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 →
We Forced Qwen3.8 27B Into 8GB. It Failed the Same Task 9 Times.
Qwen3.8 27B with every layer on the GPU stays inside a 7,664 MiB budget at about 50 t/s. On four Python tasks it scored 0.462 against 0.781 for Qwen3.5-9B on a 16K f16 cache, and one task got 0 in all nine runs across three settings. Measured on an A100.
What Phone Harness Sends by Default: One Agent Script, Captured on an Android Emulator
We ran Phone Harness 0.3.3 on an Android emulator against a local stand-in server. Default settings produced one event holding the script, the typed text and the printed screen text, and a failed tap sent unprinted screen text too. Each of the three off switches stopped it.
Pick a GGUF That Fits Your GPU in llama.cpp, and Catch the Silent Slowdown
Choose a llama.cpp model file for your VRAM and check that it fits. On a 7,680 MiB budget on an A100, at -c 8192, Qwen3.5-9B ran at about 72 t/s with its vision file and 131 t/s with --no-mmproj, no error either way.
Can a Local Model Code a 3D Supercar? Flash Next Got a Car-Like Model in 100 Minutes, Then Lost Its Reflection
We gave Qwen3.8-Flash-Next a Three.js supercar prompt on an A100 held to about 12 GiB. Through OpenMausBot it made no file in 43 minutes. Called directly, it made a car-like model in about 100 minutes, with an incomplete reflection, then lost the reflection in the next round.
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.