Articles for people who
ship the stack
Rewrites on React, Node.js, TypeScript and AI. A translated body is shown when it exists. If it does not, you see the English text and a note that says so.
Tagged: machine-learning
What Really Runs a Neural Network: AI Compilers Beneath Python
What really runs a neural network: follow an LLM from a Python expression through computation graphs, operator fusion, MLIR-style IRs and lowering to hardware.
6 min readSwitching RAG Embedding Models Without Reindexing: How rebasis Works
Switch RAG embedding models without reindexing: rebasis maps new query vectors into the old index with a small adapter and tells you when bridging won't pay.
18 min readDS-STAR vs General Coding Agents: Google's Data Analysis Harness Explained
Why a general coding agent stalls on data analysis and how Google's DS-STAR harness fixes it: seven agents, DS-STAR+ reports, feature engineering, Claude Code.
22 min readSoft-Label Distillation for a 0.8B SQL Agent: Why a Weaker Qwen Teacher Won
LoRA fine-tuning on top-k soft labels from frozen traces took a 0.8B SQL agent to 55 of 220 tasks, and a weaker Qwen teacher beat the stronger GPT model.
20 min readPlaying Volleyball With AI Agents to Predict Restaurant Turnover
A restaurant turnover hackathon built on human-gated AI agents: data audit, tail-aware validation, feature engineering, ensembling and the gap to leaderboard scores.
18 min readEight Linear Algebra Ideas Behind Embeddings, LoRA and PCA
Vectors, matrices, eigenvectors, cosine similarity, norms, SVD and rank explained with small house-price examples and the AI techniques that depend on them.
7 min readWhat GPT Really Stands For and Why AI Engineering Is Still Engineering
Unpack Generative Pre-trained Transformer, attention, fit and predict, and see why good AI work still depends on ordinary engineering judgment and better questions.
4 min readFive Free Ways to Start Learning AI Without Drowning in Roadmaps
A short list of free AI learning resources, from Google's ML crash course to fast.ai, Hugging Face, Kaggle and YouTube, with advice on what to pick first.
7 min readHow an LLM Works: Tokens, Attention and Next-Token Prediction
How an LLM works in plain terms: tokens, embeddings, self attention, causal masking and next-token generation, and why training and inference really differ.
8 min read
About these articles
Request a 24h estimate
Need the same stack in a production operator layer? Send the brief — estimate within 24 hours.