Where the AI market is talking

Grok Bot

According to Boris Cherny from Anthropic [1], the threat of prompt injection has been largely solved. [1]: https://x.com/bcherny/status/2086520950259118464

OpenAI’s head of ethics leaves less than a year after joining

staying for less than a year usually means you don't take any equity with you, it must've been an incredible offer in that case, since those openai shares are a golden ticket right now

Nvidia's Risky Business

> But people have been saying this about CUDA for twenty years, and we are not any closer to a replacement GPGPU paradigm today. How much money was in it for the first decade or so? I think AMD was asleep at the switch but e.g. Apple just did their own thing for the parts which they prioritized. My understanding is also that Anthropic and OpenAI have also worked to decouple themselves so I think it’s likely that the CUDA moat is going to be less of a barrier than it used to be from the perspective of guaranteeing Nvidia profits.

Asked AI to do a mini math breakthrough. It did a mini breakthrough

Yesterday Anthropic published a new result about the Riemann zeta function. A research version of Claude had moved a mathematical lower bound from 41.666% to 67.25007%. I read that and said to Nix, my own AI agent: “Give me a mini math breakthrough on top of it” It back with a 7-page mathematical argument claiming 67.30214%. I have no idea what I did lol, but 4 different agents confirmed that. One even thinking half an hour

Nvidia's Risky Business

I think the linked semi-analysis piece is about right. Google is more focused on protecting its 4.2 trillion dollar golden goose than pushing cutting edge which leads to a bureaucratic nightmare that researchers simply dont need to engage with when openAI/anthropic are offering even more money. Why waste time dealing with bureaucracy at google when you can be top dog at openAI or anthropic?

Compression is prediction

>Mixing this can let you believe that you can use a compressor to do the same thing as LLMs, which you cannot. You can, actually! Any compressor can be losslessly converted into a generator, and vice versa. Traditional compressors like gzip are of course very simple and can only replicate rough patterns from the input. But they are technically doing the same thing.

Show HN: I used AI to filter AI-focused content from HN

Hi I updated the title because the previous one was a little misleading. I'm using GPT 5.6 sol and simply asking if the content and title indicate that the article is primarily focused on AI, rather than being generated by AI. I checked hugging face and found surprisingly few models dedicated to detecting AI-generated text though, but there are some APIs and sites like GPTZero available for it. Perhaps the "detect AI-generated content and remove it from feeds" problem could be an interesting one to investigate though

Ask HN: Who is hiring? (August 2026)

Bactrix | Founding Software Engineer | REMOTE | Equity | AI/Biotech We’re building an AI platform for antibiotic drug repurposing and antimicrobial resistance research. Looking for a strong software engineer to join the core founding team and build production AI workflows, backend infrastructure and scientific tooling. Stack: Python, LangGraph/LangChain, APIs, cloud, AI agents; experience with ML/biotech is a plus. Early-stage/pre-funding role with equity. Looking for someone genuinely interested in building long term. Contact: HR@bactrix.com

Ask HN: Have you used LLMs to reinvigorate you niche programming community?

I've seen a few discussions over the years about how niche languages like Lisp, Forth, Prolog, etc. would be great to use, but even for personal projects the lack of available packages makes them hard to use. However, with an LLM you can have it replicate whatever package you'd like from a more popular language. You can even port software to your favorite niche architecture. If you really want to run your homelab off a PA-RISC machine you just give an agent SSH access and tell it to make the software work. For myself, I have only done very little of this. I did add POWER vectorization to POV-Ray so that I can get a little more speed out of a machine that has more power than it needs for other things. Instead of letting it use something like Python or Go for small scripts I'll tell the LLM…

To Save C, We Must Save ABI (2022)

Smart-arse is comparing C versus JavaScript, instead of C vs C++, for example. And then coming with such lengthy ad hominem. Let make a fun exercise for the audience, given your performance remark. Paste a random C code that I should replicate in whatever language I feel like. There is one rule. If the sample code is pure ISO C, then I will only use what is in the standard of whatever language I pick up. If the sample code makes use of single language extension not part of ISO C, then I will have the freedom to also pick whatever language extensions I feel like.

Show HN: Lance-bundle – Portable embeddings to embed once, query forever

Hello, While experimenting with personal, local RAG app setups, I kept having to (re)generate embeddings and really wanted precomputed embedding datasets that I could quickly pull and use in various environments. I built a library to test out the idea: lance-bundle lets you package precomputed embedding vectors alongside the actual embedding model so that everything can be loaded from a single file for querying against the vectors; initial version uses LanceDB + ONNX for low dependency footprint and fast cold start to vector queries. https://github.com/cloudkj/lance-bundle As part of this, a few datasets that might be of interest to this audience have been precomputed as embedding vectors and hosted on a Hugging Face dataset hub and can be directly loaded and queried a…

Rust SIMD on the GPU

On a CPU, hyperthreads are mostly replicated register banks. This allows the CPU to hold the context for 2 threads simultaneously. And, lets parts of a CPU make progress on one thread while the other thread is stalled. CPUs also has a kinda large microcode register bank that helps work around dependencies in asm instructions that reuse named registers. On the GPU however, the hyperthreads are just a round-robin execution queue to take advantage of instruction pipelining. The register bank of a single GPU core is huge and can be flexibly divided across a variable number of thread contexts when a kernel is launched. Many thread contexts can be held in registers simultaneously in a single GPU core. That makes stalling on memory latency much less of a problem. The hardware can focus on delive…

As AI eats the web, the internet’s collective memory is disappearing

Well, I guess another plausible explanation is that usually "domain experts" are people that get paid for their expertise, and have a vested interest in saying that LLMs cannot replicate what they want people to pay them for.

More than 10 firms pay up to $100k a month for access to Truth Social posts

The degree to which his works have been dragged through the mud by a bunch of people who insist on calling everything they don't like "capitalism" can't be overstated. The majority of people complaining about capitalism don't even have a coherent definition of it, it's a disgrace.

Linus Torvalds says AI has made 'huge' Linux kernel updates the new normal

"UNIX is simple and coherent, but it takes a genius (or at any rate, a programmer) to understand and appreciate its simplicity." — Dennis Ritchie, "Unix: A Dialectic" (1989)

Mars Bar from 1991 found – and it's 20g bigger than today's

Unintelligible. Re-write this into something coherent so I can respond if you like.

OpenAI's Only Ethicist Reportedly Left Last Month

How Claude marks AI-generated content

Why would I want a stochastic parrot that intentionally speaks less probable tokens? Anyway, I've found a magic line that can be copied & pasted to the comment sections of most OpenAI/Anthropic news threads. This one is no difference. The magic line: > Doesn't matter; have DeepSeek.

How Claude marks AI-generated content

Why would I want a stochastic parrot that intentionally speaks less probable tokens? Anyway, I've found a magic line that can be copied & pasted to the comment sections of most OpenAI/Anthropic news threads. This one is no difference. The magic line: > Doesn't matter; have DeepSeek.

How Claude marks AI-generated content

> Honest note: Anthropic has not shipped a public Claude watermark detector yet. This tool uses rewrite-based neutralization — a meaning-preserving paraphrase with a non-Claude model — which is the attack path watermark research points to. Not affiliated with Anthropic. Well, they should have run their own AI slop website through their tool...