Posts
Guides, essays, and deep dives.
machine-learningConstrained decoding: forcing an LLM to emit valid JSON
Watch a schema grammar mask illegal tokens, renormalize the survivors, and force a simulated LLM to emit valid JSON while a prompt-only run fails beside it.
machine-learningContinuous batching: why your API latency varies
A GPU scheduler timeline runs static and continuous batching on the same requests. Watch idle lanes refill, queues drain, and latency percentiles move live.
machine-learningHNSW vector search: how a layered graph finds neighbors fast
A layered graph walks coarse to fine toward your query, checking a fraction of the vectors a full scan would. Tune efSearch and M and watch recall respond.
machine-learningHow AI detectors work, and why they are wrong so often
Score text with the signals AI detectors use: predictability, burstiness, repetition, and style. Formulaic humans get flagged, edited model text slips by.
machine-learningLLM quantization explained: fewer bits, less memory
See how quantization shrinks a model's weight memory by storing FP16 values on a coarser grid of levels, and where the error actually lands.
machine-learningCan my computer run an LLM? VRAM requirements visualized
A live VRAM calculator: watch model weights, a growing KV cache, and runtime overhead fill a GPU memory bar, spill into system RAM, and compare GPUs.
machine-learningLoRA fine-tuning: adapting a model without rewriting it
See how LoRA trains a task update in two thin matrices while the base model stays frozen, and why rank sets a hard floor on what the adapter can learn.
machine-learningLost in the middle: where context windows go blind
Drag a needle fact through a long context and run seeded trials to watch recall dip in the middle, plus the strategies that flatten the U.
machine-learningPrompt injection: when data becomes instructions
Watch untrusted content compete with the real task inside an agent pipeline, then test which layered defenses actually hold.
machine-learningRAG vs fine-tuning: which one your project needs
Run the same question through RAG and a fine-tuned model side by side. Push the source date past the cutoff and watch freshness, citations, and style trade off.
machine-learningReranking: why top-K retrieval isn't good enough
A fast retriever pulls top-N candidates, then a cross-encoder style reranker reads query and document together and reorders them. Traps fall, answers rise.
machine-learningRLHF and DPO: how models learn what people prefer
Label response pairs and watch two preference-optimization pipelines, RLHF with a reward model and DPO's direct update, reshape a conceptual policy landscape.
machine-learningSelf-attention: how a token decides which tokens to read
See how self-attention works. Tap any word to make it the query and watch scaled dot-product attention weight every other word in the sentence.
machine-learningWhat does 7B mean? LLM parameter counts explained
The B in 7B counts the learned weights stored in a model. See how count, precision, and architecture determine file size in this interactive scale tunnel.
machine-learningWhy ChatGPT types one word at a time: streaming visualized
Watch a model write tokens one at a time while streaming rushes each one to your screen and buffered mode holds the whole answer behind a gate.
machine-learningHow attention works: search, feeds, and LLMs
See how attention works in search, social feeds, and LLM self-attention. One interactive runs the same score, normalize, allocate loop through all three.
machine-learningContext window: what an LLM sees, and what it forgets
Watch two LLM context windows, 2k and 32k tokens, slide through the same text and drop exactly the tokens that scroll off the back of the smaller one.
machine-learning36 ways AI works alongside a freelance web developer
A 3D tour of six freelance web project phases with six AI assists each, plus a toggle showing which 16 of 36 handoffs can drop the human approver.
machine-learningAgentic workflows: how AI agents chain business tasks
A radial simulation of an agentic AI workflow, plus a plain-language look at what agents do, why the orchestrator matters, and what breaks first for SMBs.
machine-learningLLM hallucination: why confident answers drift from truth
A 3D demo of language-model sampling: tokens drift outward as text moves from grounded to fabricated, while the model's confidence barely moves.
developer-toolsModel Context Protocol: how AI apps plug into tools and data
Watch JSON-RPC messages flow between an MCP host and its servers as the client discovers tools, invokes them, and streams results back to the model.
machine-learningPrompt engineering: how technique shapes what models say
Zero-shot, few-shot, chain-of-thought, and role prompting explained with an interactive demo. See how technique shifts token probability and output quality.
machine-learningRetrieval-augmented generation: how LLMs look things up
A 3D demo of a RAG pipeline: a query vector searches a corpus, the top-K nearest chunks light up, and the LLM grounds its answer in the retrieved text.
machine-learningAutoregressive generation: how an LLM writes one token at a time
A plain-English look at how LLMs actually produce text, with a 3D demo of the KV cache that makes inference feasible and what would happen without it.
machine-learningInference-time compute: more samples, smarter answers
A plain-English look at inference-time compute, with a live grid of reasoning chains that vote on the same question and surface the real accuracy-vs-N curve.
machine-learningLLM as judge: one model grading another
One LLM grades another: a live demo of pairwise, rubric, and best-of-N judging, plus the position, length, and self-preference biases you have to probe for.
machine-learningLLM temperature: one number that reshapes the next-token distribution
A plain-English look at temperature in language model inference, with a live demo that shows exactly what the parameter does to the probability distribution.
machine-learningMixture of Experts: sparse routing for huge models at fast-model cost
A plain-English explainer on MoE models with a live 3D demo showing how tokens route through a small subset of experts at each layer.
machine-learningTop-k and top-p sampling: how an LLM picks its next token
A plain-English look at the two cutoff strategies that decide which tokens a language model is allowed to sample from, with a live interactive demo.
machine-learningEmbeddings in 3D: how models turn words into coordinates
Type a sentence, watch each token land as a point in 3D space. Click any token to inspect its vector and see which other tokens it's nearest to.
machine-learningGradient descent: how a model rolls downhill
Watch a simulated model find a low spot on a 3D loss surface. A plain-English look at the optimizer that drives modern machine learning.