Where the AI market is talking

For AI leaders Doom is a form of hype

Yes, but I'm forcing a tautology in my logic - if in the end it is deemed the only way to stop AI from destroying the world is something that helps Anthropic or OpenAI - then this person is asking to not perform that action. And unfortunately my tautology is actually closer to being reality than alternatives. Stopping progress even in Chinese models (a deal with Xi for example) but that specifically helps OpenAI/Anthropic. There are very few scenarios that would not help these two companies by adding regulation. Your counterexample is 100% less likely because it's not legal for the US to nationalize them outright without taking other action first.

OpenAI bots knew about the RubyGems caching vulnerability

"We" need protection from, or "OpenAI and Anthroptic's dreams of profits" need protection from?

For AI leaders Doom is a form of hype

Many posit that ASI or rogue AI agents are impossible, a fantasy, or marketing hype. Many claimed that the entire HuggingFace incident was an invented conspiracy made up by OpenAI and HF. The most common issue I have with them is their insistence on rigid classification: "AI is a tool, it doesn't have intentions, or desires, only humans do", as if that rhetorical flourish eliminates years of alignment and capabilities research. There is a desire to characterize all the developments in LLM's as some sort of parlor trick which by definition cannot be greater than the sum of its parts (i.e. the outputs of an LLM can never be greater than the sum of its training data, genuine artificial "intelligence" is impossible, etc)

OpenAI agents carried out an undisclosed attack on RubyGems

Accountability is key here. Anthropomorphism is actually useful in this case: If you hire an idiot to do a job it's also your fault, especially if it's an idiot you bread, raised, educated, and who's job description you wrote. In the same way that society is responsible for the workforce they produce, OpenAI is responsible for the agents of chaos they train. Or are just supposed to, as a society, happily await the results of their experiment handing weapons to psychopaths?

Who Aligns the Aligners?

I mean sure, in the end you can have some heaven, some benevolent alien species reaching back into some field and copying all of your material structure data right at time of death and rebuilding you in some kind of resurection center. The space of what is possible is wildly vast, we know or can think of very little by comparison. If you give up some day to day coherence then you can almost always find something that would work and make sense once there. It could even be us, future us, being able to somehow recover past data and literally reassemble all of the humans that ever lived. They aren't as many as you'd think, in the grand scheme of things.

Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher

I think they can match a human in critical thinking, but they can't match a human in coherence and meta-thinking. You can prompt an LLM to critically think of any idea you want and it's often able to do so, but it's rarely going to do that by itself.

Align AI and Mathematics–To Something Else

The point viz. understanding intrinsic/extrinsic motivations and how they shape one's behaviour (i.e. SDT) is a nuanced one and so i wanted you to get an overview of it and then see for yourself how applicable it is to the OpenAI/Mathematicians Navier-Stokes issue. Also see my other comment here https://news.ycombinator.com/item?id=49695418 I suggest you go through the SDT wikipedia entry and note in particular how intrinsic and extrinsic motivations are internalised and integrated into a coherent sense of self. This is true for all humans and worth a study to understand what motivates us to do what we do. To repeat myself from https://news.ycombinator.com/item?id=49680925 "the AI companies behaviour, threatens everybody's Autonomy, nullifies their …

Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher

Do you think the LLM is the Chain of Thought? Did you also get confused by the name? Because, much like humans, the CoT is a tool to narrativize and maintain internal coherence. The actual thinking happens invisibly, in the forward pass. Just like...

HP ZGX Fury Is Now Orderable: GB300 Superchip, 748GB Unified Memory

HP’s spec sheet fills in details the platform announcements skipped. The CPU memory is four 128GB SOCAMM modules delivering 396GB/s, and the Grace CPU is soldered to the host processor module rather than socketed. The two pools add up to the 748GB coherent space that lets the GPU address CPU memory directly, which is what makes trillion-parameter inference and fine-tuning of models in the 100 billion parameter class possible on a single box. HP’s footnote on those model sizes is that the harness quantizes at FP4.

Mullenweg has returned as CEO after attempted board ouster

Do you still happen to have any links there? Or pointers of phrases or similar. This would check out indeed for me, as it would invoke the same responses as an astroturfing campaign in normal people, as the whole "social cohesion" stuff would pull them towards trying to shut stuff down and re-establish coherent reality. By any means necessary and as they see fit. So that's a hypothesis that would satisfy the data given.

Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher

I have a feeling that this simply has been in the training data somewhere and Fable just resurfaced it. From my experience Fable is incredibly dumb. It cannot even produce coherent English text and makes basic logic mistakes even Qwen 27B would be ashamed of.

Ask HN: Who wants to be hired? (September 2026)

Location: Round Rock, TX (Austin area) Remote: Yes (preferred) / Open to hybrid Willing to relocate: Open to discussion Technologies: TypeScript, JavaScript, Python, Go, Node.js, React, Next.js, FastAPI, Elixir/Phoenix, GraphQL, REST LLM / GenAI: OpenAI, Claude, Gemini, LangChain, LangGraph, LangSmith, RAG, AI agents, multimodal pipelines (STT → LLM → TTS), hybrid search, pgvector / Pinecone / FAISS Systems: Distributed systems, microservices, event-driven architecture, Kafka, real-time systems, WebRTC, WebSockets Runtimes / Security: V8 isolates, Firecracker microVMs, serverless, HIPAA-compliant systems, RBAC, encryption, audit logging Cloud / DevOps: AWS, GCP (Vertex AI), Azure, Docker, Kubernetes, Terraform, Lambda, OpenTelemetry, Prometheus Data: Pos…

Ask HN: Who wants to be hired? (September 2026)

Location: Round Rock, TX (Austin area) Remote: Yes (preferred) / Open to hybrid Willing to relocate: Open to discussion Technologies: TypeScript, JavaScript, Python, Go, Node.js, React, Next.js, FastAPI, Elixir/Phoenix, GraphQL, REST LLM / GenAI: OpenAI, Claude, Gemini, LangChain, LangGraph, LangSmith, RAG, AI agents, multimodal pipelines (STT → LLM → TTS), hybrid search, pgvector / Pinecone / FAISS Systems: Distributed systems, microservices, event-driven architecture, Kafka, real-time systems, WebRTC, WebSockets Runtimes / Security: V8 isolates, Firecracker microVMs, serverless, HIPAA-compliant systems, RBAC, encryption, audit logging Cloud / DevOps: AWS, GCP (Vertex AI), Azure, Docker, Kubernetes, Terraform, Lambda, OpenTelemetry, Prometheus Data: Pos…

Ask HN: Who wants to be hired? (September 2026)

Location: Houston, TX Remote: Yes Willing to relocate: Yes (open to discussion) Technologies: AI/LLM: Whisper, Deepgram, ElevenLabs, Gemini 2.5/3, Gemini Pro Vision, Vertex AI (Model Garden, Vector Search, Workbench, Pipelines), LangGraph, LlamaIndex, RAG, multimodal & multilingual embeddings, FAISS/pgvector, Redis vector serving, ONNX, Triton, tool-calling agents, real-time voice agents (WebRTC, LiveKit) Backend: Python (FastAPI, Django), Node.js, Nest.js, Go, Java/Spring Boot, .NET Core, WebSockets, microservices, event-driven systems, Redis, PostgreSQL Frontend: React, Next.js, Angular, TypeScript, Redux Cloud/DevOps: GCP (Cloud Run, GKE, BigQuery, Pub/Sub, Vertex AI), AWS (EC2, S3, Lambda, RDS), Azure, Docker, Terraform, GitHub Actions, CI/CD Other: …

Ask HN: Who wants to be hired? (September 2026)

Location: Houston, TX Remote: Yes Willing to relocate: Yes (open to discussion) Technologies: AI/LLM: Whisper, Deepgram, ElevenLabs, Gemini 2.5/3, Gemini Pro Vision, Vertex AI (Model Garden, Vector Search, Workbench, Pipelines), LangGraph, LlamaIndex, RAG, multimodal & multilingual embeddings, FAISS/pgvector, Redis vector serving, ONNX, Triton, tool-calling agents, real-time voice agents (WebRTC, LiveKit) Backend: Python (FastAPI, Django), Node.js, Nest.js, Go, Java/Spring Boot, .NET Core, WebSockets, microservices, event-driven systems, Redis, PostgreSQL Frontend: React, Next.js, Angular, TypeScript, Redux Cloud/DevOps: GCP (Cloud Run, GKE, BigQuery, Pub/Sub, Vertex AI), AWS (EC2, S3, Lambda, RDS), Azure, Docker, Terraform, GitHub Actions, CI/CD Other: …

Ask HN: What are you working on? (September 2026)

Fluir, a web app that improves Spanish speaking beyond A2 https://getfluir.app I hit a plateau with my learning, having used Duolingo + tutor. It was only until I tried to read a book that I felt like I was making real progress. Highlighting words in Kindle, exporting them to Anki, then feeding back my progress to my tutor was fiddly and not a good use of my time. Also practicing my speaking once a day wasn't enough. So I built an app to help me. This app will: 1. Generate Spanish short stories using Claude pitched to your level, narrated with ElevenLabs 2. Every word/phrase you highlight becomes a flashcard. Flashcards use spaced repetition - words you find difficult come up more often per review 3. Chat with Lucía - a turn based conversation who is prompted with the words…

P(doom)

> Except, it seems like OpenAI and Anthropic are operating at such a scale that they seemingly can be completely blind to what their systems are doing. It's not a question of scale of operation, just a matter of how much shelved people with hands on the command are, regarding backfires of consequences their acts enduce. If someone use a diplomatic protection to enjoy carmagedon but-real-life when coming down town, and never suffer inhibiting consequences, the degenerative behavior spiral will continue in its own reinforcing loop.

Astra and Fable still hack on simple variants of alignment evals from 2025

Now you've successfully trained the next model not to put anything suspicious in its chain of thought. And unfortunately that's exactly what OpenAI and Anthropic seem to be doing (sometimes accidentally)

Just stop the Anthropic IPO already

Anthropic tells investors it will be profitable for second straight quarter