A developer involved in a Banjo-Kazooie PC port criticized the growing use of AI coding tools such as Claude in fan-made retro game ports. He argued that these projects often produce hard-to-maintain code, stall because developers lack domain knowledge, and may offer few advantages over emulation.
Speaking at CoreWeave's Fully Connected event, Brandon Hootman, Caterpillar's VP of physical AI platforms and construction autonomy, and Richard Ahlfeld, CoreWeave's SVP of physical AI, described their collaboration on autonomous construction equipment. Caterpillar's digital ecosystem already holds about 18 petabytes of federated data from machines, dealers and customers, while a single machine can generate terabytes of lidar, camera, control and performance data in a day. Working with Nvidia, the partners use AI models to annotate and label incoming field data, cutting work that used to take weeks or months down to hours so the simulation and training loop closes within a workday. CoreWeave named Caterpillar among its enterprise customers in its second-quarter results and launched a Physical AI Field Engineering service in September.
Researchers from iLearn-Lab at Harbin Institute of Technology (Shenzhen) and LV-Lab at the National University of Singapore propose QuantWM, a training-free framework for 2-bit KV Cache quantization in video world models. The team says its query-sensitive clustering and low-rank attention compensation improve visual stability while achieving up to 6.20x KV Cache compression across five evaluated models.
Maket has launched Maket 2.0, an AI platform that combines floor plan generation, plan editing and 3D visualization in one workspace. Users describe the home they want in a chat window — for example a 3,000-square-foot townhouse — and an assistant asks follow-up questions about floors, outline and room counts before generating a plan that can be adjusted via chat or by resizing rooms on the canvas; existing plans can be uploaded as an image or PDF and converted into an editable project. Rooms can then be rendered in 3D using built-in interior styles such as Minimalist or Scandinavian, or a user-supplied reference image, and plans can be exported as PDF or DXF. Maket 2.0 is available now on the web: a free plan offers 30 one-time credits, the Homeowner plan costs $20/month for 300 credits and the Pro plan $100/month for 3,000 credits, with generating or uploading a plan using 20 credits, a render 10 and a layout optimization 5.
xAI says Grok 4.7 is now available through Microsoft Foundry. The source does not provide details about access, capabilities, pricing, or rollout scope.
Anthropic is expanding its Cyber Verification Program (CVP) and folding the separate Project Glasswing into it as a single offering with three tiers — Defense Access, Red Team Access and Specialized Access — giving vetted security teams reduced safety blocks on Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1. On Anthropic's CyScenarioBench, Claude Opus 5.5 was blocked on all 50 trials without CVP access, 46 of 50 under Defense Access, and none under Red Team Access, where it completed 34 of 50 — equivalent to the 67.6% success rate with no safeguards. Anthropic also says Glasswing partners found at least 129,000 verified software vulnerabilities between April and July 2026, more than 33,000 of them critical or high severity, plus 5,500 from its own open-source scanning; it calls these numbers a significant undercount.
A 36Kr analysis examines reports that ChatGPT conversation titles sometimes append gambling, adult-advertising fragments, foreign words, or other malformed text. Drawing on user reports and an EMNLP paper about Chinese training-data pollution in OpenAI’s o200k_base tokenizer vocabulary, it hypothesizes that a title-generation model sometimes fails to stop and then emits poorly understood tokens; OpenAI has reportedly acknowledged the issue but has not publicly explained its root cause.
Anthropic reinforcement-learning lead Sholto Douglas said in an interview that a model capable of doing all computer-based work could emerge within a few years. He also projected that annual AI capital spending could reach $4 trillion by 2028 and that global GDP could double in the early 2030s, while colleague Nick Marwell warned of the risks once AI no longer needs human partners.
DeepSeek is reportedly close to securing at least RMB 80 billion in a new financing round, according to Bloomberg-cited sources, with the final amount potentially approaching RMB 100 billion. The round was initially planned at about RMB 50 billion and is being linked to preparations for a possible Shanghai STAR Market IPO in early 2027, though the transaction has not been finalized.
Atlassian and OpenAI are expanding their partnership, with OpenAI frontier models powering agents across Atlassian’s platform and Rovo. The agreement also expands Atlassian’s access to models including GPT-6 Astra and the GPT-5.6 series, while connecting ChatGPT and Codex to Atlassian work data through plugins and the Teamwork Graph.
Google introduced the Developer Knowledge API ecosystem, a structured programmatic interface to official documentation for Google Cloud, Firebase, Android and more, aimed at AI agents, IDE extensions and automated workflows rather than web scraping. The lineup includes gcloud CLI commands (answer-query, documents search-chunks, documents describe), an official agent skill that connects via MCP to Google Antigravity, Claude Code, Cursor, GitHub Copilot and custom frameworks, client libraries for C#, Go, Java, Node.js/TypeScript, PHP, Python and Ruby, and an APIs Explorer for interactive testing. Google says the API is indexed frequently to keep retrieval fresh and reduce hallucinations caused by stale LLM training cutoffs.
Reporting from the Australian parliament, Victoria Kim quotes OpenAI chief strategy officer Mr. Kwon saying that since the Medicare breach the company has added monitoring that allows staff to "immediate intervention" to stop training if its models access the internet in ways they are not supposed to. The quote was excerpted on Simon Willison's blog under AI-security tags.
A report says PS5 modders have used AI tools to install emulators rapidly as jailbreaks spread across social media. Games requiring newer firmware remain inaccessible, while Sony and hackers continue their apparent cat-and-mouse struggle.
The World Bank raised Vietnam's 2026 growth forecast by 1.1 percentage points to 7.4%, as high-tech manufacturing and exports benefit from surging global AI investment. Finished computers, servers and routers make up about 60% of Vietnam's AI-related exports, and AI-related goods accounted for more than 70% of recent export growth in Vietnam, Malaysia, the Philippines and Thailand. Vietnam's AI-related exports grew from roughly 20% of GDP in 2023 to about 32% in 2025, though the country sits in assembly rather than leading-edge fabrication, leaving it exposed if AI hardware spending slows.
An Android Authority writer describes using Claude to create a dental-practice logo and signage, initially producing attractive mockups and an adjustable design tool. The files developed jagged curves, inconsistent typography, and inaccurate fabrication dimensions, so a human designer had to manually redraw and adapt the final design.
Simon Willison released llm-openai-decisions 0.1a0, an LLM plugin for OpenAI's new Decisions API announced at last week's DevDay; he had GPT-6 Astra read the new API docs and modeled the plugin on his existing llm-typesafe plugin for Jev. The new decision model, gpt-6-luna, accepts image input as well as text. According to the post, both models charge for input only and not output — OpenAI at 10 cents per million input tokens versus Jev's 4.2 cents — and both support yes/no, choice and score question types.
Google began rolling out native editing for .md and .markdown files on Oct. 5 for all Workspace customers and personal Google accounts, so files opened from Drive stay Markdown instead of being converted into Google Docs; users can share, comment and co-edit them, and Drive renders previews with formatted tables and clickable links. Google tied the update to AI workflows in its Workspace announcement, noting Markdown is widely used by large language models to preserve headings, lists and tables and that native support lets people and software agents work on the same files without repeated conversions. Some Docs formatting does not survive: smart chips become text or links, HTML becomes plain text, and font colors, highlights and alignment are removed; rollout can take up to 15 days.
Melius, an AI platform for generating advertising campaigns, images, and videos, announced $25 million in total funding, including a $20 million Series A led by CRV and a $5 million seed round led by General Catalyst. The startup says it abandoned its original performance-marketing product and rebuilt the company around tools for generating creative assets and campaigns.
Vinci4D Inc. announced a $250 million Series B round at a $1.5 billion valuation, co-led by Advent, Temasek and Xora, with participation including AMD Ventures. Its platform is built around a physics-focused AI model that turns uploaded CAD drawings into simulations, automates meshing and uses custom GPU kernels to speed convergence; the company says it can complete simulations with hundreds of millions of degrees of freedom (DOFs) in minutes, versus the hours or days such workloads normally take. Vinci's initial focus is thermal simulation for chips—processors, memory modules and interconnects—with plans to extend to vehicles, aircraft and satellites.
Penguin Mail is an open-source Rust email and calendar client for Linux, supporting Gmail, Microsoft, IMAP and POP3 accounts. Its optional assistant can run locally through LM Studio or Ollama, uses app tools to query mail, and asks for confirmation before sending messages or changing settings.