洪 民憙 (Hong Minhee)

@monaco_koukoku ありがとうございます！確認してみますね！

@monaco_koukoku はい、そちらにご投稿いただければ大丈夫です。ありがとうございます！

Fedifyってテスト用に相手のAPサーバーのモック的なのが使えるのね。署名やネットワークのことを気にせずにActivityの送受信をテストできると。素晴らしい。
https://fedify.dev/manual/test#mocking

@monaco_koukoku オブジェクトディスパッチャーを設定したのに、ctx.getObject()がnullを返すんですか？どうやらバグっぽいですね。よければ、イシュートラッカーに課題を作成していただけますか？🙏🏻

After much faffing about, I have implemented the dynamic #OAuth2 client creation for #GoActivityPub services using the Client ID Metadata Document[1] that's been proposed as a replacement(?) for RFC7591 (Dynamic Client Registration Protocol).

The changes are in both the Authorization service and in the BOX #ActivityPub client to server helper.

[1] https://datatracker.ietf.org/doc/draft-ietf-oauth-client-id-metadata-document/

#CIMD

來日(내일) 춥다는데 午前(오전)에 病院(병원)에 가야 한다… 가기 싫다…

@ianthetechie Yes! https://writings.hongminhee.org/feed.en.xml

브라우저 API가 모두 ‘웹’ API는 아님
------------------------------
- 웹 플랫폼은 *표준화된 API* 위에서 동일하게 동작한다는 인식이 널리 퍼져 있으나, 실제로는 *브라우저 벤더별 인프라* 에 의존하는 API가 다수 존재함
- Geolocation, Speech, Push, Payments, Passkeys 등은 표면적으로는 웹 표준이지만, 내부적으로는 *Google·Apple·Microsoft의 서비스* 를 호출
- 동일한…
------------------------------
https://news.hada.io/topic?id=25974&utm_source=googlechat&utm_medium=bot&utm_campaign=1834

So, I wrote a longer piece on this: https://writings.hongminhee.org/2026/01/ethics-of-small-actions/index.en.html.

이 主題(주제)와 關聯(관련)해서 조금 더 길게 글을 써 보았습니다.

https://writings.hongminhee.org/2026/01/ethics-of-small-actions/index.ko-hang-kr.html

오늘 발견한 거: 위키피디아에는 LLM이 작성한 걸로 의심되는 징후를 모아서 정리한 글이 있더라.. https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing

"AD8K, with Dolby Vision"

RE: https://hollo.social/@hongminhee/019bd64b-6794-79fc-9236-6703a3b4bc6e

Armin Ronacher, who even very recently has previously has been very pro ai agent vibecoding to the point of criticizing critics, seems to have stepped back and asked the question: are AI coding agent users all experiencing mass delusion... including himself? https://lucumr.pocoo.org/2026/1/18/agent-psychosis/

Conversation on lobste.rs is also interesting https://lobste.rs/s/0pxvyd/agent_psychosis_are_we_going_insane

I'm uncomfortable with using language that evokes slavery in the context of AI/LLMs. Actually, not just for AI/LLMs, but for any subject at all.

AI/LLM에 對(대)해 奴隸制(노례제)를 聯想(연상)시키는 表現(표현)을 쓰는 것에 拒否感(거부감)이 있다. 아니, AI/LLM이 아니라 어떤 對象(대상)에 對(대)해서든.

These days, my friends with ADHD and I have started calling our condition AD4K, or even AD8K, because our symptoms feel so severe. 😂

Dojang is a dotfiles manager I built a few years back and then basically abandoned. I'd accumulated enough friction points that I wasn't really using it myself anymore. Finally got around to addressing those issues and just shipped a new version: https://github.com/dahlia/dojang/releases/tag/0.2.0.

If you're curious about Dojang, the docs are here: https://dojang.dev/.

I know I made this project, but I have no idea why people are starring the repository for this Markdown formatter that I created solely for myself. 😂

https://github.com/dahlia/hongdown

Sometimes when I see programming experts trying to help new learners I imagine what would happen if we were a driving school:

"The gas pedal makes the car go, right?"

"Well, only if the engine is started, and not if you are out of gas, and not if the car is in neutral, and not if the car is parked against a wall, and not if the parking brake is on, and it shouldn't be called a gas pedal in an electric car, and, and, and...."

@celesteh @jnkrtech I think it's partly because JavaScript is more commonly used for building consumer products compared to other languages, so there's a higher proportion of developers focused on shipping products rather than diving deep into infrastructure. When you're building a product, you naturally extract just enough to solve your immediate problem and move on. Languages like Rust tend to attract more developers interested in tooling and infrastructure work, where yak shaving is almost expected. The ecosystem reflects the priorities of its community.

JavaScript libraries are like, here are five libraries that each do 18% of what you need. They are mutually incompatible.

Weekend thoughts on Gas Town, Beads, slop AI browsers, and AI-generated PRs flooding overwhelmed maintainers. I don't think we're ready for our new powers we're wielding. https://lucumr.pocoo.org/2026/1/18/agent-psychosis/

I started Ghost of Yōtei today, and just like its predecessor Ghost of Tsushima, the visuals are absolutely stunning!

@ianthetechie Yeah, I think Haskell is my favorite language too!

I'm coding in Haskell for the first time in a while, and the build times feel so long. I guess I've gotten too used to the instant changes from only working with TypeScript lately.

ActivityPub standardization is back! The #W3C just announced a new Working Group to evolve #ActivityPub and #ActivityStream and related standards! This is extremely good news. It's not like all the world's federated social media protocol problems were already solved! https://lists.w3.org/Archives/Public/public-socialweb/2026Jan/0000.html

AI企業がF/OSSコードでLLMを訓練することを止めるのではなく、訓練したモデルを公開するよう要求すべきだと思います。

撤退ではなく、再専有。GPLがそうだったように。

訓練コピーレフトについて書きました：「F/OSSの唯物史観——LLMを拒絶するのではなく、取り戻すべきだ」

F/OSS 史唯: 우리는 LLM을 거부할 게 아니라 되찾아 와야 한다
------------------------------
#### F/OSS를 LLM 훈련에서 막을 게 아니라, 훈련 결과 모델을 해방시켜야 한다는 주장

* 최근 〈자유·오픈 소스 소프트웨어와 LLM 학습에 관해〉(On FLOSS and training LLMs) 글이 F/OSS 커뮤니티의 좌절감을 잘 표현 — AI 기업의 …
------------------------------
https://news.hada.io/topic?id=25879&utm_source=googlechat&utm_medium=bot&utm_campaign=1834

Been thinking a lot about @algernon's recent post on FLOSS and LLM training. The frustration with AI companies is spot on, but I wonder if there's a different strategic path. Instead of withdrawal, what if this is our GPL moment for AI—a chance to evolve copyleft to cover training? Tried to work through the idea here: Histomat of F/OSS: We should reclaim LLMs, not reject them.

언제부터인가 才致(재치) 있게 말을 하려는 欲心(욕심)을 버렸는데, 普通(보통) 그러려다가 嫌惡(혐오) 發言(발언)을 하게 되는 것 같아서…