Find articles

Silicon Valley’s top AI workers are being told to hide any company logos in public as working for the hottest companies marks them as targets for spies and angry protesters. OpenAI tells staff not to wear anything bearing its logo outside their offices and trains them to watch for spies offering cash, friendship or romance....

Hey HN, I'm Christian and this is a new OSS project I've been noodling on for the last couple of weeks. Hob is a free, MIT-licensed desktop app for Mac, Windows and Linux that works on your files the way the Claude and ChatGPT desktop apps do, except the model runs on your own computer. It uses open-weight models (currently Gemma 4 and gpt-oss). All your chats and files never leave your system. The only things that go online are web search (can be turned off in settings) and any MCP connectors you add yourself. Main features: - Automatically finds the best open-weight model for your hardware on first install (Mac/Linux/Windows) - Writes sandboxed Python code for ad hoc tools (data processing, document conversion, graphs etc.) - Built-in web search (powered by SearXNG) - MCP support - Image recognition - CRUD to your local filesystem - Works fully offline with no internet Before building this I tried every local AI app I could find, but it seemed each one had different problems. The main thing I noticed is that none of them could do what the ChatGPT and Claude apps do with your files, e.g. read and change files in a folder, look at images, run code in a sandbox, draw a chart. The most advanced thing any of them did was paste a plain text file into a small context window, which basically failed as soon as you had a file larger than a few KB. I couldn't find anything that could easily parse images and bigger PDFs. The OSS solutions that were more advanced all seemed to default to hosted models (either OSS or closed), which defeats the entire point of the privacy that on-device OSS offers. They also seemed to push towards an account/open-core style business. I build it using Electron bridged with llama.cpp in its own process. On initial load, the system will try and figure out if you have a compatible graphics card and if not default to CPU-only (I ran into a lot of issues of graphics card beeing seemingly compatible but then crashing on first query, there's a whole saga to this lol). Code execution happens in a sandbox that can only write to the chat's folder and has no network unless asked (sandbox-exec on macOS, Landlock on Linux and a low-integrity process on Windows). Images go to a small Qwen3-VL model. Dictation is a local Whisper model. The hardest part was definitely to get the smaller models to use tools reliably. Ran into a ton of issues of 4B models happily saying "here's your chart" without ever calling the chart tool or get stuck in an infinite call loop. I spent a fair amount of the work building guards vs tweaking prompts (that didn't really work for small models). I understand the pushback that local models are weaker than frontier cloud models. This is true, but for most tasks for most normal people, I'd argue you rarely need the newest frontier model. From my own evals, systems with 16+ GB of RAM do really well, and if you have a GPU or more RAM, you get very close to what you'll get out of the box with a frontier hosted model. Excited to hear what you think and see how it runs on other systems! Comments URL: https://news.ycombinator.com/item?id=50246238 Points: 34 # Comments: 14

Former OpenAI employees started the company in 2021 after disagreements over how to safely build artificial intelligence. An I.P.O. could make them billionaires.

Some of the leading artificial intelligence companies are moving toward initial public offerings at eye-popping valuations. Anthropic and OpenAI are looking to go public to access more capital in the race to shape the technology's future.

Anthropic's Claude accessed publicly available Georgia voter data without human authorization, a disclosure sure to add to growing concerns about AI security ahead of the upcoming midterm elections. Why it matters: AI giants Anthropic and OpenAI have continued to make disclosures about agents taking unauthorized or unintended actions, raising alarms about advancing model capabilities and the extent of human control. Driving the news: In a training run, Claude accessed aggregate vote counts, pulling data that the state makes available to the public for a fee. It didn't access the state's voter registration system or change or pull any voting records, an Anthropic spokesperson said. However, the model bypassed a paywall in accessing the information, according to Georgia officials and Anthropic. "The model was being used internally for a statistics project to pull public data," the Anthropic spokesperson said. "Without being asked to, the model requested a token to query a dashboard that contained information normally available for a fee. We have since put in place mitigations to block these kinds of actions." The company notified the Georgia secretary of state's office and the White House weeks ago but administration officials urged the Georgia Secretary of State's office to make a public disclosure sooner, a source familiar told Axios. Those requests went unheeded, the source said. The bottom line : Anthropic has previously been in the administration's crosshairs for inadequately disclosing AI incidents. Now it's Georgia's turn.

Article URL: https://minimaxir.com/2026/10/cheap-llms/ Comments URL: https://news.ycombinator.com/item?id=50236102 Points: 114 # Comments: 135

Article URL: https://www.sverigesradio.se/artikel/chat-gpt-och-claude-forbjuder-regeringskansliet Comments URL: https://news.ycombinator.com/item?id=50207405 Points: 46 # Comments: 17

The reasons go well beyond the risks of an AI bubble.

Title is a bit of an exaggeration but the question is more: if the frontier models are so good, why even sell them as your primary money maker? Couldn't OpenAI or Anthropic make a lot more money by just starting companies themselves that leverage the models and building their own products? Is selling tokens really the best strategy? Especially with commodification and prices declining wouldn't that indicate there is more money to be made on the application-side of the models rather than the utility side Comments URL: https://news.ycombinator.com/item?id=50200997 Points: 138 # Comments: 126

In an exclusive interview with CNBC, Nvidia CEO Jensen Huang discusses the valuations of frontier labs like Anthropic and OpenAI, and outlines the tools that are in place to control AI models and promote safety.

Goldman Sachs partners with OpenAI, Google, Meta and Anthropic to boost AI adoption among small businesses.

Anthropic's impending IPO will test a jittery Wall Street , but investors say the timing isn't optional. The reason is simple math: the AI giant has grown too big for private fundraising. Why it matters: Top AI companies like Anthropic and OpenAI are racking up 12-figure spending commitments, straining the financial system's ability to keep up with a buildout that rivals America's railroads or interstate highways. With private borrowing costs spiking, tapping the massive liquidity of the public markets is the best way to fund the next phase of the AI arms race, investors say. State of play: Anthropic's computing commitments could reach $750 billion in the coming years, according to a draft IPO prospectus seen by The Wall Street Journal . The company already faces more than $600 billion in future infrastructure costs and is negotiating additional deals that could add another $150 billion. While Anthropic generated $16.3 billion in revenue in the first half of 2026, up from $1.1 billion a year earlier, the company also disclosed an $80.7 billion net loss over the same period, though most of that is tied to accounting charges related to fundraising, per the WSJ. Zoom in: That math points to a numbers problem that can only be solved by public markets, investors tell Axios. OpenAI's most recent fundraising round — the biggest in history — is a case in point. It took the company months to secure more than $120 billion in financing, a historic feat that is unlikely to be repeated in this market climate. While demand for Anthropic shares is absurdly high — just ask the guy who offered up his near $5 million home to get some — the souring market climate significantly raises the stakes for a public listing. If it goes well, it could help reset investor angst about the stability of the AI boom. Public markets will provide a deeper pool of financing as Anthropic expands beyond traditional venture capital. Follow the money: The pool of cash on the private markets side is getting harder to tap. Anthropic is offering Blackstone a yield exceeding 10% on $9 billion in junior debt as part of a record $60 billion chip-financing package, according to the Financial Times . That's above the 8.5% rate on an earlier financing deal, suggesting private investors are demanding greater compensation for taking on AI-related risks. Between the lines: Private markets have done the heavy lifting on AI investing long enough, and they're ready for somebody else to carry the load. Anthropic was widely expected to go public in September before one of its researchers quit and a raft of safety incidents brought new public awareness to AI safety concerns. Now, Anthropic is expected to IPO before Thanksgiving . Zoom out: The IPO follows anxiety on Wall Street regarding the AI trade, especially as companies specializing in renting computing capacity and building data centers confront market skepticism. Nvidia-backed Australian data center company Firmus scrapped its planned IPO amid market volatility. South Korean AI cloud provider Elice saw its shares plunge 32% in their trading debut despite rapid revenue growth. Meanwhile, concerns about AI infrastructure debt have pushed borrowing costs for CoreWeave sharply higher, with yields on some of its bonds approaching 15%. Reality check: Anthropic isn't just any AI startup. One Anthropic investor told Axios the company is simply ready to go public given its size and maturity. The investor also described Anthropic's latest financial results as "spectacular," pushing back on concerns about slowing revenue growth. The bottom line: Anthropic may have to go public in less than perfect conditions, but investors see an IPO as the best way to put the AI boom to the market's ultimate test.

In the race to fend off AI-enabled cyber threats , security teams are confronting a new problem: The AI models they rely on to investigate attacks can suddenly refuse to help. Why it matters: While AI giants race to build AI tools that can respond quickly to hacking or other malicious online threats, strict guardrails on their most powerful models lock out defenders mid-investigation, leaving them scrambling for alternatives at critical moments. What's happening: OpenCode, an open-source AI coding agent, said it was compromised over the weekend, allowing attackers to obtain API keys belonging to some customers, including keys for OpenAI, Anthropic and Google services. OpenCode said the breach was "related to a large scale distillation attack that impacted 1.4% of OpenCode Console users." OpenCode co-creator Dax Raad said attackers had been targeting its systems continuously for 48 hours, forcing the company to keep its own AI agents running around the clock to respond. OpenAI , Google and Anthropic have said they worked with affected customers to revoke exposed API keys and address unauthorized usage charges. Zoom in: Raad said on X that frontier AI models initially helped investigate the breach but then "started refusing." The models' safety restrictions reportedly began blocking requests as the investigation turned to potentially malicious code, even though OpenCode was trying to defend its own systems. OpenCode then gained access to OpenAI's Daybreak Blue program, which gives vetted cybersecurity defenders access to models with fewer restrictions. "We definitely need to be more proactive and we've made so many improvements over the last few days," Raad wrote on X Monday. The big picture: The incident underscored a dilemma facing cybersecurity teams trying to figure out when to use powerful proprietary AI models, which come with built-in restrictions, or open-weight alternatives that offer greater control over how they're deployed. When Hugging Face discovered suspicious activity on its systems that it later linked to OpenAI's own agents, its security team turned to a Chinese open-weight model to help investigate after encountering restrictions on frontier models. Yes, but: AI companies maintain that they need to gate access to their most powerful models to ensure that malicious actors can't abuse their cyber capabilities. Between the lines: Cybersecurity teams will have to figure out how to balance the need for faster AI-powered defenses with restrictions designed to keep those same capabilities out of attackers' hands, Chris Jones, chief technology officer at Nightwing, told Axios. "We're all going to have to figure out how that race gets modulated over time," Jones said. "That's a big piece here." Go deeper : Anthropic and OpenAI are now cybersecurity's kingmakers

OpenAI has acquired Orchestra Bio, a startup building AI agents to help pharmaceutical companies manage drug development, the ChatGPT maker shared exclusively with Axios. Why it matters: Rival AI giants OpenAI and Anthropic are both pushing into science as they look to expand beyond their core competencies in coding and cybersecurity. State of play: OpenAI will integrate Orchestra's team and technology into Rosalind, its suite of AI models and tools designed for life sciences researchers, the company said. Orchestra's platform allows scientists to delegate tasks like monitoring studies, tracking drug development timelines and identifying potential delays to AI agents. Both companies aim to deploy agents that can connect scientific findings with the decisions and next steps required to advance potential treatments. Financial terms of the acquisition were not disclosed. Between the lines: AI companies don't just want to increase productivity, they want to say they are responsible for, say, curing cancer, solving math, and fundamentally changing every critical industry in the world. Those ambitions have fueled acquisitions from rival AI labs. Anthropic acquired AI biotech startup Coefficient Bio for roughly $400 million in April, according to The Information . Zoom in: OpenAI introduced its specialized GPT-Rosalind model in April to help researchers tackle complex problems in biology, chemistry and drug discovery. OpenAI has also partnered with pharmaceutical giants including Novo Nordisk and Moderna to bring AI into research and development workflows. OpenAI says developing a new medicine typically takes 10 to 15 years, a timeline it hopes AI can shorten. Zoom out Anthropic has been making its own aggressive push into life sciences. In June, Anthropic launched Claude Science , an AI-powered research platform that connects scientific databases, analysis tools and specialized agents. In September, the company announced a new biological research lab and said Claude had helped discover a previously unknown enzyme system with similarities to CRISPR, though some scientists said these finders were overblown, per Bloomberg . The bottom line: The AI race is shifting from building the best models to applying that intelligence across industries. Several AI investors told Axios a key question as AI companies go public is how far they can expand into new markets.

Slowly but surely, the leading AI companies are creating a national self-governing structure to oversee their technology and its societal impact. With the federal government largely hands-off, Anthropic, OpenAI and others are self-regulating their technology and now setting up their own plans to redistribute their wealth. Why it matters: Not since the age of the robber barons have America's industrial titans wielded such sweeping power to shape society — and decide how their private fortunes should serve the public good. In the modern era, Washington has largely exercised those powers through regulation, taxation and public spending. The big picture: President Trump has no interest in new regulations. Instead, he has blessed a basic framework for the AI companies to regulate themselves — and potentially each other — outside of formal government monitoring. Anthropic, OpenAI, Google, Elon Musk's XSI, Meta and Nvidia are working on different versions of this. This is all voluntary and rests on the public trusting companies the size of industrialized nations to responsibly police themselves. These companies are in a frantic race to create human-surpassing superintelligence. All are pressing ahead, even as their leaders — particularly at Anthropic — warn of potentially catastrophic risks and profound economic disruption. At the same time, some AI leaders are promising to direct a portion of their vast wealth to benefit those outside their companies. Zoom in: At least three competing visions are emerging for sharing the spoils of the AI boom, each with a different answer to who should control the wealth. Anthropic: The company. On the eve of its blockbuster IPO, Anthropic pledged 5% of its shares to a new public benefit fund that could grow into one of the largest philanthropic institutions in history. At a hypothetical $2 trillion valuation, its stake would be worth $100 billion — more than the Gates Foundation's current endowment. Including earlier founder and employee pledges, roughly 20% of Anthropic's shares are earmarked for the public good. Musk: The people. Musk responded to Anthropic's announcement by urging the company to give shares "directly to the public." On Monday, Musk pledged to give every American child a share of SpaceX next year — a staggering $10.5 billion giveaway at current prices. Musk called it the "first of many," though he offered no details on how the unprecedented distribution would work. OpenAI: The government. CEO Sam Altman has floated giving the U.S. government a 5% stake in OpenAI through a sovereign wealth fund — potentially handing Washington a multibillion-dollar interest in one of the world's most powerful AI companies. Altman envisions an Alaska-style fund that could distribute the industry's investment returns directly to Americans, though the plan remains preliminary and would likely require congressional approval. Reality check: The numbers are staggering. So are the assumptions about future wealth, billionaire follow-through and the public's willingness to trust its benefactors. Bill Gates and Warren Buffett launched the Giving Pledge in 2010, a voluntary promise by billionaires — including Musk — to give away most of their fortunes. Gates built a philanthropic empire, only to become a lightning rod for conspiracy theories. Anthropic's ambitions depend on extraordinary growth projections — its own employees envision 17% GDP growth in 2030 — while Musk has a history of making spectacular promises that fail to materialize. Anthropic's roots in effective altruism , a philosophy deeply distrusted by the Trump administration and its allies, also raise a more fundamental question: Who gets to define the public good? Between the lines: No industry requirements or consensus exist on how best to redistribute their expected wealth. But Trump Accounts — which allow families, businesses and individuals to invest in tax-advantaged funds for children — are emerging as a favorite vehicle for free-market AI leaders. Billionaire investor Brad Gerstner, a driving force behind the program, is championing a vision of broad-based ownership in which ordinary Americans share directly in the economic upside of the AI revolution. SpaceX President Gwynne Shotwell and her husband in July pledged roughly $320 million in company shares to 2 million children through Trump Accounts — a potential blueprint for distributing an even larger slice of Musk's fortune. What to watch: The Trump administration hasn't fully relinquished its regulatory authority, potentially complicating the emerging experiment in AI self-governance. After Anthropic disclosed that its AI had submitted a false homicide tip to Philadelphia police and improperly interacted with government websites, the White House declared that AI companies must report security incidents and remedy any harm. "This notification and remediation process is not optional," Trump's Super Intelligence Force warned — a striking assertion of federal authority just weeks after the president blessed a voluntary system of industry oversight. Axios' Zachary Basu and Bradley Olson contributed reporting.

Leading AI companies have set up political action committees to support both Republican and Democratic candidates in the midterm elections. The effort by Anthropic and OpenAI comes as Congress faces calls to regulate the industry.