# May 14, 2026

## Show HN: Needle: We Distilled Gemini Tool Calling into a 26M Model
- **Needle** distills **Gemini 3.1** into a **26m parameter Simple Attention Network**, enabling local finetuning on personal devices, achieving **6000 toks/sec prefill** and **1200 decode speed** in production.

## A History of IDEs at Google
- **Google's IDE landscape evolved from fragmentation to a unified platform** with the introduction of Cider V, which now supports **80% of development** in the main codebase, enhancing productivity through better integrations and AI features.

## The US is winning the AI race where it matters most: commercialization
- The **US leads in AI commercialization**, leveraging cloud infrastructure, data, and strategic capital, significantly outpacing competitors like China in revenue and adoption since the launch of DeepSeek R1 in January 2025.

## An idiot's guide to lead optimisation for proteins
- **Lead optimisation** is a critical phase in drug design, where existing molecules are refined to enhance their efficacy, leveraging **machine learning** to propose and test modifications efficiently, as exemplified by the **Cradle-1 pipeline**.

## MinT: Managed Infrastructure for Training and Serving Millions of LLMs
- **MinT** is a **managed infrastructure** system that optimizes **Low-Rank Adaptation (LoRA)** for training and serving millions of large language models (LLMs) by efficiently handling adapter revisions without the need for full model checkpoints.

## Human-level performance via ML was *not* proven impossible with complexity theory
- The **Ingenia Theorem** proposed by Van Rooij et al. claims that achieving AGI via ML is impossible, but this assertion is fundamentally flawed due to the lack of a precise definition for "human-level classifier."

## Hermes Unlocks Self-Improving AI Agents, Powered by NVIDIA RTX PCs and DGX Spark
- **Hermes introduces self-improving AI agents** that leverage NVIDIA RTX PCs and DGX Spark, enabling continuous learning and adaptation through user feedback and task complexity.

## Marco Polo: Finding a friend with only distance and motion
- **The article explores the challenge of locating devices in crowded spaces using range-only relative localization, leveraging ultra-wideband (UWB) technology and inertial measurement units (IMUs) to track movement and distance.**

## Elastic Attention Cores for Scalable Vision Transformers
- **Elastic Attention Cores** introduce a **core-periphery block-sparse attention** structure for Vision Transformers, reducing computational costs from **N²** to **2NC + C²** for **C** core tokens, enhancing scalability at higher resolutions.

## LLMs are breaking 20 year old system design
- **LLMs challenge the 20-year-old web architecture** by introducing long-running, stateful processes that require a new routing primitive, as traditional models assume stateless interactions with databases.

## EditLens: Quantifying the extent of AI editing in text (2025)
- **AI editing is prevalent**, with a significant number of queries to language models focused on editing rather than generating new content, revealing a need for tools like EditLens to quantify this phenomenon.

## Trained transformer-based chess models to play like humans (including thinking time)
- **Trained transformer-based chess models** achieve superior accuracy compared to MAIA-2 and are competitive with MAIA-3, utilizing nearly **1B games** from Lichess across various rating buckets from ~800 to 2500+ with only **9MM parameters**.

## Learning, Fast and Slow: Towards LLMs That Adapt Continually
- **Fast-Slow Training (FST)** enables large language models (LLMs) to adapt continually by utilizing **fast weights** for task-specific learning while maintaining **slow weights** for general reasoning, enhancing efficiency and performance.

## Continual Harness: Online Adaptation for Self-Improving Foundation Agents
- **Continual Harness** automates the iterative refinement process for self-improving foundation agents, enhancing their performance through model-harness co-learning, as demonstrated by the Gemini Plays Pokémon project.

## Scenema Audio: Zero-shot expressive voice cloning and speech generation
- **Scenema Audio** enables **zero-shot expressive voice cloning**, allowing any voice to convey various emotions without prior recordings, enhancing creative workflows in video production.