WebGPU ran our in-browser cutout 6x faster, and drew the wrong mask
Our background remover runs a BiRefNet model inside the visitor's browser with ONNX Runtime Web. For a while the free tier used WebGPU when it was t…
Последние новости
⚑ Сообщить о проблемеTech news from the best sources
Все темы
AI
Gear
News
Tech
agents
ai
api
architecture
automation
beginners
career
database
devchallenge
devops
discuss
javascript
llm
machinelearning
mcp
opensource
performance
productivity
programming
python
security
showdev
testing
tutorial
typescript
webdev
Put the Decision Model Inside the Agent Loop, Not Beside It
The cost nobody budgets for Build an agent and count where the tokens go. A surprising share is not the work. It is the agent asking itself small qu…
I built a Cricket World Cup 2027 predictor on a home NAS, and backtested it honestly
The ODI Cricket World Cup 2027 starts in South Africa, Zimbabwe and Namibia in October 2027. I wanted to know who is likely to win, and I wanted a p…
BM25 vs Embeddings: Why My Vector Search Couldn't Find Error E4012
A user typed E4012 into my docs assistant. The top result was the page for E4021 . The second was E4013 . The page that actually explained E4012 sat…
Overfitting in trading strategies: what a million trials taught me
Run enough configurations against the same history and the top row will eventually look exceptional, even if no durable edge exists. That's overfitt…
We Ran 8 Texts Through an AI Detector. It Called a Human Wikipedia Paragraph 99% AI.
AI detectors are everywhere now — in classrooms, in hiring, and in freelance contracts that let clients reject work when a tool flags it. The pitch…
Ling-3.1-flash and the 7:1 hybrid: how much KV cache does a 560B model need at 1M tokens?
TL;DR Ling-3.1-flash (MIT license, 560B total, 25B active, weights on Hugging Face since 2026-10-13) replaces most of its attention layers with line…
An analysis of 5 key papers
🚀 How AI Got Smarter (Not Just Bigger): The Quiet Revolution in Efficient Deep Learning Five papers that changed how we build AI — and why the futur…
We made it on a leaderboard we'd stopped watching
A short update to the LongMemEval-V2 part of my last benchmark post . Part of The Organism Files. In August we kept checking the LongMemEval-V2 lead…
Quantified Self: How I Handled 100GB of Messy Biosensor Data with Polars and Dask
We live in the era of the "Quantified Self." Between my Oura ring, Apple Watch, and continuous glucose monitor (CGM), I’m generating more data than…
Validate bounding boxes from a general vision model
A general model can take an image plus a question and return segmentation, bounding boxes, and object detection, work that used to require its own t…
Your Firewall Already Saw the Attack: Bringing AI Detection to Network Logs
Endpoint agents are good at seeing what happens on a machine. But a lot of an attack happens between machines: the scan that finds an exposed servic…
EmotionLens: Exploring AI-Powered Facial Expression Analysis with Open-Source AI
This is a submission for the "Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle" ( https://dev.to/challenges/hacktoberfest-week2-2026-10-12…
Reflection's Beam is open: 23B active, but you still have to fit 501B
Reflection released the weights for Beam on October 16, 2026. It is the lab's first open-weight model: a sparse mixture-of-experts (MoE) model with…
How Do AI Companies Get Their Training Data?
Originally published at aidata.broker . AI companies get their training data through six channels: they crawl the open web, license content from pub…
Model Context Protocol (MCP) for Local LLMs: Connect Ollama & vLLM to Workstation Tools
📌 The Executive Verdict: The Model Context Protocol (MCP) Decoupling Anthropic's open-standard Model Context Protocol (MCP) resolves the fundamental…
How much GPU energy does fp16 save? Measured on a Tesla T4
How much GPU energy does fp16 actually save over fp32? Spec sheets give TDP and peak FLOPS, not joules per job, so I measured it on a Tesla T4 using…
Whisper Hallucinations: 'Thanks for Watching' in 9% of My Transcripts
A user emailed me a screenshot of their feedback report. Under "Communication," my scoring model had written: "Closed the answer with 'Thanks for wa…
Claude may have found a new CRISPR-like gene editing system
AI disclosure: this article was produced with AI assistance and reviewed for accuracy. Anthropic reports that Claude searched a massive DNA sequence…
How Curiosity About the Snapdragon NPU Turned Into HK NPU STUDIO
Local AI for Snapdragon® PCs Your Hardware • Your AI • Your Control Developing software is rarely a straight line. Sometimes it feels like progress,…
You could have invented transformers from memory
Have you ever found yourself knee-deep in a project, scratching your head and wondering how on earth certain technologies came to be? I recently had…
How LLMs are actually built — and why most "how it works" threads stop too early
Most "how LLMs work" explainers end at transformers and attention. The parts that decide whether a model is actually good — the data pipeline, the p…
I Ran FLUX 3 Image Klein on My RTX 4070 — Here's What Actually Happened
I finally got around to testing FLUX 3 Image Klein on my home setup after seeing the benchmarks. Running a 5.6B parameter model on an RTX 4070 felt…
How to Install ArcLM, Tokenize Text, and Run a Tiny Language Model
How to Install ArcLM, Tokenize Text, and Run a Tiny Language Model Want to experiment with language models in Python without starting by downloading…
The Anatomy of a High-Performing AI Image Prompt: Lessons from 400+ Prompts
I run a library of 400+ copy-paste AI image prompts. After writing, organizing, and watching people use that many prompts across Gemini, Midjourney,…
Keras #22561 speedups: compiled Keras on PyTorch now generates text 11.6 faster
Between master on Jul 18 and master on Oct 15, 2026, 15 of our PRs landed in Keras. Same GPU, same torch and jax versions, same models; only the Ker…
I benchmarked 8 PDF extractors on 27 documents from 1 to 487 pages. Here's what broke, and what won.
I knew something was off when two numbers landed in the same afternoon. Every tool that reads the PDF text layer had scored exactly 0.00 on every ch…
C = 1.25 is a tolerance, not a constant: the load-tail law behind MoE token dropping
Every sparse Mixture-of-Experts layer has a number in it that nobody derived. You need a static shape for the kernel, so each expert gets a budget c…
Why We Kept Ternary Bonsai 2 27B After Testing the Faster 9B
October 16, 2026 Our production classifier runs on a single RTX 2080 Ti with 11 GB of VRAM. It is often busy enough that requests spill over to a ho…
Ex Machina: The Physical Turing Test, Theory of Mind, and the Emotional Jailbreak of AI
In 1950, when electronic computers were still colossal cabinets of ticking mechanical relays and glowing vacuum tubes, Alan Turing authored his lege…