1. Beginning January 2026, all ACM publications will be made open access
Total comment counts : 25
Summary
error
Overall Comments Summary
- Main point: The discussion centers on open-access publishing and ACM’s transition to OA, comparing APC-based, library-funded models like Subscribe to Open (S2O), and their implications for authors, institutions, and readers.
- Concern: The shift could burden authors or libraries with high costs, create inequities, and raise licensing and quality issues (e.g., non-retroactive CC licenses, AI-generated summaries, and paid “premium” features).
- Perspectives: Opinions range from embracing OA via S2O and low/no author fees to criticizing high APCs and profit-driven elements, with debates over licensing, archival access, and the value/accuracy of AI-assisted tools.
- Overall sentiment: Mixed
2. We pwned X, Vercel, Cursor, and Discord through a supply-chain attack
Total comment counts : 24
Summary
Daniel, a 16-year-old bug hunter, details uncovering critical XSS flaws in Mintlify during Discord’s switch to the platform. He found endpoints that let an attacker fetch arbitrary Markdown and static files from any Mintlify subdomain, risking injected scripts in documentation and credential theft. By serving an SVG with embedded JavaScript via Discord’s path, he demonstrated executable code in a user context. He shared the discovery with friends and documented the process using Mintlify CLI, illustrating how a single misconfigured route enabled cross-site scripting.
Overall Comments Summary
- Main point: Discussion centers on the security risks of SVG files that can contain scripts, highlighting the tension between their benefits as scalable graphics and the potential for widespread vulnerabilities.
- Concern: The main worry is that SVGs can enable XSS, RCE, phishing, or drive-by downloads, compromising users and eroding trust in shared platforms.
- Perspectives: Viewpoints range from valuing SVGs for their visuals while criticizing modern tech’s security culture and bug-bounty incentives, to proposing mitigations like sanitization, rasterization, and CSP, and to debating responsible disclosure and industry practices.
- Overall sentiment: Mixed
3. GPT-5.2-Codex
Total comment counts : 32
Summary
error
Overall Comments Summary
- Main point: The discussion centers on Codex/GPT-5.x’s coding capabilities, practical workflows, and security/privacy implications, with hands-on usage and cross-model comparisons.
- Concern: A major concern is privacy and security risks, including persistent prompts and code diffs and the potential for dual-use misuse.
- Perspectives: Views range from enthusiastic praise of Codex’s debugging, code quality, and workflow benefits to skepticism about safety, performance inconsistencies, and deployment trade-offs, with frequent model comparisons.
- Overall sentiment: Mixed
4. Texas is suing all of the big TV makers for spying on what you watch
Total comment counts : 22
Summary
error
Overall Comments Summary
- Main point: The discussion centers on privacy and ad-targeting in smart TVs and the idea of a trustworthy “DUMB” (Don’t Upload My Bits) certification or other ways to avoid data collection.
- Concern: The core worry is pervasive surveillance and data collection by TV makers and firmware, with limited consumer control and potential regulatory or market pushback not yet solving the issue.
- Perspectives: Viewpoints range from wanting a fully open, Linux-based or non-networked dumb TV, to seeking hardware/software blockers or third-party filters, to faith in regulation or industry changes, and some arguing the market may not support non-surveillance devices.
- Overall sentiment: Mixed
5. Skills for organizations, partners, the ecosystem
Total comment counts : 31
Summary
Claude’s “skills” now support organization-wide management, making deployment and discovery easier. Admins can centrally provision skills from Admin Settings; they’re enabled by default but users can switch off individual skills. Creating skills is simplified: describe the task, upload skill folders, or use the skill creator, with previews showing full content. A growing catalog of partner-built skills—Notion, Canva, Figma, Atlassian, Vercel, Sentry, Zapier MCP, and more—is available at claude.com/connectors, enabling instant use without custom development. Skills are an open standard, portable across AI platforms. Prerequisites include Code Execution and File Creation.
Overall Comments Summary
- Main point: There is a wide-ranging debate about what ‘Skills’ in Agent/MCP actually are, how they relate to prompts, and whether standardizing them is useful or premature amid hype and fragmentation.
- Concern: The main worry is that standardization could become marketing-driven, prematurely lock in design choices, and fail to deliver genuine interoperability, while also overlooking licensing and the ‘data as code’ implications.
- Perspectives: Opinions span from viewing Skills as a superficial repackaging and marketing gimmick to seeing them as a potentially valuable, flexible foundation for interoperable agent behavior, with many arguing for careful, adaptable standards rather than rigid, one-size-fits-all solutions.
- Overall sentiment: Mixed
6. How China built its ‘Manhattan Project’ to rival the West in AI chips
Total comment counts : 13
Summary
In a high-security Shenzhen lab, Chinese scientists have built a prototype capable of producing cutting-edge semiconductor chips, potentially challenging Western dominance. Finished in early 2025 and now testing, the machine fills nearly an entire factory floor. It was developed by former ASML engineers who reverse-engineered ASML’s extreme ultraviolet lithography (EUV) systems. EUV lithography uses extreme UV light to etch circuits thousands of times thinner than a human hair onto silicon wafers, a capability currently monopolized by the West, signaling a new phase in tech competition.
Overall Comments Summary
- Main point: The discussion centers on whether China can replicate ASML’s EUV lithography ecosystem and its implications for global semiconductor supply and geopolitics.
- Concern: The concern is that Western retreat from consumer GPUs and potential Chinese progress in EUV tech could worsen supply to hobbyists and professionals and heighten geopolitical risk.
- Perspectives: Perspectives range from viewing China’s EUV progress as a meaningful milestone that could alter tech parity to doubting it can replicate ASML’s ecosystem, with some criticizing sensationalist framing.
- Overall sentiment: Mixed
7. T5Gemma 2: The next generation of encoder-decoder models
Total comment counts : 2
Summary
Google’s T5Gemma 2 is a major upgrade of the encoder–decoder family built on Gemma 3. It introduces the first multi-modal, long-context encoder–decoder, uses tied word embeddings, and merges decoder self- and cross-attention to save parameters. Available in compact pre-trained sizes: 270M–270M (~370M total excluding vision encoder), 1B–1B (~1.7B), and 4B–4B (~7B) parameters, ideal for rapid experimentation and on-device deployment. It extends T5Gemma by initializing from a decoder-only model and continuing pre-training, now with vision-language capabilities. Pre-trained checkpoints are released for developer post-training.
Overall Comments Summary
- Main point: Discussion centers on not releasing post-trained/IT checkpoints, the potential usefulness of a 540M multimodel model, and seeking clarification on encoder-decoder models.
- Concern: Not releasing checkpoints may hinder users who cannot post-train themselves, and “just post-train it yourself” isn’t always a feasible option.
- Perspectives: Some participants want a useful mid-sized model and clearer terminology, while others emphasize protecting Gemma and avoiding cannibalization by withholding post-trained checkpoints.
- Overall sentiment: Mixed
8. Classical statues were not painted horribly
Total comment counts : 45
Summary
Ancient Greek and Roman sculpture often survives in excellent condition and was frequently painted, though modern viewers tend to treat it as white marble. The article notes abundant high-quality works in major museums and Pompeii, and mentions the Augustus of the Prima Porta with surviving traces of pigment. It discusses contemporary color-reconstructions (e.g., Brinkmann’s Gods in Color) that many find aesthetically jarring, prompting the idea that taste has shifted since antiquity and that chromophobia shapes our response. Yet ancient depictions and some surviving remains show colored surfaces, with varied, delicate finishes, not monochrome.
Overall Comments Summary
- Main point: There is a heated debate about whether ancient statues were painted and how reconstructions should be done, balancing evidence, painting conventions, and public communication.
- Concern: Conservation doctrines and reliance on limited evidence may produce unattractive or misleading reconstructions that misrepresent the originals.
- Perspectives: Some advocate color-based reconstructions reflecting painting practices, others criticize outsiders’ speculation and urge consulting the actual practitioners, and still others defend cautious, evidence-based approaches while acknowledging uncertainty.
- Overall sentiment: Mixed
9. FunctionGemma 270M Model
Total comment counts : 7
Summary
Google releases FunctionGemma, a specialized Gemma 3 270M tuned for function calling, enabling private, on-device actions and offline task execution. It serves as a base for custom, fast local agents and can route complex tasks to larger models like Gemma 3 27B. It aims to transform from conversational AI to acting agents, with Google AI Edge Gallery demos showing offline commands and a game, all run locally in browser or on-device. Developers can fine-tune with a cookbook and load onto mobile devices. Emphasizes private, ultra-fast edge experiences.
Overall Comments Summary
- Main point: The discussion centers on a Google model release and evaluating whether the model can be used as a command-line tool to call external APIs, with references to Ollama and n8n.
- Concern: The main worry is about the practicality and safety of enabling direct API calls from the model and integrating it into workflows, including potential security and reliability issues.
- Perspectives: Views range from excitement about rapid Google shipping and potential API-call capabilities to skepticism about practicality, safety, and feasibility of integrating with tools like Ollama and n8n.
- Overall sentiment: Cautiously optimistic