Acing AI — AI education, tutorials, research and datasets for data scientists
Test-Time Compute: Where More Thinking Stops Paying
Test-time compute scaling explained: best-of-N, self-consistency, and verifier-guided search, where each saturates, and when more inference compute is wasted.
Latest Intelligence
Curated technical papers and hands-on implementation guides for the modern AI engineer.
Preference Optimization After DPO: IPO, KTO, and SimPO Compared
Runnable comparison of preference optimization after DPO: IPO, KTO, and SimPO. What each objective changes, when reference-free wins, and how to switch in TRL.
Self-Hosting a Frontier Open MoE: The Real GPU Bill for a 1.6-2.8T Open-Weight Model
Self-hosting a frontier open MoE like DeepSeek V4 or Kimi K3 means holding all 1.6-2.8T parameters in VRAM, not the active count. The real GPU bill and when the API wins.
ArticlePagedAttention Is Gone — What vLLM's Model Runner V2 Replaced It With
TutorialRLVR in Practice: Build a Verifiable-Reward Loop with GRPO
ArticleLinear Attention at Frontier Scale: Kimi K3's KDA Claim, Fact-Checked
Browse by Type
Tutorials
Step-by-step guides from neural network basics to advanced LLM fine-tuning.
Research Papers
Peer-reviewed insights and white papers defining the frontier of artificial intelligence.
Datasets
High-fidelity training sets for natural language processing and computer vision.
Start Learning
Guided sequences through our best content — structured to build understanding from the ground up.
Building Production AI Agents
Go from an agent demo to a system that holds up in production. This path moves through the four decisions that determine whether an agent ships: the deployment reality gap, tool and data access with MCP, persistent memory, and evaluation that measures reliability rather than luck.
LLM Inference in Production
A practical route through the serving stack. Start with the map of where inference cost and latency actually come from, take the quantization lever apart, speed up decoding with speculative drafts and measure it yourself, confront what long context really delivers, and finish at the 2026 state of the art. Every step names its tradeoffs.