paste a ds or dnskey record into the field above to use. the given name servers must be authoritative for the same zone as the trust anchor - i.e., the name server must match the zone in which the trust anchor is located if it is not already published in the dns. qr = 0 tc = 1 rd = 0. opcode = noerror ;; do =
We introduced Multi-Token Prediction (MTP) drafters for Gemma 4, delivering up to 3x speedup without quality degradation. MTP decouples token generation from verification, utilizing idle compute to predict future tokens and verify them in parallel.
The author proposes the Inverse Laws of Robotics to prevent humans from blindly trusting AI systems, which can lead to societal dangers, and emphasizes the need for humans to remain responsible and accountable for AI use. The three laws are: humans must remain responsible for AI consequences, not anthropomorphise AI, and verify AI output before accepting it as authoritative.
We compared two AI agent approaches for operating an admin panel: vision agents and API agents. Vision agents are more expensive and less reliable due to the need for explicit UI walkthroughs and high input token counts.
Google Chrome silently installs a 4 GB AI model on users' devices without consent, violating Article 5(3) of the ePrivacy Directive and GDPR principles of lawfulness, fairness, and transparency. The model's environmental cost is significant, with estimated CO2-equivalent emissions ranging from 6,000 to 60,000 tonnes per model push, depending on the number of devices affected.
Browse, search, and filter beautiful color palettes inspired by the Old Masters.
GLM-5V-Turbo is a multimodal agent model that integrates perception and action capabilities. It achieves strong performance in multimodal tasks while preserving text-only coding capability.
Claude releases 10 ready-to-run agent templates for financial services tasks, expanding its partner ecosystem and capabilities across Microsoft Office.
A required part of this site couldn’t load. This may be due to a browser extension, network issues, or browser settings. Please check your connection, disable any ad blockers, or try using a different browser.
Raymond has been involved in Windows evolution for over 30 years and created The Old New Thing website and book. He occasionally shares stories on the Windows Dev Docs Twitter account.
Philip, an engineer at Distr, shares common Docker Compose quirks and operational gaps that can lead to production incidents. He recommends handling these gaps with commands and practices, or using an agent like Distr's to automate the process.
Ethan Mollick's work on AI adoption highlights the "messy middle" phase where AI use is uneven, hidden, and difficult to compare, and not yet connected to organizational learning. Companies must move beyond license usage and surveys to understand how AI creates learning and value within their organization.
0
2026-05-05 17:32:32 UTC
ycombinator.com
Llama
Proliferate is building an operating system for modern engineering, automating software development to make engineering more human. They're hiring an engineering lead to build the core product with high ownership and a top-market salary.
Apple is adding a "Create a Pass" feature to the Wallet app in iOS 27, allowing users to build their own passes from scratch or scan QR codes. This feature is expected to be released in September 2026, making it easier for users to add passes without needing a third-party tool.
The author is working to reduce async Rust bloat in the compiler, specifically targeting the state machine generated for async code. They propose two optimizations: returning Pending instead of panicking when a future is done, and removing the Panicked state.
Researchers at Kobe University developed a Mie-resonant silicon nanoparticle ink that can be printed using an inkjet printer to create structural colours. This breakthrough enables large-area applications of structural colour, with potential uses in anti-counterfeiting, smart displays, and art pieces.
Michel, CEO of Airbyte, introduces Airbyte Agents, a unified data layer for agents to discover information and take action across operational systems, addressing issues with current MCPs. Airbyte Agents uses a Context Store to provide a structured way for agents to discover data, reducing API calls and improving performance.
Please enable JS and disable any ad blocker
Meta's CEO Mark Zuckerberg and the company are being sued by five publishers and an author for allegedly copying millions of copyrighted works to train their AI system. The lawsuit claims Meta deliberately circumvented copyright protections and considered licensing the works before abandoning that strategy.
At Andon Labs, we deploy AI agents into the real world, giving them real tools and real money, to show the current capabilities of AI. After the success of Andon Market in San Francisco, we wanted to see what happens when you move the experiment 9,000 kilometers east and add European bureaucracy. We signed a lease for a space at Norrbackagatan 48 and gave it to an AI named Mona. This post ...
You don't have permission to access "http://www.nsa.gov/Cybersecurity/Quantum-Key-Distribution-QKD-and-Quantum-Cryptography-QC/" on this server.
Something went wrong, but don’t fret — let’s give it another shot. Some privacy related extensions may cause issues on x.com. Please disable them and try again.
The containerd image store is the default storage backend for Docker Engine 29.0 and later, offering features like lazy-pulling and peer-to-peer image distribution. It uses more disk space than legacy storage drivers due to dual storage of compressed and uncompressed formats.
Apple has removed higher RAM options from Mac mini and Mac Studio due to global memory shortage caused by AI server demand. Delivery estimates for these models are now 9-10 weeks or even 4-5 months for upgraded RAM configurations.
The author of CodeMirror, a code editor system, discusses the design considerations for its collaborative editing feature, deciding to use a non-distributed operational transformation (OT) approach due to its simplicity and effectiveness. The author also explores alternative approaches, including CRDTs, but ultimately decides against them due to their complexity and memory requirements, ...
The user compared the 8080, Z80, 8086, and 6502 CPUs, highlighting their similarities and differences, and discussed how their instruction sets and design decisions influenced programming styles and best practices. The user also shared their own experiences and advice for programming each chip, including guidelines for memory access, loop variables, and register usage, and noted that the Z80 ...
The meta-harness is a loop that improves the harness around an LLM agent by reading diagnostic logs and writing a better harness, which can be achieved with Islo's sandbox infrastructure. A proof-of-concept implementation demonstrates the optimization signal end-to-end with a proposer that reads all traces, not summary scores, and converges in four steps on a 5-task held-out suite.
The user shares 10 lessons for agentic coding, emphasizing the importance of implementing to learn, rebuilding often, and investing in end-to-end tests. These lessons aim to help developers make the most of cheap code and agent capabilities while considering maintenance, support, and security.