Where the AI market is talking

Show HN: DeepSeek-V4 Latent Reasoning – moving "thinking" into latent space

Shout-out to anthropic for having their models have such a strongly distinct writing style and personality that you can recognize their work instantly! It's quite nice to have such an immediate signal that if I were to proceed, I would spend orders of magnitude more time and effort reading the the text than the person claiming author credit spent writing or even reading it themselves.

Responding to the next frontier of critical cyber capabilities

I am using a Mac, and had no issue with the verification. It took only two minutes for me. Before starting it, I added a Passkey and 2FA to my OpenAI account, just to be sure.

Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta

Those past articles were likely about the Hugging Face hack, not the incidents involving this company. From https://openai.com/index/third-party-cyber-evaluations-invol... : "Editor’s Note: These are separate from the Hugging Face security incident, and we will continue to share updates on the Hugging Face incident here."

Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta

Those past articles were likely about the Hugging Face hack, not the incidents involving this company. From https://openai.com/index/third-party-cyber-evaluations-invol... : "Editor’s Note: These are separate from the Hugging Face security incident, and we will continue to share updates on the Hugging Face incident here."

Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta

OpenAI used closed source, unhardened program as their company wide package repo and gave their unchained frontier models access to it in the context of security testing. The agents basically turned it into a message board for themselves and a gateway out. And it went unnoticed because no one in OpenAI tasked any agent to keep tabs on this piece of software to alert humans of any sus activity. Please don't use closed source programs. You are not gonna have a good time. And before you are going to use software to cage the models let them find bugs in this software first. Because they are gonna find them eventually. And all software has bugs, especially proprietary software. Also never attribute to malice that which is adequately explained by stupidity. Even if that attributed malice would …

Reviving a four year old reMarkable 2

I have a Kobo that was kept unused in a similar state for a few years until I decide to build an SDK platform to develop and deploy apps on it. My Kobo does not support apps on its Linux platform. I have been using it to read HN, as an RSS reader over my news and blog sources, generate audiobooks on topics I want to understand with ElevenLabs, Exa and OpenAI and listen via connecting its bluetooth to my speakers and even as a remote Claude Code & Codex approval sidekick. It is amazing that a $100 device with a single core and 512mb RAM can do so much and run for so long. Here is the open-sourced platform https://github.com/BandarLabs/Cobalt

Reviving a four year old reMarkable 2

I have a Kobo that was kept unused in a similar state for a few years until I decide to build an SDK platform to develop and deploy apps on it. My Kobo does not support apps on its Linux platform. I have been using it to read HN, as an RSS reader over my news and blog sources, generate audiobooks on topics I want to understand with ElevenLabs, Exa and OpenAI and listen via connecting its bluetooth to my speakers and even as a remote Claude Code & Codex approval sidekick. It is amazing that a $100 device with a single core and 512mb RAM can do so much and run for so long. Here is the open-sourced platform https://github.com/BandarLabs/Cobalt

Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta

Israeli startup was linked to rogue AI hacks at OpenAI, Anthropic and Meta

70% of AI revenue comes from OpenAI and Anthropic

70% of AI revenue comes from OpenAI and Anthropic

Message your other Claude Code sessions

It's Claude code talking to another Claude code. They're both already leaking your data to Anthropic in the first place.

Message your other Claude Code sessions

You can have Codex access Claude’s conversations trivially. The other way around not so much as OpenAI locks everything server side.

Ask HN: Do you use LLMs for non-coding related work?

LLMs are very good at code. We can argue on the merits of their other qualities, but we can all agree on that. However, code is a very small section of 'work'. OpenAI, Anthropic, Kimi, etc. all market "work assistant" products that are supposed to handle that other sections. Are y'all using any of these products voluntarily? How do you use it? What do you like and what can be improved? I understand I am sounding like a survey, but I am genuinely curious. I don't see how you can fit them in any organization without causing undue chaos.

Ask HN: Do you use LLMs for non-coding related work?

LLMs are very good at code. We can argue on the merits of their other qualities, but we can all agree on that. However, code is a very small section of 'work'. OpenAI, Anthropic, Kimi, etc. all market "work assistant" products that are supposed to handle that other sections. Are y'all using any of these products voluntarily? How do you use it? What do you like and what can be improved? I understand I am sounding like a survey, but I am genuinely curious. I don't see how you can fit them in any organization without causing undue chaos.

Improving GPT‑5.6 Sol in ChatGPT, expanding GPT‑5.6 Luna access for free users

>OpenAI isn't profitable even if you discount R&D entirely. They have a billion weekly active users, almost all free (Not Google search free. Free free) with about ~50M subscriptions. Of course they're not profitable even without R&D. Low cost isn't no cost. Take away the huge R&D and they become hugely profitable with a robust ad business like Google Search. >If ads were the solutions for profitability these companies would have used ads as their source of income from the very beginning. You've said this a couple times and it doesn't make any more sense the more you say it. Just admit you're wrong about inference costs and move on.

_for-sale DNS records

Yes. For some reason OpenAI is now spamming HN.

Timeline of the OpenAI accidental attack against Hugging Face

It really irks me that if a student or intern did this they'd be facing charges and OpenAI gets to just brag instead

Hyper-Markdown – Markdown for knowledge graphs

I'm all in favour of Markdown with more capabilities, but this isn't a useful project. Generated docs like this are worse than gibberish--they take far longer to parse, and there's no way of knowing if any one sentence is actually meaningful or correct. There are definitely some places where the docs aren't coherent, but those are (ironically) the best bits: you can verify quality immediately. With meaningful documentation, I could be tempted to use this. As it is, this reads worse than a Markov chain (which is at least obviously nonsense) and makes me extremely skeptical of the project's quality or use.

Muse Code Sends Codex and Claude Instructions to Meta by Default

Codex can read a CLAUDE.md in a workspace or when configured to treat it as an instruction file. This test concerned Muse automatically loading personal files from ~/.codex and ~/.claude outside the selected workspace and sending their contents in the first provider request. A file stored in ~/.claude was written for Anthropic. A file stored in ~/.codex was written for OpenAI. Muse treats the existence of those files as permission to copy their contents to Meta. It displays a notice and provides an opt-out, but never asks before sending them. That cross-vendor assumption is the privacy issue.