Hacker News
1Elon Musk has lost his lawsuit against Sam Altman and OpenAI
- Elon Musk lost his lawsuit against OpenAI co-founders Sam Altman and Greg Brockman, with nine California jurors unanimously ruling his claims were filed too late under statute of limitations. Musk had accused them of "stealing a charity" by creating a for-profit affiliate, seeking $78.8-135 billion in damages. He plans to appeal the decision.
Statute of limitations ruling
Musk lost because the jury found he waited too long to file his claims. The 2019/2021 Microsoft deals were too similar to the 2023 deal he sued over, meaning he should have brought the case earlier. Appeals are unlikely since jury fact-finding is rarely overturned.Legal technicality vs substance
The case was dismissed on statute of limitations grounds before addressing whether OpenAI actually breached its charitable mission. Many felt this left unresolved questions about OpenAI's transformation from nonprofit to for-profit entity.OpenAI's nonprofit-to-profit transition
Discussants questioned whether OpenAI's transfer of IP from its nonprofit to for-profit entity sets concerning precedent for exploiting tax-deductible donations for R&D, then commercializing results through a separate for-profit structure.
2Show HN: Files.md – Open-source alternative to Obsidian
- Files.md is a simple, local-first markdown note-taking application that stores notes, documents, journals, checklists and tasks in plain .md files without sending data to servers. It works offline in browsers, offers optional sync via cloud services or self-hosted servers, and emphasizes simplicity over complex features to foster genuine thinking and knowledge building rather than just collecting information.
Open source vs proprietary software
Users debate Obsidian's closed-source nature despite feeling "open source-like" due to markdown storage and plugin ecosystem. Discussion covers monetization models, data ownership, and whether developers deserve profit from proprietary software.Alternative note-taking tools and approaches
Community shares various markdown editors and knowledge management tools including native Qt implementations, VS Code extensions, TiddlyWiki, and self-hosted solutions, comparing features and philosophies.Second brain methodology effectiveness
Debate over "second brain" approaches, with criticism of Tiago Forte's "remember nothing" philosophy versus using external notes to strengthen rather than replace human memory and learning.
3Garry Tan, the CEO of YC, accused me of unethical reporting
- Garry Tan promoted a book by SF reporter Dion Lim, claiming she was unfairly targeted for exposing anti-Asian crimes. Journalist Radley Balko disputes this narrative, explaining he fact-checked Lim's viral carjacking story that falsely claimed charges were dropped against a juvenile defendant. Balko argues his reporting was legitimate journalism, not a coordinated attack, and that Lim had pressured sources based on inaccurate information from police sources.
Progressive prosecutor competency issues
Discussion centers on how progressive prosecutors like Chesa Boudin and Kim Foxx failed due to basic management problems rather than ideology - losing veteran staff, poor promotion systems, and inability to execute day-to-day operations effectively.Wealth and power corruption
Commenters debate whether extreme wealth causes mental breakdown and need to control others, or if the journey to wealth selects for people with sociopathic traits who exploit systems and refuse to help others.Journalistic objectivity and bias
Discussion about whether reporters can maintain objectivity while also acting as commentators, with debate over how all journalism involves selective fact presentation and inherent bias regardless of stated neutrality.
4AI is a technology not a product
- The author criticizes Steven Levy's article arguing Apple needs a "killer AI product," calling it hype-driven nonsense. They argue AI will be pervasive technology integrated into existing devices like phones rather than requiring revolutionary new products, comparing it to wireless networking that's now embedded everywhere.
AI should be a feature, not a product
Users want practical AI improvements like making Siri actually work for basic tasks (calendar events, music playback) rather than flashy AI products. Companies are pushing random AI features without considering user needs or workflows, when the real value lies in seamlessly integrating AI into existing products to solve actual problems.Working backwards from customer experience
Steve Jobs' philosophy of starting with user needs rather than technology still applies to AI development. Apple's strength lies in focusing on customer experience over technological showcasing, which may position them well as AI becomes commoditized and the focus shifts back to product integration.Skepticism about AI agent futures
Many question whether people actually want AI agents handling routine tasks like ordering rides or making grocery lists, as these activities are often enjoyable parts of life rather than drudgery to automate. The vision of always-on AI agents may appeal more to tech executives than regular users.
5Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep
- Stephan and Thomas open-sourced Semble, a tool that solves token efficiency problems when using Claude Code on large codebases. It combines static embeddings with BM25 for fast code retrieval, using 98% fewer tokens than grep+read while maintaining 99% retrieval quality and being 200x faster than transformers.
Token usage vs. actual performance
Multiple users tested tools like semble and found that while they reduce individual search tokens, overall conversation costs often increase due to more turns needed. Some found simple PROJECT.md files work better than specialized tools.AI models' preference for familiar tools
Discussion reveals models are heavily trained on grep and don't trust results from newer tools, leading to retries that negate token savings. Users report mixed success getting models to consistently use LSPs or other alternatives.Real-world benchmarking challenges
Critics note the industry lacks proper end-to-end agent evaluation. Tools often show impressive isolated metrics but fail to demonstrate actual improvement in complete coding tasks, with some making agents perform worse overall.
6GenCAD
- GenCAD is an AI model that generates 3D CAD models and their complete parametric command sequences from images. Unlike traditional approaches using meshes or point clouds, it produces true CAD programs that maintain engineering accuracy and modifiability for manufacturing and design.
Reliability and practical limitations
Users report the system only works ~60% of the time even on training data, with many failed attempts on simple drawings. The model is limited to isometric, noise-free CAD images and only supports basic pad operations, excluding spheres, fillets, and other common features.Utility and real-world applications
Debate over whether this solves actual problems, as CAD's difficulty lies in dimensioning, constraints, and tolerances rather than basic modeling. Some see potential for carpentry/furniture design, while others question if it addresses genuine workflow needs.Alternative approaches with LLMs
Multiple users successfully use OpenSCAD with LLMs like Claude for parametric modeling, suggesting text-based CAD generation may be more practical than image-to-CAD conversion for current AI capabilities.
7I turned a $80 RK3562 Android tablet into a Debian Linux workstation
- rkdebian is a build system that creates a bootable Debian 12 image for the Doogee U10 Android tablet with Rockchip RK3562 SoC, allowing users to run full Debian from an SD card without modifying internal storage or unlocking the bootloader, with most hardware features working including display, touchscreen, Wi-Fi, audio, and NPU for local LLM inference.
AI-assisted reverse engineering
Users discuss how AI tools like Claude and ChatGPT accelerated hardware reverse engineering and driver development. Some debate whether AI-generated content should be allowed, while others share experiences using AI for device porting and firmware analysis.Hardware capabilities and software compatibility
Discussion centers on what software runs well on 4GB RAM tablets, browser performance, lightweight desktop environments, and the potential for repurposing old mobile hardware for development work and edge computing applications.Boot process and device availability
Technical discussion about how Rockchip devices boot from SD cards before eMMC, bypassing locked bootloaders. Users also discuss device availability, pricing concerns, and interest in similar projects for other tablets.
8We stopped AI bot spam in our GitHub repo using Git's –author flag
- The Archestra team faced an overwhelming influx of AI bots spamming their GitHub repository with low-quality pull requests and comments, drowning out legitimate contributors. After receiving 253 comments on a single bounty issue and 27 untested PRs for another, they implemented a nuclear solution requiring contributor onboarding and whitelisting to protect their open source project from AI-generated noise that was demotivating real developers.
Security implications of contributor status
Discussion of how GitHub's contributor approval system creates security vulnerabilities, where malicious actors can gain elevated privileges by getting simple changes merged first, then bypass approval requirements for future malicious PRs.GitHub's role in AI spam problem
Users criticize GitHub/Microsoft for not implementing better spam prevention measures, arguing the company has no incentive to block AI-generated content since it benefits their Copilot business model and server usage.Rating and reputation systems for contributors
Extensive debate about implementing Elo-like scoring systems to rank contributors based on past contributions, with concerns about manipulation, gaming by bots, and barriers for legitimate new users.
9Eric Schmidt speech about AI booed during graduation
- Former Google CEO Eric Schmidt was booed during his University of Arizona commencement speech when discussing artificial intelligence. He acknowledged graduates' fears about AI taking jobs and inheriting societal problems, arguing the future remains unwritten and they can shape AI's development. The negative reception reflects growing concerns about AI among young people.
Schmidt's comparison of AI to immigrants
Commenters criticized Schmidt's attempt to connect AI acceptance with pro-immigration values as a manipulative rhetorical trick, calling it a false equivalency designed to shame critics into supporting AI by associating opposition with anti-immigration sentiment.Student backlash and generational divide
Discussion revealed a stark divide between tech executives promoting AI and younger generations who distrust it, with students booing Schmidt at graduation. Many noted the irony of executives urging AI adoption while simultaneously threatening job cuts.Optimistic vs pessimistic AI futures
Debate over whether AI leads to post-scarcity abundance or permanent underclass on UBI. Optimists envision technological deflation making goods cheap; pessimists fear corporate dystopia where billionaires control automated economy while masses become expendable.
10Anthropic acquires Stainless
- Anthropic is acquiring Stainless, a company that builds SDKs and developer tools, to enhance Claude's ability to connect to external systems and data. Founded in 2022, Stainless has powered Anthropic's official SDKs since the early days and helps hundreds of companies generate libraries and connectors across multiple programming languages. The acquisition aims to advance agent connectivity as AI shifts from answering questions to taking actions.
Acquihire strategy and talent acquisition
Anthropic is using acquisitions as a sophisticated recruitment method to hire top engineers at premium compensation levels. Building a successful company serves as the best interview process for identifying talent capable of end-to-end product ownership.Impact on existing customers
Stainless is shutting down its SDK generation services, leaving hundreds of companies that relied on their tools without support. Customers face migration challenges since different SDK generators produce incompatible outputs, creating vendor lock-in issues.Competitive dynamics between AI companies
The acquisition is viewed as part of broader competitive tensions, particularly between Anthropic and OpenAI. Since OpenAI uses Stainless for their SDKs, acquiring the company could disadvantage competitors while creating walled gardens.