Learn AI by Building
From your first dataset to production agents — deep-dive series, hands-on notebooks, and experiments you can rerun yourself.
Paper of the Week
All Issues →Paper of the Week #8 — Agents Beat All 20 Science Benchmarks, and None of 135 Follow-up Papers Passed
This week's roundup. Agent4Sci reports that frontier agents beat human-verified baselines on all 20 scientific ML benchmarks it collected, mostly through engineering; Into the Unknown reports that none of 135 agent-written follow-up papers met every condition set by the scientists who posed the problems. Also: BizAgentBench, which grades goals and procedure separately on payment-company workflows, and CarryOn, which rebuilds an agent's context from what it has produced instead of its chat history.
Premium Series
Our Products
Courses and code built from what we measure here
Video Courses
13 hands-on courses — quantization, diffusion, RAG, on-device AI, decision models. $199 lifetime bundle, first 3 lectures of every course free
LLM Quantization and Compression Hands-On
GPTQ, AWQ, GGUF, QLoRA — fit LLMs into the memory you have. The course behind our KV-cache measurements
Free Benchmark Code
The scripts and raw results behind the measurement posts, starting with the KV cache harness. Free with an account
Premium Series
140 deep-dive posts across 21 series, bilingual KO/EN, with production-ready code and notebooks
Latest Posts
View All →
Free Grok Bot Forever? OpenMausBot + Local Qwen (No API Bills)
We tested four OpenMausBot jobs using local Qwen3.8-Flash-Next through Strata. A price-drop alert fired and the final Snake file passed our test, but none of six shell-based scheduled runs delivered a proper brief to chat. Setup, prompts and failures are included.
Can a Text Model Match Photos Without Seeing Them? Rerunning Yann LeCun’s JEPA Paper
We reran Balestriero and LeCun’s JEPA code on Flickr30k. A text encoder trained on different captions of one photo retrieved the correct caption 24.3% of the time, versus 12.5% for next-token prediction, after fitting a map to DINOv2. We explain the setup and its limits.
Who Does DeepSeek's Coding Agent Listen To? One GitHub Comment, Followed Through the Code
DeepSeek's coding agent dsh-superagent takes work from GitHub comments. By default, a request from someone not on its allowed list stops at a code gate and calls no model by itself. But a comment on an issue it's already working on reaches the agent regardless of the list. We read the code and did not run it.
Install llama.cpp on Linux and Get a First Answer on Your NVIDIA GPU
Build llama.cpp from source with CUDA, download a small model and get a first answer on your GPU. On our CMake 3.22 A100 server, building one architecture and one program cut the build from 31 minutes to about 8.
Can an 8 GB GPU Turn One Photo Into a 3D Model? Idol Peaked at 7,380 MiB on a Capped A100
We ran Speridlabs' Idol on an A100 capped at 8 GB, with no real 8 GB card. A real photo peaked at 7,380 MiB in nvidia-smi: 300 MiB left counting 8,192 MiB with a 512 MiB desktop, 263 MiB over counting 8 × 10⁹ bytes. Commands and the step that adds 762 MiB.
Which Local Coding Model Should You Run on Your GPU? 8, 12, 16 and 24 GB on 4 Python Tasks: Qwen3.5-9B, gpt-oss-20b, Flint, Qwen3.5-27B
On four Python tasks: Qwen3.5-9B for 8 and 12 GB, Flint for 16 GB (we built the agent with it), Flint or Qwen3.5-27B for 24 GB. Measured on an A100 against the capacities of the RTX 4060, 3060 12GB, 4070 Ti Super and 4090.