Tootfinder

Sources: Nvidia is developing a Nemotron 4 model with 1T parameters, up from Nemotron 3 Ultra's 550B parameters but smaller than leading Chinese open models (The Information)
https://www.theinformation.com/articles/nvidia-trying-develop-wo…

Why Nvidia Is Trying To Develop The World’s Best Open-Source AI Models
Nvidia is doubling down on its open-source efforts, pouring investment into an ambitious new in-house AI model that it hopes will drive demand for its hardware—but that could also compete with customers and partners. Nvidia is aiming for the new model, the biggest of a family that it is calling ...

Working in computer security, I've learned to be *real fucking careful* about this law and have heard of it being interpreted *very broadly.* If a client of mine described a system to circumvent access restrictions like this I would fire them. But assuming I didn't fire them, I would
1) tell them to shut it down immediately
2) tell them to talk to a lawyer.
LLMs, as a class of user, are clearly not authorized to access the information. By circumventing access controls, they have "knowingly accessed a computer without authorization." Those access restrictions exist to protect against various damages, including bandwidth use and "intellectual property theft" (another concept I'm not a fan of, but anyway). By violating those restrictions they could open themselves up to legal liability.
Now, IANAL. My job is to describe risk. That's a sufficient risk. It actually doesn't even matter if you'd win the case in the end, a lot of people just can't afford to fight.
But, in this case, it tends to be powerful people, for whom the law does not apply, abusing people who neither have the protection of the law nor the money to just ruin someone's life with a flood of lawyers.
https://www.law.cornell.edu/uscode/text/18/1030

Source: Trajectory, founded by ex-DeepMind, Apple, OpenAI, and Meta staffers to build continual learning models, raised $40M led by Sequoia at a $300M valuation (Stephanie Palazzolo/The Information)
https://www.theinformation.com/articles/tr

Trajectory, Founded by Ex-Google and Apple Researchers, Raises Funding From Sequoia in Back-to-Back Round
As closed-source models have gotten more expensive, businesses have found workarounds by customizing open-source AI for their specific needs. Some are also trying to improve the software they use that allows models to use tools and take actions, otherwise known as a harness.Now, startups that ...

🚀 Heute ist wieder unser Culture Information Ticker erschienen!
💻️️️ Jeden Dienstag veröffentlichen wir kuratierte Infos für unsere 4Culture Communities, dieses Mal zum Call for Proposals für das Open Science Festival (18. –19.09., Bamberg), Infos zum Cyberangriff auf das Badische Landesmuseum Karlsruhe, und die Dokumentation der Restauration von Flipper-Automaten, initiiert von HSBI-Studierenden am Campus Minden.
👉️️️

How is #WSocial getting on? Well, they may be violating #GDPR by exposing users full legal name, DOB and nationality (even if it's not publicly viewable on their profiles) #noxp

W Don't Have Time (@wdonthavetime.selfhosted.social)
Have you been Wned? W Social says it respects your #privacy. But how much of your personally identifiable information is published by them to the open web for anyone to find? wdonthavetime.tngl.io/wsocial-wned/ #WSocialFacts #WSocial

RE: https://grapheneos.social/@GrapheneOS/117057099753905023
To explain for non-techies: Google is actively obstructing the way tagged version information of some source code can be accessed in supposed open source source projects. Normally these vers…

If everything that’s made with it is free of copyright and available as free and open source. Not just the model weights; all of it. So Antrophic, OpenAI, Google, Microsoft, etc., are forced to release all their software as free software.
In other words, if we get rid of copyright altogether for everyone and all tech and all information enters the commons. Of course, that would also take implementing universal basic income, universal healthcare, housing, etc. (i.e., a proper social saf…

Zara Rahman, Elisa Lindinger: Is it too late to build an AI commons?
https://giswatch.org/en/floss-freelibre-open-source-software-open-source-technology/it-too-late-build-ai-commons

How mature is this repository? My long quest for open-source software metrics
When I discover an open-source software project for the first time, how can I estimate whether it is mature, healthy and usable?
I just released OSSTRL - Open Source Software Technology Readiness Level to help me calculating the maturity of open-source projects.
#opensource

🗓️ Culture Information Ticker - Issue Week 36
🗣 Call
Weiterentwicklung des strategischen Rahmens für Open Data 💾
Das @… plant, den strategischen Rahmen für Open Data weiterzuentwickeln, und hat dazu eine Online-Konsultation gestartet. Interessierte können bis zum 15. September ihre Anregungen einreichen.
🔗

Meta, Nvidia, Microsoft, a16z, and others sign a letter defending open-source AI; Jensen Huang, in his first X post, says open models strengthen cybersecurity (Leo Schwartz/The Information)
https://www.theinformation.com/briefings/me…

Meta, Microsoft, Nvidia and Others Sign Letter Defending Open-Source AI
Many of the largest U.S. tech companies signed a joint letter on Friday defending open-source AI software as the Trump administration considers possible restrictions on Chinese labs. The letter—signed by Meta Platforms, Nvidia, Microsoft and Andreessen Horowitz, along with AI startups like Reflection and Hugging Face—argued that open-weight AI models strengthen competition and expand access to

Source: Nvidia has agreed to buy Hugging Face for $12.9B (The Information)
https://www.theinformation.com/articles/nvidia-agrees-buy-open-source-model-repository-hugging-face-12-9-billion

Nvidia Agrees to Buy Open Source Model Repository Hugging Face For $12.9 Billion
Nvidia has agreed to buy Hugging Face, a company known for its GitHub-like repository of open-source AI models, for $12.9 billion, according to a person with knowledge of the agreement. The move will put Nvidia in charge of a strategic asset in the race among open-source model developers to ...

The poster session at #DH2026 was full of delicious snacks *and* hope for the future of digital humanities. Yay!
A random selection of the last four poster photos I took

Hello. Can anyone tell me if the Britannia Panopticon in #Glasgow is open for viewing Thurs or Fri afternoons? I've read that it is but I really had to dig for that information so I'm a bit wary of trusting it. I was lucky enough to see inside 27 or 28 years or so ago and would love to visit again even though I found it a very spooky space.

Something that I wrote a year ago:
"But – wasn’t the word-wide-web designed to be a … web? Where links connect sites & information? I’m sad to say: The big players have succeeded to destroy a part of this ideology by creating walled platforms. And we happlily followed. 😭"
https://www.lock…

Source: Nvidia considered a last-minute OpenRouter bid, telling its leadership it was prepared to make a generous offer, but OpenRouter didn't want to wait (The Information)
https://www.theinformation.com/articles/nvidias-100-billion-dealmak…

Where Nvidia’s $100 Billion Dealmaking Juggernaut Will Go Next
This July, Nvidia caught wind of a looming takeover of OpenRouter, a three-year-old startup that had emerged as a popular marketplace for AI models, including many cheaper open-source ones. Databricks and Stripe had already been talking to the startup about a potential sale. Nvidia was late to ...

Some news - I'll be speaking at the virtual tbd/Con organized by Gazzetta next week
My focus - the scale of #digital #surveillance and censorship in #Tibet, a relevant topic due to the information bla…

Register to join
we want talks that end with open questions. technical depth, social critique, and wild prototypes welcome.

🇺🇦 #NowPlaying on KEXP's #VarietyMix
Fastbacks:
🎵 No Information
#Fastbacks
https://open.spotify.com/track/3ljBzW4U4ndjeiVN7pLiHp

The US, China, and other APEC economies release a statement supporting open models, while emphasizing respect for security, data protection, and IP rights (Evelyn Cheng/CNBC)
https://www.cnbc.com/2026/07/24/china-ai-open-source-apec.html

U.S., other nations back open-source AI with 'strong security' at China summit
The APEC statement is the first to include open-source cooperation at a minister level, said Li Lecheng, China's industry and information technology minister.

@… IMO updating the code would be overkill, to your own point in the book: It's not supposed to be a book about writing react apps in typescript.
If I were in your shoes, I would leave an open issue and/or update the Readme on the example repo, so anyone like me in the future can relatively easily find the information that: The code still works, if you use the…

Alibaba prices Qwen3.8-Max at $2 per 1M input tokens and $6 per 1M output tokens via its API, below Kimi K3's $3/1M input tokens and $15/1M output tokens (Henry Siu/The Information)
https://www.theinformation.com/briefings/alibaba-offers-ne…

Alibaba Offers New Flagship Model at Lower Prices Than Kimi K3
Chinese tech giant Alibaba Group made its new flagship model, Qwen3.8-Max, widely available on Monday through application programming interfaces at lower prices than its Chinese competitor Moonshot AI’s popular Kimi K3 model. Alibaba will also make the open-source model available for download next week. The API for Qwen3.8-Max, Alibaba’s largest-ever model with 2.4 trillion parameters, costs

Check out today's Metacurity for the most critical infosec developments you may have missed since Friday, including
--OpenAI missed rogue agent,
--Hugging Face breach reshapes open-v-closed AI,
--Firms abandon AI lab loyalty,
--Open-weight AI's Kubernetes moment
--China pushes open-weight AI,
--Washington splits on Chinese AI,
--N. Korea expands its intelligence apparatus,
--Hackers hijack hotel Wi-Fi to steal Microsoft 365 accounts,
--DOJ tests privacy limits with GrapheneOS prosecution,
--Malware found in Steam’s biggest indie game,
--Google unveils a simpler threat actor naming scheme,
--Dutch Olympic speed skating venue hit by The Gentlemen,
--Tribeca Festival leak exposed celebrity contact information,
--Companies return to entry-level human hiring in the AI era,
--Comic-Con bans Meta smart glasses,
--AI guardrails fail on mass-casualty prompt
https://www.metacurity.com/ai-watch-openais-hacking-agent-breach-signals-a-fracturing-ai-order/

China-based Moonshot filed confidentially for a Hong Kong IPO; source: the company is in talks to raise a pre-IPO funding round at a $50B valuation (Juro Osawa/The Information)
https://www.theinformation.com/briefings/moonshot-ai-files-confidentially-h…

Moonshot AI Files Confidentially for Hong Kong IPO
Moonshot AI, whose Kimi K3 open-source model has become a global hit, has filed confidentially an application for a Hong Kong initial public offering, Chinese media outlet LatePost reported Wednesday. Moonshot’s path to its IPO hasn’t been easy. The company went through an overhaul to unwind its original corporate structure with an overseas holding company, after Chinese regulators’ new

Wafer, which makes AI agents that optimize open-source models for a business's workload, raised a $40M Series A, a source says at a $200M valuation (Stephanie Palazzolo/The Information)
https://www.theinformation.com/articles/w…

Wafer, An Inference Provider That Uses Non-Nvidia Chips, Lands Acquisition Offers and $200 Million-Plus Valuation
Developers and chip designers are increasingly upbeat about the benefits of using AI to supercharge the process of designing and optimizing AI chips, as my colleague Phoebe wrote about last week. Now, a year-old inference provider, Wafer, is trying a similar approach: using AI to improve the way ...

Hong Kong-listed Z.ai reports H1 2026 revenue up 5x YoY to ~$142M, with "open platform and API revenue" up 28x to ~$122M, and net loss down 12% YoY to ~$308M (Juro Osawa/The Information)
https://www.theinformation.com/briefings/chines…

Chinese AI Model Firm Z.ai’s API Sales Surge in First Half
Chinese AI firm Z.ai’s revenue in the first half jumped nearly fivefold from a year earlier to 954 million yuan ($142 million), boosted by skyrocketing sales of cloud-based AI model services. Z.ai, whose stock is listed in Hong Kong, said its “open platform and API revenue,” which includes the sales of its models through application programming interfaces and other services such as coding

Reframe, which uses AI and industrial robot arms to make modular homes in a factory, raised a $40M Series A extension round led by Energy Impact Partners (Rocket Drew/The Information)
https://www.theinformation.com/articles/excl…

Reframe Raises Funds to Bring Amazon Robotics Know-How to Home Building
For all the talk of robots that look like humans, the next robots to move into people’s houses could look more like ducks. Sales of such a robot, made by open-source model platform Hugging Face, reached more than $2.5 million on its first day on Thursday.Microduck, which costs $399, waddles ...

🇺🇦 #NowPlaying on #BBC6Music's #JarvisCockersSomedayService
Bill Callahan:
🎵 Natural Information
#BillCallahan
https://billcallahan.bandcamp.com/track/natural-information-2
https://open.spotify.com/track/3dsIMfhpMddGl8YyzTN1Sx

Alibaba says its new open-source multimodal model, Qwen3.8-27B, passed 1M downloads within a few days of release, making it one of its fastest-growing models (Juro Osawa/The Information)
https://www.theinformation.com/briefings/alibabas-small-device-model…

Alibaba’s Small, On-Device Model Gains Traction
Alibaba Group said its new small-size AI model that customers can run on their own personal computers has been downloaded more than 1 million times in just a few days since its release, making it one of the company’s fastest-growing models. Alibaba’s new Qwen3.8-27B, a 27 billion-parameter open-source model that understands text, image and video prompts, can be downloaded for free, and users

Between AI slop overwhelming verified information, search engines becoming significantly worse, Cloudflare starting a payment service for service access through AI agents, and age verification for the few trending human users, I fear we are witnessing the end of the open Internet.

COBRA2026: a large-scale multicenter pelvic cone-beam computed tomography projection dataset
Adrian Thummerer, Simon Rit, Florian Kamp, Matteo Maspero, Martjin P. W. Intven, Thomas G. Bon\'e, Christopher Kurz, Guillaume Landry, Thomas Baudier, Mustafa Kadhim, Julius Arnold, Michael Rauter, Barbara Kn\"ausl, Lukas Zimmermann
https://arxiv.org/abs/2607.20037 https://arxiv.org/pdf/2607.20037 https://arxiv.org/html/2607.20037
arXiv:2607.20037v1 Announce Type: new
Abstract: The COBRA2026 dataset is a large-scale, multicenter resource of raw radiotherapy cone-beam computed tomography (CBCT) acquisitions created for the development and evaluation of conventional and learning-based reconstruction and image-correction methods. It contains data from 867 patients undergoing pelvic radiotherapy at six European centers, acquired using Elekta and Varian imaging systems. For each case, the dataset includes raw projection data, acquisition geometry, calibration and correction information, clinically reconstructed CBCT images, and corresponding planning CT images. Vendor-specific files were anonymized and converted into open formats. Planning CT images were deformably registered to the daily CBCT anatomy, and matched projections were simulated using the corresponding acquisition geometry. All cases underwent visual quality control, and cases with substantial processing or registration errors were excluded. The approximately 950 GB dataset is divided into training, validation, and test sets containing 692, 52, and 123 cases, respectively. Projection stacks and volumetric images are provided as compressed MetaImage files, with geometry and metadata supplied in XML and YAML formats. COBRA2026 supports research on full- and sparse-view reconstruction, low-dose imaging, artifact and scatter correction, motion compensation, and synthetic CT generation. The dataset is released under the CC BY-NC 4.0 license, indexed on Zenodo (doi:10.5281/zenodo.21322350), and accompanied by openly available preprocessing and baseline reconstruction code. It also forms the basis of the COBRA2026 reconstruction challenge.
toXiv_bot_toot

The US Department of Commerce's BIS is formally investigating whether Chinese companies like Moonshot are accessing advanced US AI chips to train their models (Leo Schwartz/The Information)
https://www.theinformation.com/articles/u-

U.S. Investigates Chinese AI Companies’ Access to Chips Amid Moonshot Accusations
This week, Trump administration officials have taken to the airwaves and X to blast Chinese open source AI companies like Moonshot, accusing them of stealing intellectual property from the likes of Anthropic and other U.S. AI developers. The administration’s concerns aren’t just rhetoric, ...

Sources: Chinese startup Moonshot AI is seeking formal approval from investors to start a Hong Kong IPO process, which could happen in as soon as six months (Juro Osawa/The Information)
https://www.theinformation.com/briefings/moonshot-ai-seeks-…

Moonshot AI Seeks Investor Approval to Begin IPO Process
Chinese startup Moonshot AI, whose powerful new open-source model is upending Silicon Valley, is seeking formal approval from investors to begin preparations for an initial public offering in Hong Kong, according to two people with knowledge of the matter. The Beijing-based company recently distributed a shareholder resolution regarding its IPO plan, according to the people. The timing of its

Source: Nvidia has discussed an investment in Mercor as part of a round valuing the startup at $20B; Nvidia paid Mercor tens of millions of dollars last quarter (The Information)
https://www.theinformation.com/articles/nvidia-disc…

Nvidia Discusses Funding Its AI Data Supplier Mercor at a $20 Billion Valuation
Nvidia has discussed an investment in Mercor, a data labeling provider that helps the chip designer develop its open-source AI models, according to a person with knowledge of the process. The investment would be part of a $20 billion-valuation round. Existing investor General Catalyst has been ...

Sources: DeepSeek's annualized revenue recently reached $400M to $500M as it seeks to raise ~$7.4B at a ~$74B valuation, up from ~$50B after raising ~$7B in May (The Information)
https://www.theinformation.com/articles/de

DeepSeek’s Annualized Revenue Nears $500 Million, Boosting Fundraise and IPO Plans
Strong revenue growth at DeepSeek is empowering the Chinese open-source AI star to raise its second-funding round of 50 billion yuan ($7.4 billion) and plot a potential public offering in Shanghai next year, according to three people with knowledge of the matter. The three-year-old startup’s ...

DeepSeek launches V4-Pro, its most advanced model that rivals Kimi K3 on some benchmarks at much lower prices, costing $0.44/1M input and $0.87/1M output tokens (Juro Osawa/The Information)
https://www.theinformation.com/briefings/deepseek-r…

DeepSeek Releases Flagship V4-Pro Model in Challenge to Kimi K3
Chinese AI developer DeepSeek on Wednesday launched the official version of its most advanced model, V4-Pro, which offers capabilities comparable to Moonshot AI’s popular Kimi K3 model on some benchmarks at much lower prices. DeepSeek’s latest release further intensifies the competition among increasingly powerful Chinese open-source models and U.S. frontier models from Anthropic and OpenAI. A

GCVE BCP-05-X-03: Bringing Vulnerability Handling Timelines into Vulnerability Records.
Vulnerability records usually provide a good description of what a vulnerability is, which products are affected, how severe it may be, and where additional information can be found.
They are often much less effective at describing what happened, when it happened, and who was involved during the vulnerability handling and disclosure process.
Following many discussions during VulnOptiCON 2026 in Luxembourg, this limitation became particularly clear. Vulnerability analysts, coordinators, vendors and other participants repeatedly discussed how difficult it can be to reconstruct a reliable timeline during vulnerability analysis and coordinated vulnerability disclosure.
To help address this, the GCVE initiative has published a new extension:
GCVE BCP-05-X-03 - Vulnerability Handling and Disclosure Timeline.
The extension provides a structured and machine-readable way to represent the lifecycle of a vulnerability from discovery and reporting through acknowledgement, validation, remediation and public disclosure. It's already live in our open-source tool-set.
🔗 Blog post #opensource #openstandard #cve #gcve #cra #cybersecurity #vulnerability

AI providers like Google, OpenAI and Anthropic don’t publish detailed energy‑use data for their models, viewing it as commercially sensitive operational information.
Now the Sustainable AI Group (SAIG) has developed a way to help customers compare the energy use of closed models by comparing them with open models of similar capability.
#claude #gpt #gemini #emissions #saig #energy #ai #technews

Tired of drafting vulnerability advisories from Git patches?
We developed patch2vuln to facilitate the creation of security advisories directly from Git patches.
With a single command, patch2vuln can assist an analyst throughout the advisory creation process: analyzing the patch, drafting the vulnerability title and description, identifying CWE and CAPEC mappings, proposing a CVSS v4.0 vector, extracting affected versions, remediation information and credits, and generating a structured CVE/GCVE record.
The analysis can run entirely locally using an LLM via Ollama, while deterministic processing is used for elements such as CVSS scoring and structured output validation.
The goal is not to replace the security analyst, but to remove much of the repetitive work and provide a solid structured draft for human review and publication.
patch2vuln v1.0 is now available as open-source software under the @… Lab initiative.
#VulnerabilityManagement #CVD #CVE #GCVE #OpenSource #CyberSecurity #AI