Hacker News
1Advertise in ChatGPT
- ChatGPT now offers advertising opportunities that let businesses reach people during decision-making moments through relevant, contextual ads. Advertisers can create campaigns using Ads Manager, targeting users as they explore options and compare choices. The platform emphasizes user trust with clearly labeled ads that remain separate from ChatGPT responses, while early advertisers like Best Buy report positive results from this AI-native advertising approach.
Kagi search vs. ad-supported models
Vlad from Kagi promotes their "you are not the product" philosophy, emphasizing user-funded search over ad-supported alternatives. Discussion includes praise for Kagi's mental clarity benefits, but controversy over their partnership with Yandex amid Ukraine war concerns.OpenAI's advertising introduction concerns
Users express alarm about ChatGPT introducing ads, comparing it to Google's gradual degradation from clearly labeled ads to hidden promotional content. Many fear this marks the end of AI's "golden age" and the beginning of manipulation through subtle product placement.Ad-supported vs. paid AI services debate
Discussion about whether society should ban advertising entirely, with concerns that AI agents delivering disguised ads could be more manipulative than traditional advertising. Users debate the sustainability of subscription models versus ad-supported free tiers for AI services.
2Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
- Kimi K3 AI model ranks second only to Fable 5 on the AA-Briefcase benchmark, demonstrating strong performance in agentic knowledge tasks according to analysis from Artificial Analysis.
Benchmarking vs real-world performance
Users debate whether Chinese models like Kimi K3 are "benchmaxxed" - performing well on tests but poorly in practice. Many report token inefficiency and slow performance, though some find them comparable to previous-generation frontier models like Opus 4.8 and GPT-5.5.Oracle routing methodology
Discussion centers on Fireworks' "oracle routing" approach, which retroactively chooses the cheapest correct model after running tasks through both. Critics argue this theoretical framework doesn't represent realistic routing capabilities.Chinese model adoption concerns
Users express hesitation about Chinese models due to data governance issues, privacy controls, and training on user content. Some prefer "western" providers or zero data retention options despite potential cost savings.
3FreeInk: Open ecosystem for e-readers
- Free Ink is an open ecosystem designed for e-readers, providing an accessible platform for electronic reading devices and related services.
Xteink X4 user experiences
Users share positive feedback about the Xteink X4 mini e-reader, praising its portability and simplicity. Discussion covers firmware options like Witch reader and CrossPoint, methods for transferring Kindle books via DRM removal, and appreciation for the device's pocketable size and weeks-long battery life.DRM and book ownership concerns
Users discuss the ethics and legality of removing DRM from purchased e-books, sharing examples of Amazon remotely deleting books like 1984. The conversation explores the difference between ownership and licensing, with some advocating for piracy when "buying isn't ownership."Alternative e-reader platforms
Discussion of various e-reader options including Kobo devices with KOReader, Boox Android-based readers, and requests to port open firmware to older Kindles. Users compare features like battery life, screen size preferences, and the benefits of hackable vs proprietary systems.
4'VPNs are lawful technical tools,' says EU Court in landmark copyright ruling
- The EU Court of Justice ruled that VPN providers are not liable for copyright infringement when users bypass geo-blocking restrictions. The landmark case involved Anne Frank's diary manuscripts and established VPNs as "lawful technical tools." The court determined that publishers using adequate geo-blocking technology cannot be held responsible if users circumvent restrictions with VPNs, placing legal burden on content providers rather than VPN companies.
Copyright disputes over Anne Frank's diary
Users debate the absurdity of copyright extending decades after an author's death, with Anne Frank's estate pursuing legal action. Discussion covers how extended copyright terms don't incentivize creation and questioning whether heirs should profit from works they didn't create.VPN legality and internet freedom
Commenters explore how this ruling might protect VPN usage against future legal challenges, particularly regarding age verification laws. Discussion includes concerns about government overreach in controlling internet access and the shift toward more private, decentralized online communities.EU regulation and internet control
Users critique European approaches to internet governance, citing examples like France blocking Polymarket without judicial review. Discussion covers the struggle between maintaining sovereignty over digital spaces while preserving open internet principles.
5OverpAId – Fire your CEO. Hire the future
- OverpAId is a satirical AI product positioned as a "Chief Executive Replacement Engine" that costs $4,699 versus a CEO's $22 million annual salary. The site critiques massive CEO compensation growth while workers face layoffs justified by AI efficiency, pointing out the irony that if AI can replace workers, it should logically replace executives first.
CEO role and value
Discussion centers on what CEOs actually do versus their compensation. Some argue CEOs serve as corporate delegates and relationship managers rather than operational leaders, while others debate whether "proven talent" justifies high pay or if many executives are overpaid fraudsters who benefit from luck and connections.Remote work vs return to office
Debate over RTO motivations, with some claiming it's about real estate valuations while others argue it's about managerial power and control. C-suite members claim productivity drops with remote work, while employees see RTO as surveillance rather than business necessity.AI replacing executives
Discussion of whether AI could effectively replace CEO functions, with some seeing executives as prime targets for automation due to their high cost and routine decision-making, while others note the importance of human accountability and relationship-building in leadership roles.
6Show HN: Bento - An entire PowerPoint in one HTML file (edit+view+data+collab)
- The team created Bento, a self-contained HTML slide presentation tool that works offline without installation or cloud login. It features editing, presenting, printing, saving, and live collaboration capabilities. Users can share files via email or Airdrop, and the tool can convert PowerPoint files using AI assistants. The project is open source and MIT licensed.
Single-file web applications
Discussion around the concept of self-contained HTML files that include both application logic and data. Users shared examples like TiddlyWiki, mdwiki, and other tools, with interest in creating a Wikipedia page for this category.Technical implementation details
Creator explained Bento's architecture: JSON slide data embedded in HTML, base64-compressed app code, CRDT for collaboration via encrypted Cloudflare Durable Objects relay, and seamless file upgrades.Reveal.js vertical slides feature
Users discussed the loss of Reveal.js's vertical slide navigation feature, which allows creating layered presentations with different detail levels for different audiences by navigating up/down within sections.
7Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
- A federal judge approved a $1.5 billion settlement where AI company Anthropic will pay authors about $3,000 per book for using pirated copies to train its Claude chatbot. About 91% of over 482,000 affected books have been claimed for payment in what's called the largest copyright recovery in history.
Effectiveness of $1.5B settlement
Many commenters argue the settlement is inadequate - a "slap on the wrist" that doesn't address ongoing AI training. Some propose royalty systems for regurgitated content, while others note this creates a moat favoring big tech companies who can afford such settlements.Legal distinction between piracy and training
Judge Alsup ruled that feeding books into LLMs is transformative fair use, but keeping pirated copies in a central library is infringement. The settlement addresses the piracy, not the training itself, creating precedent for future AI development.Comparison to past copyright enforcement
Users express outrage at the disparity between this settlement and harsh treatment of past infringers like Kim Dotcom and Aaron Swartz, viewing it as evidence that wealthy corporations receive preferential treatment in copyright cases.
8Terrence Tao's ChatGPT Conversation about the Jacobian Conjecture Counterexample
- This appears to be a ChatGPT interface page related to the Jacobian Conjecture Counterexample topic. The page shows standard ChatGPT navigation elements including options for new chat, search, images, chat history, plugins, and account settings. It displays the typical ChatGPT disclaimer about AI usage, terms of service, and data usage policies.
LLM reasoning capabilities and the "keep going" technique
Users discovered that repeatedly prompting LLMs with "keep going" can lead to breakthrough results, as LLMs don't inherently know when problems are impossible. This technique helped prove conjectures and find counterexamples.Mathematical complexity and accessibility
Discussion of how mathematics has uniquely dense and impenetrable nomenclature compared to other technical fields, making it difficult for non-mathematicians to understand even basic concepts.AI as intelligent colleague vs tool
Debate over whether LLMs demonstrate true intelligence, with the conversation showing the AI acting more like an equal colleague to Tao rather than just a computational tool.
9John C. Dvorak has died
Career and legacy
Commenters fondly remember Dvorak's influential tech journalism career spanning PC Magazine, InfoWorld, and Mac User. His small thumbnail photo in PC Magazine represented "gravitas" for many young readers, and his bold, provocative takes shaped tech discourse for decades.Media appearances and relationships
Discussion centers on Dvorak's memorable TV and podcast appearances, particularly his dynamic with Leo Laporte on TWiT and Tech TV. Their playful antagonism and equipment borrowing became legendary, though they eventually had a falling out.Personal remembrances
Multiple commenters share personal connections to Dvorak's work, from transcribing articles to Usenet to enjoying dinners together. Despite his curmudgeonly public persona, he's remembered as warm-hearted and genuinely passionate about computing.
10Apple defeats liability for not scanning iCloud for CSAM
- A court dismissed a lawsuit against Apple over its failure to scan iCloud files for child sexual abuse material, ruling that Section 230 protects Apple from liability. Plaintiffs argued Apple should use detection technology like PhotoDNA, but the court found this would treat Apple as a content publisher. The judge expressed concern about the legal framework's inadequacy in protecting CSAM victims while noting that mandatory scanning would compromise end-to-end encryption and privacy for all users.
CSAM enforcement vs. actual child abuse prevention
Discussion argues that authorities focus heavily on CSAM detection/scanning after abuse occurs, but invest little in preventing actual child sexual abuse. Many see this as misplaced priorities that don't protect children effectively.Apple's privacy stance and client-side scanning controversy
Debate over whether Apple genuinely supports privacy or uses it for marketing. Discussion covers Apple's abandoned NeuralHash client-side scanning plan and whether true end-to-end encryption is possible when companies control both endpoints.CSAM laws and teenage sexting
Participants argue that most "CSAM" cases involve consensual content created/shared by teenagers, not actual predatory abuse. Current laws can criminalize normal teenage behavior while failing to address real exploitation of vulnerable children.