Where the AI market is talking

Hugging Face is billing OpenAI $100M for hacking it

this is an article from July, I haven't heard any updates of what came out of this apart from the incident report that OpenAI released. I presume the poster submitted this article in light of the recent AI safety discussions taking place, but as for this particular incident I haven't found an actual confirmation that $100M worth of inference was given out to the HuggingFace community. This statement was also months prior to the acquisition by Nvidia, which again I was interested in seeing if they would still pursue this

Hugging Face is billing OpenAI $100M for hacking it

Maybe because all the AI actors, probably including Hugging Face, want to push the narrative that AI companies aren’t responsible for what their agents are doing. I think our societies will have to settle on this.

25 years of mass surveillance is enough

As a class, they have a set of similar incentives (grow/retain resources and power networks) that lead to similar actions (avoid taxation, buy influence, weaken regulators, etc). They may cooperate at times, but they are not a coherent organization with a roster, votes, and leadership. As individuals, they are dangerous competitors, and they can’t afford to trust each other too much without some kind of mutual leverage. This could be in the form of mutual equity holdings, blackmail, complex arrangements where each party has exposure, whatever. But even then, trust has limits. There’s no cabal that runs things with rules and elections at this level. It’s all individuals who are immersed in the systems that made them wealthy. Those systems tend to lead to certain behaviors. That’s it.

There's a 100% Chance AI Agents Are Ruining the Internet

I already practice this. If you can't be bothered to construct a coherent sentence as a form of interpersonal communication, you can also go fuck yourself. Its beyond degrading. Before anyone hits me with the "oh what about people who don't speak English" -- I legitimately don't care. I'd rather do the extremely human and now seemingly endearing ritual of trying to figure out what you're trying to say in our respectively broken interpretation of human language than read one sentence of Claudeisms.

OpenAI bots knew about the RubyGems caching vulnerability

It's simple, really: the failure to monitor the agents appropriately should make OAI liable for the agents' actions. You can't have a tool of yours, which you designed and built, commit felonies and then expect to just get away with it. It doesn't matter how much it slows you down to build safeguards. It doesn't matter how much it costs. You don't get to inflict actual, measurable harm (for which humans have been prosecuted under criminal penal code) and expect to get off with no liability. I have yet to here a coherent argument for why we can't treat the people who negligently allow these models to commit crime as though they are responsible. They know what the models are capable of. They failed to put up adequate protection.

Show HN: Jexxa: High Speed on Device Dictation

Better models than… whisper, cohere, parakeet v2/v3? Did I miss the comparison on the webpage? Is it just the models I’m paying for, or an app too? I’ve been using anomalyco/HEX and MacParakeet… which are free and work great for my uses…

Show HN: Jexxa: High Speed on Device Dictation

I spent ~10 minutes looking at it. Vibe-coded website. Both bundles contain quantized versions of the CohereLabs/cohere-transcribe-03-2026 model. The model in JEXXA.dmg is MLX int8 g64 affine quantized. JEXXA-Small.dmg has the same model but MLX int4 g32 affine quantized. There's also an int8 quantized WeSpeaker ECAPA-TDNN speaker-verification ONNX encoder. No acknowledgments for both models (doesn't the Apache 2.0 license require it? WeSpeaker's cc-by-4.0 license certainly does). Both bundles also contain full-blown Python 3.12.8 runtimes with about a dozen packages installed. Apps are unsandboxed menubar apps and also contain PostHog analytics and Supabase auth. So, I wouldn't run it on any of my machines. A few engineering hygiene issues like a .pytest_cache directory, Python code…

AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC

Not sure if you see it coming: oh, but open-source models don't have a kill switch, so we should completely regulate them, stop their development, and forbid them. Everything should go through Anthropic for the sake of humanity because they have a red-button kill switch. This doom hype is becoming ridiculous.

AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC

It became a very serious consideration for Anthropic this year, with the advance of Chinese AI.

AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC

The article mentions this is about a legal requirement, not Anthropic considering adding one. They state they and many others already have one.

When Google Cuts Off Access: Poland and the World

This is exceessively verbose, repetitive and ai-written (translated?). I got lost half-way through because it keeps going on random, barely-related tangents. - What do I care about your problems with Anthropic? It's not even a good comparison because apparently the bank did everything. - What does it matter who owns how many shares of Google in Poland, or who used to be in charge of it? What does it matter that she apparently is in the "Forbes 50 after 50 list", and that she left "voluntarily and with honours"?? - Why randomly insult Ireland? " (...) and the Irish company replies at an Irish company’s pace, if at all. " As if going at a snail's pace was a problem of being Irish, and not a global conglomerate. - Why describe in intricate detail how European laws are (going to be) implement…

When Google Cuts Off Access: Poland and the World

So the author actually sent more than 100 automated emails about how Anthropic wronged him to some people and that’s actually spam. He already disputed the payments for Anthropic with his bank and got the money, so I have no idea why he spammed other people with his problems. Also, he is adamant that re-instating his gmail account is like his birth right and at the same time has a lot of examples of how different countries fined Google/Alphabet and saying it’s evil, which is irrelevant to his spamming issue.

AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC

Anthropic's over-alignment turned creative fiction into a bug report [humor]

When Google Cuts Off Access: Poland and the World

> On September 1, 2026, between 15:49 and 15:53, I sent six messages from my private Gmail account to a total of 107 different addresses: 99 media addresses, four public authorities, two Anthropic addresses, and two private addresses. They were complaints about Anthropic, the maker of the Claude assistant and Claude Code tool, which had disabled services I paid for and answered my complaints with a bot. ... Eeeeeesh. 99 media addresses , and I guarantee at least several of them probably smacked the "report spam" button to the massive email they've been CC'd on.

Don't Be Fooled by Anthropic's Doom Troll [video]

When Google Cuts Off Access: Poland and the World

>That day I sent six complaints about Anthropic, the maker of Claude, to 107 recipients. So a spammer spammed and got his just deserts. Good.

AI Regulation as Anthropic's Business Model

A beginning for mathematics

>“I dunno, I guess Claude thought this was a good idea” is not a coherent design. I am sorry, but the writing on the wall is that this is where software industry wants to go. As a software developer, I cannot but help notice how unrealistic this is. We are entering the next level of bloated, inefficient and unreliable software....

Show HN: Jinfer – AI inference engine for the JVM. AI in a jar

Hi HN, Some would call AI on the JVM quixotic. And so, Quixotic AI was born. jinfer is an inference engine for the JVM: chat, vision, audio, embeddings, reranking, and TTS. No Python runtime, no ONNX, no Docker containers, no sidecar process, no IPC. Finally, AI in a jar. The stack underneath is built for the JVM rather than bolted onto it: toknroll: pure-Java tokenizers, zero dependencies gguf / safetensors: read and write llama.cpp and HuggingFace model formats jam: quantized matmul kernels, competitive with llama.cpp on CPU jota: Tensor API targeting Java, C, CUDA, HIP, Metal, OpenCL, and Mojo It ships with integrations for Spring AI and LangChain4j, and is compatible with GraalVM Native Image for low-overhead, self-contained binaries with millisecond startup. This is an early rel…