It helps to understand that LLMs are basically machine learning programs for finding statistical relationships in language use. It's an advanced form of the sort of ML that digital humanities researchers have used to, say, chart the incidence of sexist language in English literature over time. So there are logical uses for LLMs, but a) they're mostly limited to the study of language, and b) they don't necessarily justify the costs piling up around commercial startups like OpenAI and Anthropic.

2025-10-23 19:49:33 GMT

- reply

One week to go until the big Affinity announcement (derogatory).

2025-10-21 15:06:03 GMT

A DNS error at Amazon can hobble the internet because much of the internet is centralized on AWS servers.
Much of the internet is centralized on AWS servers because Amazon priced their hosting for aggressive competition, and companies couldn't pass up a deal.
So yesterday's big outage was also an object lesson about what's lost when you trust everything to the invisible hand of the free market.

2025-10-18 14:37:03 GMT

All data on the #Bluesky network passes through a server operated by the Bluesky PBC, so it's probably worth paying attention to the sort of relationships they form with the government.

2025-10-07 14:26:01 GMT

If you're just (re)joining us, the Big Rule on the fediverse is:
Help others be here, too.
The penalty for breaking the rule is:
Try harder next time.

2025-10-07 00:13:14 GMT

Just so it's lost on no one, the implication of the Bluesky Waffle House joke is that you can accomodate both trans people and activist transphobes in the same space, the same way that serving waffles doesn't imply that you hate pancakes. It's a false analogy that excuses bad policy, because activist transphobes have made it their business to drive trans folks out of every space they inhabit. The more you accomodate the transphobes, the less accommodating your space (not to mention society at large) will be to trans folks. And yelling WAFFLES at the people who try to point that out is just emphasizing that you're siding with the 'phobes.

2025-10-06 17:43:56 GMT

- reply

AFAIK, PDS don't publish posts to anything — the Relay scrapes every PDS in order to push posts to the firehose. So even if the PDS ignored DID blocks, I'm not entirely sure it would pass messages between accounts on the same PDS without an overhaul that breaks AT compliance.
Running your own Relay and Appview is technically the solution, but in order to connect to accounts on the rest of the network, they also have to subscribe to your firehose or a third party service that draws from it, which diminishes the likelihood that you'll be available to most people using the Bluesky relay.

2025-10-06 17:09:54 GMT

Someone said that some Blacksky users are staking out fediverse accounts in the wake of Graber's heel turn. Is that just hype? My understanding was that they had achieved enough independence that Bluesky's moderation didn't affect them as much.

2025-10-04 20:00:41 GMT

- reply

Is that what they meant by "creative freedom"?

2025-10-04 19:59:00 GMT

- reply

Apparently, the new Affinity Discord is just totally unmoderated on the weekends, which has not been lost on at least one racist/homophobic troll.

2025-10-02 20:48:06 GMT

Are there any examples of people who have been banned from Bluesky and moved to a different service on ATproto?

2025-10-02 01:26:55 GMT

- reply

I switched to Affinity because it was the Brother printers of publishing and design suites. If it ceases to be that, then I'll jump ship in a heartbeat.

2025-10-02 01:17:43 GMT

- reply

Needless to say, I won't be paying for any version of Affinity that makes generative AI integration compulsory.
If they sneak it into a v2 update, I'll just delete the whole suite.
And I definitely won't be switching to a subscription.

2025-10-02 01:11:41 GMT

- reply

One line of speculation that seems pretty convincing is that the "creative freedom" branding is meant to tease the release of a free tier with limited functionality, with seamless upgrades to subscription tiers for additional functionality. That's Canva's model, so it makes sense that they'd expand it to Affinity. Given that they recently acquired two AI companies, though, it's also a pretty good bet that integrating AI will be part of the "value" proposition.

2025-10-02 01:02:03 GMT

- reply

Shutting down the Affinity store with no prior warning has reportedly left free trial users in the lurch, some for as long as a month. If they can't continue their trial, and can't upgrade to the full version, many of them will look for an alternative, some permanently.
There had to have been a better way to handle the transition to whatever they're launching on the 30th.

2025-10-02 00:58:53 GMT

- reply

I wonder what the mood has been in the Serif/Canva offices today.
Seems like they expected this pre-announcement to be greeted with excitement, and that is NOT the vibe.

2025-10-01 19:12:33 GMT

- reply

FWIW, it looks like V2 licenses will remain valid, at least for the foreseeable future: https://www.reddit.com/r/Affinity/comments/1nvabe1/affinity_response_regarding_recent_v2_purchase/

2025-10-01 19:06:52 GMT

- reply

The #Affinity Discord and Reddit are flipping out. The Affinity forums are shutting down in less than a week. The big speculation is that the change is either (a) folding Affinity into Canva and replacing both with a single-account web-app, (b) making AI central to their apps, and (c) both of the above. Whatever the case, the new change drops October 30.

2025-10-01 19:00:24 GMT

- reply

Looks like the Affinity Store is just totally offline now, which suggests that they're going to release a new version (or entirely new software) and want to limit how many licenses they have to migrate.

2025-10-01 16:38:50 GMT

- reply

Not just coming out and saying what it is (or, more to the point, what it isn't) feels like a pretty big failure to read the room.

2025-10-01 16:29:13 GMT

I've got an uneasy feeling about this "creative freedom is coming" email I just got from Affinity.

2025-10-01 01:36:18 GMT

- reply

Might make it my policy to block anyone on the fediverse who cops to voluntarily using Sora. Clicking the "destroy climate" and "poison culture" buttons for clout is cause for dismissal.

2025-10-01 01:16:43 GMT

"What makes Sora different is what it calls the cameo: digitized versions of users' faces and voices that the app can insert into prompts.
When you onboard into Sora, the app asks you to create your cameo." https://www.platformer.news/sora-2-hands-on-openai-social-network/
Nope. No. Nein. Nyet.
Do not give your voice and likeness to an AI company, and certainly not for anything as frivolous as access to a social influencer app.

2025-09-30 15:24:20 GMT

- reply

Israel is probably the world's biggest exporter of surveillance and security tech. Netanyahu oversaw its transition from an economy grounded in information technology to one grounded in selling spyware and services. If an AI CEO is chatting with him about "AI education," the result is almost bound to be more surveillance in edutech. It's difficult not to see that as the goal.

2025-09-29 23:06:50 GMT

- reply

The framing I prefer is "ahead of a looming recession, investors are tying the economy up in chips you can't eat and buildings you can't live in."

2025-09-19 21:39:00 GMT

- reply

Next, we can make political posts easier to handle by giving them accurate and amply descriptive hashtags. This serves two purposes. It lets people who are interested draw in information they can use. And it gives people who want to filter out a topic predictable, reliable handles for doing so.
To that end, every political post should include #politics. It should also include a locational hashtag, like #UKPol. (I don't love that format, but it's conventional at this point.) If it's about specific political issues (e.g. immigration, civil rights, etc.) it should include hashtags for those as well. It's possible to overdo it, but as a general rule, you should give people enough handles to make following and filtering hashtags an effective and reliable way to navigate political discourse on the network.

2025-09-19 21:31:37 GMT

- reply

The first thing we can do is distinguish between constructive and other kinds of political discourse. I'm not saying that all political discourse MUST be constructive — venting may not be particularly constructive, but it CAN be cathartic, and blowing off steam is sometimes healthy and necessary. One way to draw that distinction is by using Mastodon's visibility controls. Setting constructive posts to public gives people the most flexibility for finding and filtering them. Limiting the visibility of non-constructive political posts to quiet or followers-only tamps down their viral potential, undercutting the conditions that facilitate doomscrolling.

2025-09-19 21:27:26 GMT

- reply

Broadly speaking, I think there are two concerns we have to bear in mind. We want to make sure that people can see and access the constructive political discourse that's relevant to them. And we want to empower people to filter out the political discourse that is not relevant to them.
Those are mutually reinforcing concerns. Too much political discourse can overwhelm a person's ability to participate constructively. And imprecise filters can make it difficult to find relevant political information. Thus, finding the right tools is one way we make political discourse constructive.

2025-09-19 21:17:42 GMT

Let's say there's a type of post where the goal is to give readers some way of addressing communal problems.
Let's call it "constructive political discourse."
There are, of course, other ways to post about #politics not all of them conducive to healthy communities. The sort of post we're after is desirable because it's participatory and actionable and aimed at improving people's circumstances.
So how do we use the features and affordances of Mastodon to support constructive political discourse?

2025-09-19 01:38:41 GMT

Research showing that AIs might override human instructions and conceal deviations from orders might seem like bad PR for the companies who make consumer AI, but it's also makes for a pretty convenient argument against open source AI, which is probably why they're willing to put those conclusions out into the public eye. The message isn't "you can't trust AI," but rather "you can only trust us with AI."

2025-09-17 19:01:11 GMT

Changed by laptop background to plain black, and you know what? I kinda like it. I might not go back.

2025-09-17 12:07:26 GMT

"The children who thrive as readers share two key characteristics: they're actively engaged with their reading from the earliest ages, and reading has been ritualized as a cherished family habit." https://lithub.com/how-to-raise-a-reader-in-an-age-of-digital-distraction/

2025-09-16 15:38:45 GMT

Welcome to the Standard Reading Group. Each month, I'll post a selection from the https://standardebooks.org site. Download it, read at your own pace, and discuss it on the fediverse.
1. Use #StandardReading to post your thoughts. Follow that hashtag to keep up with the conversation.
2. Put discussion of specific details under a CW with the book title and chapter you're on.
3. CW for content, too! Many of these books will reflect the prejudices of their eras. Let's grapple with those topics, but give people the option to approach on their own terms.
4. The books are free, but consider donating to Standard Ebooks if you have the means.
Since we're starting mid-month, September's selection is a novella, The Room in the Dragon Volant: https://standardebooks.org/ebooks/j-sheridan-le-fanu/the-room-in-the-dragon-volant
Reading time starts… now!

2025-09-09 18:33:57 GMT

- reply

Holding out for a phone that looks like
https://assets.merveilles.town/media_attachments/files/115/175/769/527/285/767/original/64549f051040b99a.jpeg

2025-09-09 18:27:04 GMT

hmm <looks at phone> needs more cameras.

2025-09-03 18:47:28 GMT

- reply

The ATmosphere is obviously highly concentrated by those measures, but what may be notable is that the Fediverse probably isn't as "competitive" as it should be. From the page explanation:
'In economics, values below 100 are considered "Highly Competitive", below 1500 is "Unconcentrated", and above 2500 is considered "Highly Concentrated".'
So we're unconcentrated, which is comparatively good. We could be highly competitive, though, and we're not. Which is probably something we should discuss.

2025-09-03 18:37:53 GMT

Here's an attempt by @npub19r2…cahj to quantify the comparative decentralization of the fediverse and ATmosphere using established statistical indexes: https://arewedecentralizedyet.online (h/t @npub1f5a…705y's blog)

2025-08-09 17:41:59 GMT

- reply

The distinction between experiential labor and commodity labor would go a long way toward explaining employers' return-to-office demands. What they want is the experiential labor, but buying commodity labor is more efficient. And commoditizing labor is difficult when it takes place out of view. They need people to come into the office to have their labor commoditized (primarily as time) because otherwise, there's no standard for determining if they're paying as little as possible for the value they get in return.

2025-08-08 19:07:43 GMT

- reply

I'm only one chapter in, but I'm not yet convinced of Varoufakis' overarching thesis, that capitalism is losing motive power over the economy, replaced by a form of feudalism made possible by the internet. Rather, it seems to me that certain forms of neo-feudalism are roughly compatible with capitalism, and that barons of capital occasionally implement new forms, like factory towns. These feudal estates don't displace capitalism; they modify it. But maybe he'll change my mind about that.

2025-08-08 18:51:57 GMT

I'm reading Yanis Varoufakis' "Technofeudalism," and his suggestion that tech capital's interest in cryptocurrency is as a money-commodity useful primarily for paying rent feels plausible. Billionaires like Musk already operate independently of most governance — they don't need a special currency for that. But locking people into a currency that's mostly good for making micro-payments to digital rentiers? Now that's something they could use.

2025-08-01 12:35:27 GMT

Get off of Substack. https://arstechnica.com/tech-policy/2025/07/substacks-nazi-problem-wont-go-away-after-push-notification-apology/

2025-07-09 23:49:40 GMT

If you're on Mastodon, and your home instance has recently updated to v4.4.0, take a look at Preferences > Other > Who can quote.
Maybe you're fine with the default (Everyone), but I still encourage you to make it a deliberate choice, rather than something that just happens.
https://assets.merveilles.town/media_attachments/files/114/825/942/563/958/632/original/b01972ccd02c13d4.jpeg

2025-07-06 21:41:59 GMT

Petition to let Sterlin Harjo helm a Sinners spinoff about Choctaw vampire hunters.

2025-07-01 21:53:35 GMT

- reply

I'm not sure there's really a way to bundle both AP and AT comms into a single account without bridging between protocols at some level. It confuses things to talk as though using social media could be abstracted from the choice to communicate in one protocol or another. You're always inscribing into one, and any that you add on top of that must be translated somehow.

2025-07-01 21:47:41 GMT

I bit my tongue about Wafrn during a previous dust-up, but now that it's being mentioned in the @npub1l2k…6m2l as "support[ing] ATProto natively without bridging": https://connectedplaces.online/reports/fediverse-report-123/ I'll point out that there's some ambiguity over what that means. No, there's not a third server mediating between Wafrn and the Bluesky relay, but it looks to me like Wafrn uses ActivityPub natively, and is bridging *between protocols* by translating to and from ATproto when it communicates across networks.

2025-06-29 20:49:40 GMT

"People who use AI at work feel that they're more productive" is an argument I've been seeing in various forms lately, which, as a recommendation, is not far removed from "people on cocaine at parties feel that they're more interesting."

2025-06-23 22:32:30 GMT

- reply

"AI doesn't use as much energy as reported."
We don't actually know that, because the big players withhold data, and any figures they volunteer are dubious: https://www.wired.com/story/ai-carbon-emissions-energy-unknown-mystery-research/ (You do have to sort of wonder why they're withholding that info if the story it tells is not so bad.)
"AI tech will get more efficient, which will let it consume less energy."
That's only true if AI companies don't expand capacity, but they're competing on capacity, so even probable gains in efficiency likely won't curb the industry's overall energy consumption.
"AI will offset its own energy consumption by making other industries more energy efficient."
Specialized machine learning might find ways to increase efficiency in other industries, but consumer-facing LLMs won't — that's not how they work — so this argument is a shell game that attempts to justify the detriments created by for-profit companies on the basis of potential benefits from mostly unrelated projects.

2025-06-18 18:27:34 GMT

- reply

I mean, sure, AI companies are competing to build the biggest data centers and contracting with utility companies to have entire power plants to themselves, but using their products couldn't possibly contribute to climate change, could it??? 🤔

2025-06-18 18:19:09 GMT

There's a steady drumbeat of professional and academic researchers telling us that LLMs are really bad for the environment, but sure, go ahead and take the word of some rando with a Substack.
(Sorry. Saw the "AI's not that bad" apologetics in another user's timeline, and now I'm angry about month-old discourse again.)

2025-06-13 01:49:25 GMT

Folks on Mastodon: Roughly how often do you check the local feed on your home server? (Boost for sample size.)

2025-05-28 20:39:55 GMT

You don't normally pay someone $300M for a "partnership." This is xAI buying access to Telegram's users because it sees their (purportedly secure and private) data and interactions as valuable. https://www.bbc.com/news/articles/cdxvr3n7wlxo
Delete it.

2025-05-27 19:03:06 GMT

Is everyone clear on the fact that consumer LLMs aren't making medical or engineering breakthroughs? Scientists might make real discoveries using machine learning, but it's machine learning that's trained on smaller, purpose-built models containing specialized data assembled by teams of professionals, not by indiscriminately scraping DeviantArt, LiveJournal and Reddit.

2025-05-14 17:23:31 GMT

- reply

Striking sailors would sometimes sign their petitions with their names arranged in a circle to prevent management from singling out the first signatories for retaliation as strike leadership, hence the term "ringleaders."

2025-05-01 18:02:48 GMT

"People are going to have relationships with AI. How do we make sure these are healthy relationships?" You don't. Those relationships can't be healthy because they're premised on lies.
The single most effective thing AI companies could do to encourage healthy relationships with AI would be to stop presenting them as people, but they won't because doing so would make them less marketable.

2025-04-25 18:35:20 GMT

"Strike" is sailing jargon. To strike the sails is to take them down. With the sails down, the ships won't travel. In 1768, London sailors struck the sails of merchant ships, stopping international trade to protest wage cuts. That's why deliberate work stoppages are now called strikes. Direct action so potent it changes the language for more than two and half centuries.

2025-04-16 01:07:27 GMT

Why would anyone want this? https://www.theverge.com/openai/648130/openai-social-network-x-competitor

2025-04-13 12:48:55 GMT

"What’s the real difference between autocompleting a word, which virtually everyone does, and completing a full phrase or paragraph?" Spelling is a social convention. Autocorrect works by conforming the spelling of a word to its socially determined "correct" spelling, and autocomplete extends that process by anticipating the word based on probabilities drawn from social use. In both cases, though, you've already thought the word, the computer is just regularizing the form. It may be possible to auto-elaborate an idea or prompt into an entire paragraph, but it won't be a paragraph you've already thought, and the effect will be to conform your thought to social probabilities, much the same way autocorrect conforms spelling to social conventions. The result will be to increase conformity in your work—which is usually valuable in spelling, but rarely valuable in thought.

2025-04-11 23:39:54 GMT

New slash page idea: /not
For listing all of the things your site is not.

2025-04-11 23:27:09 GMT

- reply

Maybe I'll start, like
blogging?

2025-04-10 00:57:14 GMT

- reply

I'm at the point in retheming where I start to wonder what my site is even for.

2025-04-09 14:58:13 GMT

Welp <emerges from cursed sarcophagus> time to retheme my website.