@RosaCtrl Agreed. A good language forces you to ensure you know what it is you're trying to say, and that's key to programming languages. Using a verbose, ambiguous and vastly redundant natural language makes that really difficult. It may make you think, but not so much about the program, more about how to say something unambiguous in a tool that was often adapted to do the opposite.

Anyone who's tried to express logic or anything complex in English surely knows how hard it makes it. #LLMs

@RosaCtrl Agreed. A good language forces you to ensure you know what it is you're trying to say, and that's key to programming languages. Using a verbose, ambiguous and vastly redundant natural language makes that really difficult. It may make you think, but not so much about the program, more about how to say something unambiguous in a tool that was often adapted to do the opposite.

Anyone who's tried to express logic or anything complex in English surely knows how hard it makes it. #LLMs

Nightmare: Why #LLMs kill author's skills: "What began as a jagged, precise Romanesque structure of stone is eroded into a polished, Baroque plastic shell: it looks "clean" to the casual eye, but its structural integrity – its "ciccia" – has been ablated to favor a hollow, frictionless aesthetic. We are building a world on a hollowed-out syntax that has suffered semantic ablation" https://www.theregister.com/2026/02/16/semantic_ablation_ai_writing/

#authors #writing #writerslife #editing #writersCommunity #LLM #genAI #semantics #language

New data from amplifying.ai shows Claude Code defaults to Railway for deployment, hand-rolls auth from scratch, and never picks AWS/GCP as a primary choice. For government projects with mandated platforms and compliance requirements, those defaults are wrong. That's what AGENTS files are for.

https://amplifying.ai/research/claude-code-picks

#AI #ChatGPT #Claude #ClaudeCode #LLMs

Nightmare: Why #LLMs kill author's skills: "What began as a jagged, precise Romanesque structure of stone is eroded into a polished, Baroque plastic shell: it looks "clean" to the casual eye, but its structural integrity – its "ciccia" – has been ablated to favor a hollow, frictionless aesthetic. We are building a world on a hollowed-out syntax that has suffered semantic ablation" https://www.theregister.com/2026/02/16/semantic_ablation_ai_writing/

#authors #writing #writerslife #editing #writersCommunity #LLM #genAI #semantics #language

Large-scale online deanonymization with LLMs (using HN posts)

https://arxiv.org/abs/2602.16800

#HackerNews #LargeScaleDeanonymization #LLMs #HNPosts #PrivacyResearch #OnlineSafety #AIethics

LLM=True

https://blog.codemine.be/posts/2026/20260222-be-quiet/

#HackerNews #LLM #True #LLMs #ArtificialIntelligence #MachineLearning #TechNews

Green data centers
cannot bend Climate physics.
Growth can't be sustained.

#haiku #senryu #poem #ShortPoem #SmallPoem #SmallPoems #Climate #ClimateDenial #denial #green #energy #GreenEnergy #AI #LLMs

Green data centers
cannot bend Climate physics.
Growth can't be sustained.

#haiku #senryu #poem #ShortPoem #SmallPoem #SmallPoems #Climate #ClimateDenial #denial #green #energy #GreenEnergy #AI #LLMs

I'm looking to expand on how it is not the #AI that is really driving the mad rush to build data centers. But what I am not is a grid expert, or even a data center expert though I have observed the latter since the 1990s. What makes more sense to me is the death of Moore's Law turning compute from an exponentially expanding commodity into something additive. It will be centralized now because it CAN BE. #LLMs are just a parlor trick to get capital to concentrate compute and rent it out. No?

New benchmark data on AI agent Skills: human-curated instruction sets improved performance by 16 points on average.

When AI coping agents generated their own procedural knowledge instead? Essentially zero benefit.

The domains with the biggest gains were the most specialized — exactly where government legacy systems sit.

Human experts aren't just essential for verifying specs. They're also the ones who have to write the Skills.

https://spec-ops.ai/blog/posts/skills-research/

#AI #Claude #ChatGPT #LLMs #Codex

I'm looking to expand on how it is not the #AI that is really driving the mad rush to build data centers. But what I am not is a grid expert, or even a data center expert though I have observed the latter since the 1990s. What makes more sense to me is the death of Moore's Law turning compute from an exponentially expanding commodity into something additive. It will be centralized now because it CAN BE. #LLMs are just a parlor trick to get capital to concentrate compute and rent it out. No?

I've always managed my tasks in #orgmode, but I kept losing track of the work I was handing off to #LLMs. I added the rules below to the top of my `CLAUDE.md`, and now Claude Code automatically logs its own work directly into my org files.

https://gist.github.com/typester/92eccd38b5bc1179a9aed0b42d8d6295

It's been super handy for me, so I'm sharing the instructions! The AI's work is recorded as tagged subtasks, so filtering them is really easy.

#emacs #productivity #claudecode

I think we need to stop blaming #ai and #LLMs for the training the humans involved are responsible for.

When it comes to training data, humans select it.
When it comes to the tokenization, humans are responsible.
When it comes to the weights, humans are responsible.

We can blame what they call 'ai', but that makes the responsibility disappear in the mist.

Name the corporation.

I think we need to stop blaming #ai and #LLMs for the training the humans involved are responsible for.

When it comes to training data, humans select it.
When it comes to the tokenization, humans are responsible.
When it comes to the weights, humans are responsible.

We can blame what they call 'ai', but that makes the responsibility disappear in the mist.

Name the corporation.

"A.I. is clearly not technology that is being universally encouraged as inevitable. Corporations often report that, so far, it does not seem to do much."

https://www.nytimes.com/2026/02/21/technology/ai-boom-backlash.html?unlocked_article_code=1.OVA.Ynqq.XAB-lNTc16MB&smid=url-share

#news #TechNews #technology #AI #GenAI #LLMs

"A.I. is clearly not technology that is being universally encouraged as inevitable. Corporations often report that, so far, it does not seem to do much."

https://www.nytimes.com/2026/02/21/technology/ai-boom-backlash.html?unlocked_article_code=1.OVA.Ynqq.XAB-lNTc16MB&smid=url-share

#news #TechNews #technology #AI #GenAI #LLMs