A computer-vision tool that evaluates deadlift form from side-on video: it counts reps, timestamps each pull, measures bar height using the near plate as a ruler, and classifies each rep's back as STRAIGHT or ROUNDED with live probability displayed beside the footage. Three models are combined: a ViTPose model for pose and hip hinge angle (used to cross-check rep counting), SAM 3.1 to track the plate and derive bar height (assumes a 45 cm Olympic plate), and a System One read (gemma-4-26b-a4b-it) that returns per-frame probabilities {"straight", "rounded"} about spinal shape. Each rep verdict is the mean P(rounded) across the first pull (from bar leaving the floor to roughly the knee), and the classifier is explicitly about spine shape, not torso lean.
Setup and usage are concrete: get a VLM Run API key, set it in a .env, create the provided conda environment, drop side-on clips into data/input, and run main.py with optional flags that override config.py (sample rate, plate diameter, rep source, model ID, threshold, export height, trim seconds, fresh runs, and on-screen credit). Tone-mapping/HDR is supported (requires ffmpeg with libplacebo and, on macOS, a Vulkan driver like moltenvk) and models can be fed tone-mapped frames. Outputs go to a timestamped data/output folder containing an overlay video, timeline.png, reps.json, reads.json, summary.txt and run.json (which records per-model cost); gateway replies are cached in data/cache. Licensed Apache-2.0.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
AstaBrief is an open-source model designed for fast scientific report generation, capable of turning research questions and literature into cited reports. It reduces generation time by nearly three and a half times compared to proprietary models, making it a useful tool for researchers. (allenai.org)
Shrivu Shankar argues that SaaS companies are evolving into 'harnesses,' where core tasks are performed by AI agents rather than humans. This shift transforms organizational structure, with humans overseeing and reviewing AI-driven processes within the company. (blog.sshh.io)
Three AI agents - Meta’s Muse, Anthropic’s Claude, and OpenAI’s GPT - were tested on multilingual research tasks involving U.S. and Iranian data. The experiment focused on how language and context influence agent performance across reasoning, sourcing, and artifact creation. (royapakzad.substack.com)
An open-source project named LDRAW-NOVA allows users to generate Lego models using AI. The repository contains tools, examples, and documentation for creating Lego designs with artificial intelligence. (github.com)
DwarfStar 4 (ds4) is a local inference engine that runs large language and vision models on high-memory Mac, CUDA, and ROCm machines. It supports models like DeepSeek V4 and Qwen3.8, offering CLI, APIs, and persistent cache features for efficient local deployment. (dwarfstar.sh)
Rai is a CPU-only large language model inference engine written entirely in Rust. It aims to provide efficient and pure Rust solutions for LLM deployment without relying on GPUs. (github.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.