#machine-learning

30 stories

⬢ HYVE SPOTLIGHT

The Owens AI Institute is giving K-12 AI education away free, forever

Hyve Spotlight·22m·HyveCares

Memory now dominates AI chip costs as GPU prices plateau

AI

Memory now dominates AI chip costs as GPU prices plateau

Hacker News·5h·intelkishan

AI

llmrequirements.com maps local LLMs to the hardware that can actually run them

Hacker News·6h·truetotosse

AI

KVBoost brings 5–48x speedups to HuggingFace via chunk-level KV cache reuse

Hacker News·2d·pythongiant

Developer argues AI training amounts to large-scale plagiarism without consent

AI

Developer argues AI training amounts to large-scale plagiarism without consent

hackernews·3d·speckx

Running LLMs offline on Apple Silicon can cost more than cloud inference

AI

Running LLMs offline on Apple Silicon can cost more than cloud inference

Hacker News·1w·datadrivenangel

Tool ranks local LLMs by actual benchmark performance on your hardware

AI

Tool ranks local LLMs by actual benchmark performance on your hardware

Hacker News·1w·andyyyy64

Running local LLMs on M4 MacBook: practical limits at 24GB

AI

Running local LLMs on M4 MacBook: practical limits at 24GB

hackernews·1w·shintoist

AI

Making the case for local AI as the default

hackernews·2w·cylo

Google upgrades reCAPTCHA with machine learning fraud detection

SaaS

Google upgrades reCAPTCHA with machine learning fraud detection

hackernews·2w·unforgivenpasta

Google releases multi-token prediction drafters for Gemma 4

AI

Google releases multi-token prediction drafters for Gemma 4

hackernews·2w·Google

AI

Bonsai 1.7B ternary model hits 442T/s on M4 Max

Hacker News·2w·hhuytho

Devtools

Predicting HN front page success with machine learning

Hacker News·3w·margotli

Apple's SHARP ML model now runs in-browser via ONNX

AI

Apple's SHARP ML model now runs in-browser via ONNX

Hacker News·3w·bring-shrubbery

AI

Study finds AI hiring tools systematically favor candidates similar to their training data

Hacker News·3w·laurex

Grok 4.3 arrives with longer context and cheaper inference

AI

Grok 4.3 arrives with longer context and cheaper inference

hackernews·3w·simianwords

TRiP: A Complete Transformer Engine Written from Scratch in C

Devtools

TRiP: A Complete Transformer Engine Written from Scratch in C

Hacker News·3w·carlovalenti

AI

Talkie: A 13B Language Model Trained on 1930s Text

hackernews·3w·jekude

Mercor breach exposes 4TB of voice data from 40k AI contractors

AI

Mercor breach exposes 4TB of voice data from 40k AI contractors

Hacker News·3w·Oravys

AI

Researcher outlines path toward unified theory of deep learning

hackernews·1mo·jamie-simon

DeepSeek v4 offers competitive inference at lower cost

AI

DeepSeek v4 offers competitive inference at lower cost

Hacker News·1mo·impact_sy

Alibaba's Qwen 3.6-27B brings flagship coding performance to smaller model size

AI

Alibaba's Qwen 3.6-27B brings flagship coding performance to smaller model size

Hacker News·1mo·Alibaba

Google releases eighth-gen TPUs designed for AI agents

AI

Google releases eighth-gen TPUs designed for AI agents

Hacker News·1mo·Google

AI

Meta plans to log employee keystrokes and mouse movements for AI training

Hacker News·1mo·Meta

AI

Darkbloom turns idle Macs into a private AI inference network

Hacker News·1mo·twapi

AI

Aphyr on why ML's future is fundamentally unpredictable

Hacker News·1mo·aphyr

Apfel brings local AI to macOS without subscriptions

AI

Apfel brings local AI to macOS without subscriptions

Hacker News·1mo·franze

AI

Google Open-Sources Gemma 4 Models

Hacker News·1mo·Google DeepMind

Ollama switches to MLX backend for faster Apple Silicon inference

AI

Ollama switches to MLX backend for faster Apple Silicon inference

Hacker News·1mo·Ollama

AI

GitHub will train AI on private repos unless you opt out by April 24

Hacker News·1mo·vmg12