# News Highlights

## Spectral Labs Releases SGS-1
- **SGS-1** is the first generative model capable of producing **fully manufacturable and parametric 3D geometry** from images or 3D meshes, enabling seamless integration into traditional CAD software.

## Vibe Coding Cleanup as a Service
- **Vibe Coding Cleanup** is emerging as a vital service in tech, addressing the reality that **most AI-generated code is production-unready**, leading to a surge in demand for specialists to rectify these issues before they escalate into significant technical debt.

## Show HN: I Parallelized RNN Training from O(T) to O(log T) Using CUDA
- **Parallelizing RNNs**: The project implements simplified GRUs and LSTMs, termed **minGRU** and **minLSTM**, which allow for parallel computation, transforming the training process from **O(T)** to **O(log T)**, thus enhancing GPU performance significantly.

## Philips Announces Digital Pathology Scanner with Native DICOM JPEG XL Output
- **Philips has launched the Pathology Scanner SGi**, the first to offer **native DICOM JPEG XL output**, which reduces file sizes by up to **50%** while maintaining high image quality, enhancing data management in pathology labs.

## Evals in 2025: Going Beyond Simple Benchmarks to Build Models People Can Use
- **In 2025, evaluations will focus on building models that are not just intelligent but also practical, emphasizing their utility in real-world applications.** This shift is driven by the need for models that effectively manage ambiguity, follow instructions, and adapt to dynamic environments, as highlighted by recent reports from [Anthropic](https://www.anthropic.com/research/anthropic-economic-index-september-2025-report) and [OpenAI](https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f142e/economic-research-chatgpt-usage-paper.pdf).

## The Beginner's Textbook for Homomorphic Encryption
- **Fully Homomorphic Encryption (FHE)** allows computations on encrypted data, producing results that match those from plaintext operations, thus ensuring data privacy during processing.

## Apple Silicon GPU Support in Mojo
- **Mojo now supports Apple Silicon GPUs**, enabling developers to utilize GPU capabilities on all modern Macs, which could democratize access to GPU-accelerated algorithms and AI model development.

## [R] MiniGrid DoorKeys Benchmark Active Inference
- The **Active Inference Framework** demonstrates impressive performance on the **MiniGrid DoorKeys (MG-DK)** benchmark, achieving an average of **<19 steps** for an 8x8 grid and **<60 steps** for a 16x16 grid without extensive training or benchmarking.

## The Deadweight Loss of Strictly Isotonic Regression
- **Strict isotonicity in calibration can lead to significant deadweight costs**, as it often merges distinct scores into fewer steps, diminishing the model's resolution and informativeness in downstream tasks.

## [P] Benchmarked EpilepsyBench #1 Winner - Found 27x Performance Gap, Now Training Bi-Mamba-2 Fix
- The **SeizureTransformer** achieved a remarkable **26.89 FA/24h**, revealing a **27x performance gap** compared to previous benchmarks on the Temple EEG dataset, showcasing significant advancements in EEG machine learning.

## [D] Strategies for Routing LLMs
- **LLM routing strategies optimize performance by directing queries to the most suitable model based on their capabilities**, utilizing a router that evaluates performance and efficiency profiles derived from clustered query embeddings.

## [P] Introducing LabelMob: Connecting ML Teams with Expert Data Annotators
- **LabelMob.com** connects ML teams with **expert data annotators** in specialized fields, addressing the challenge of finding high-quality, domain-specific annotation for complex datasets like quantum physics and biological sequences.

## [r] Governed Multi-Expert aka (GME)
- The **Governed Multi-Expert (GME)** architecture transforms a single large language model into a **dynamic team of specialists** using Low-Rank Adaptation (LoRA) modules, enhancing response quality and safety while optimizing computational resources.