Austria Seeks EU Backing to Host Anthropic Amid US Access Restrictions

·1 views

Austria is lobbying the European Union to support hosting AI company Anthropic within its borders. The push comes in response to US access curbs that have complicated European use of American AI services. Austria appears to be positioning itself as a European hub for the Claude-maker's operations. The move reflects broader EU efforts to secure independent access to advanced AI infrastructure amid transatlantic tensions.

Read the full story at Hacker News

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Why LLMs Fail in the Real World: The Overfitting Problem in RAG Evaluation

Overfitting is a common machine learning issue where a model performs well on training data but poorly on new, unseen inputs — a problem that also affects large language models (LLMs). In Retrieval-Augmented Generation (RAG) evaluation, overfitting can cause models to memorize training examples rather than learning generalizable patterns. AI platform Narrivo highlights that overfit models are prone to failing on out-of-distribution data and can be overly sensitive to minor input variations. To counter this, experts recommend strategies such as regularization techniques like dropout, data augmentation, early stopping, and evaluating models on diverse test sets. Addressing overfitting is considered critical to building LLMs that perform reliably in real-world deployment scenarios.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Dev builds bare-metal PCI, NVMe, and FAT32 drivers for custom OS from scratch

A developer working on V.E.L.O.C.I.T.Y.-OS, a bare-metal operating system designed to run inside the CPU's L3 cache, has published Part 9 of a 12-part build series. After transitioning the kernel to Ring 0 in the previous installment, the developer faced the challenge of having no drivers to read from storage or load files. To solve this, they wrote a PCI configuration space scanner in Rust that queries buses, slots, and functions via legacy I/O ports to detect attached hardware. Using the scanner's output, an NVMe block storage driver was built by locating the mass storage controller, mapping its MMIO registers, and implementing the full NVMe startup and read sequence. A FAT32 parser was also developed, giving the OS the ability to locate and read files directly from disk without relying on any existing driver stack.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Using OpenAI's tiktoken for Claude Token Counts Can Skew Cost Estimates by 20%

Developers using OpenAI's tiktoken tokenizer to estimate token counts for Anthropic's Claude models risk cost and context budget errors of 15–20%, and more on code or non-English text. This happens because Claude uses its own tokenizer, which splits text differently than tiktoken, causing systematic undercounting. Anthropic provides a dedicated countTokens API endpoint in its SDK that returns accurate, model-specific token counts before inference is run. Token counts also vary across Claude model versions, meaning cached counts from older models should not be reused when switching versions. The recommended fix is to always call countTokens against the specific Claude model being used, and never apply a blanket multiplier to convert counts between models.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Blockchain Failures Often Stem From Cross-Layer Protocol Breakdowns

Blockchain outages are rarely caused by a single component failure; instead, they typically emerge from interactions between layers such as transaction decoding, execution, and consensus. A single malformed transaction can freeze a chain entirely if validators keep rejecting the same invalid block without purging it from the mempool, causing leadership rotation to stall progress without ever producing a fork. Consensus reaching unanimous agreement on rejection does not mean the protocol is functioning correctly, as it can indefinitely halt progress while preserving safety. Proper error classification is critical — distinguishing permanently invalid transactions from local infrastructure faults prevents operational errors from being broadcast as consensus decisions. Robust testing must simulate multi-node fault scenarios including process kills, disk failures, and mixed software versions to reveal whether local failures can cascade into a global network outage.

0 comments Read more at DEV Community