All 88 articles, sorted alphabetically
Advanced RAG
Beyond basic top-K retrieval. Techniques for hard questions.
Read article →Agentic Prompting
Structure prompts for autonomous multi-step task completion.
Read article →Chain-of-Thought (CoT) Prompting
'Let's think step by step.' Unlocks reasoning in large models.
Read article →Chain-of-Verification (CoVe)
Model generates → generates verification questions → answers them → revises. Reduces hallucination.
Read article →Constitutional AI
Model critiques + revises its output per explicit principles. Bootstraps harmlessness.
Read article →Constrained Decoding
Force model to produce output matching a grammar. Zero-cost format guarantees.
Read article →Context Window Management
Chunking, summarization, retrieval. When your data doesn't fit.
Read article →Cost Optimization
Route by difficulty. Cache. Compress. 10-50x cost savings possible.
Read article →Emergent Abilities
Some capabilities appear only above threshold model size. Contested but influential.
Read article →Dynamic Few-Shot
For each query, retrieve most similar examples from bank. Big accuracy jump on hard tasks.
Read article →Few-Shot Prompting
Show the model the format via examples. Foundation of in-context learning.
Read article →Function Calling
Define tool schemas, model produces JSON call. Foundation of modern agents.
Read article →Grounding + Citations
Hallucination defense: model must quote provided sources with tags.
Read article →Instruction Hierarchy
OpenAI's formalization. Model trusts sources by role. Foundation of injection defense.
Read article →Long Context Prompting
Gemini/Claude/GPT-5 support 200K-2M tokens. New patterns emerge.
Read article →Meta-Prompting
Use LLM to generate/refine prompts for other tasks. Iterative prompt improvement.
Read article →Model-Specific Tricks
Each model family has quirks. Prompts that win on one may lose on another.
Read article →Multi-Agent Orchestration
Team of specialized agents. Orchestrator routes. Cheaper than one giant agent.
Read article →Negative Prompting
Explicit prohibitions. Sometimes works, sometimes highlights the forbidden.
Read article →Agent Prompt Architecture in Depth
A 2500-word walkthrough of agent prompt architecture: system prompt, tools, memory, response format, safety, few-shot, CoT, budget, versioning, portab…
Read article →Chain-of-Thought Prompting
How chain-of-thought (CoT) prompting elicits reasoning, why it improves accuracy on reasoning tasks, and its limitations.
Read article →Chain of verification architecture
Deep-dive on Chain of Verification (CoVe): draft, per-claim verification in isolated context, compare, revise, cost.
Read article →Context packing architecture
Deep-dive on context packing: salience scoring, summarization, retrieval augment, working memory, priority ordering, budget.
Read article →Dynamic few-shot prompting
Deep-dive on dynamic few-shot prompting, where the demonstrations shown to a model are selected per request from an exemplar store rather than frozen …
Read article →Prompt Evaluation Architecture in Depth
A 2500-word walkthrough of prompt evaluation: golden sets, runner, metrics, regression gates, CI, cost/latency, adversarial, rollout gates, curation.
Read article →Prompt evaluation architecture
Deep-dive on prompt evaluation: test cases, rubrics, judge models, regression runner, golden set, human review, live shadow.
Read article →Prompt Evaluation Metrics
How to measure prompt quality: automated metrics, LLM-as-judge, human evaluation, and A/B testing.
Read article →Few-Shot Prompting
How few-shot prompts teach models via examples, how to choose examples, and when few-shot outperforms zero-shot.
Read article →Few-shot prompting architecture
Deep-dive on few-shot prompting: exemplar pool, selection, ordering, retrieval, K, formatting, evaluation, storage, cost.
Read article →LLM hallucination guardrails architecture
Deep-dive on layered hallucination guardrails for LLM systems: why fabrication is structural and trust asymmetric, grounding via retrieval, the cite-o…
Read article →Least-to-most prompting architecture
Deep-dive on least-to-most prompting: decomposing a hard problem into an ordered easiest-to-hardest subproblem queue, passing each committed subanswer…
Read article →Meta-prompting -- using an LLM to write and improve prompts
Deep-dive on meta-prompting: using an LLM to generate and refine prompts, the optimization loop (generate/evaluate/refine), the meta-prompt, automatic…
Read article →Multimodal prompting architecture
Deep-dive on prompting with images: how pixels become tokens via resize and tiling, low vs high detail cost, image placement and interleaving, the cro…
Read article →Output Formatting
How to specify output formats (JSON, XML, markdown) and enforce structure for machine-readable output.
Read article →Output parsing architecture
Deep-dive on LLM output parsing: constrained decode (JSON mode, grammars), parser, repair, retry with feedback, validation, fallback.
Read article →Prompt caching architecture
Deep-dive on LLM prompt caching: storing the deterministic KV attention state of a stable prompt prefix so later requests prefill only the new tail, t…
Read article →Prompt compression architecture — segment budgets, compression cascades, and cache-aware assembly
Deep-dive on compressing LLM prompts at scale: segment inventory and per-class token budgets, relevance scoring, the extractive-abstractive-pruning ca…
Read article →Prompt-injection defense architecture
Deep-dive on defending LLM applications against prompt injection: why no prompt can stop it, how to separate trusted instructions from untrusted conte…
Read article →Prompt Pipeline Architecture in Depth
A 2500-word walkthrough of a production prompt pipeline: templates, variables, compiler, structured outputs, model router, validation, registry, eval,…
Read article →Prompt registry architecture
Deep-dive on prompt registries: immutable prompt versions with variable schemas, eval-gated promotion through environment labels, cached runtime resol…
Read article →Prompt Templates
How prompt templates with placeholders enable maintainable, reusable prompts across features and users.
Read article →ReAct
How ReAct interleaves reasoning and tool calls, enabling agents that think and act iteratively.
Read article →ReAct prompting
Deep-dive on ReAct prompting: the thought-action-observation loop, tool grounding for reduced hallucination, prompt format vs native tool calling, err…
Read article →Reflexion architecture
Deep-dive on the Reflexion pattern: how an LLM agent improves within a single session through verbal reinforcement — writing natural-language self-cri…
Read article →Role Prompting
How assigning a role or persona to the LLM shapes its output style, expertise, and tone.
Read article →Prompt routing architecture
Deep-dive on prompt routing: intent classifier, policy engine, LLM registry, selection, fallback ladder, quality gate, and A/B.
Read article →Self-Consistency
How self-consistency samples multiple CoT traces and takes the most common answer, boosting reliability.
Read article →Self-consistency -- sample many reasoning paths, vote
Deep-dive on self-consistency: sampling multiple chain-of-thought reasoning paths and taking the majority-vote answer, marginalizing over the reasonin…
Read article →Semantic routing architecture
Deep-dive on semantic routing for LLM apps: embedding queries, labeled route centroids, cosine similarity and confidence thresholds, LLM fallback for …
Read article →Skeleton-of-Thought prompting architecture
Deep-dive on Skeleton-of-Thought: a cheap skeleton call that lists an answer's key points, parallel expander calls that flesh…
Read article →Step-back prompting architecture
Deep-dive on step-back prompting: the two-stage abstraction-then-reasoning pipeline, using the derived principle as a retrieval key, self-verification…
Read article →Structured Output Architecture for LLMs in Depth
A 2500-word walkthrough of structured output: schema, provider features (function calling, JSON mode), constrained decoding, validation, retry, stream…
Read article →Tree of Thoughts
How Tree of Thoughts (ToT) explores multiple reasoning paths in parallel with backtracking, and when it beats linear CoT.
Read article →Tree of Thoughts architecture
Deep-dive on Tree of Thoughts prompting: thought decomposition, candidate generation, value vs vote state evaluation, BFS/DFS search with beam width a…
Read article →Verifier architecture
Deep-dive on verifiers: post-hoc check, calibration, threshold, reject-and-retry with feedback, fallback, domain-specific tuning.
Read article →Zero-Shot Prompting
How zero-shot prompting works, when it's sufficient, and the design patterns that improve zero-shot quality.
Read article →Prompt A/B Testing in Production
Route traffic split. Measure metric. Statistical significance before shipping.
Read article →Analogical Prompting
'Think of a similar problem you've solved.' Model retrieves similar patterns.
Read article →Anatomy of a Prompt
Every modern LLM uses 3 roles. What each role does, when to use what.
Read article →Prompt Caching
APIs cache prefix tokens. 90% cost reduction on repeated system prompts.
Read article →Prompt Chaining
Multiple LLM calls, each doing one thing. Cheaper + more reliable than one mega-prompt.
Read article →Prompt Compression
Compress prompts 2-20x preserving downstream task quality.
Read article →Debate Prompting
Two agents argue different positions. Judge picks winner. Improves reasoning.
Read article →Debugging Broken Prompts
When prompt fails: minimal example, ablation, model comparison. Not vibes.
Read article →Prompt Delimiters
Separate instructions from data. Reduces injection + confusion.
Read article →Prompting for Code Generation
Language, style, tests, review. Foundation of Copilot, Cursor.
Read article →Domain-Specific Prompts
Schema, hints, few-shot examples, self-correction loop. Text-to-SQL production patterns.
Read article →Prompt Engineering Team Structure
Who owns prompts? How to review + ship. Emerging discipline.
Read article →Prompt Evaluation
How to measure prompt changes objectively. Reject vibes-based improvement.
Read article →Extraction Prompts
Given text, extract entities, relationships, key-value pairs. JSON schema-guided.
Read article →Multilingual Prompting
Cross-lingual tasks. English pivot. Translation fidelity.
Read article →DSPy
Declare task signatures. Framework compiles prompts + optimizes. No hand-tuning.
Read article →Prompt Design for Streaming UX
How prompt affects streamed output quality. First-token latency. Format for progressive rendering.
Read article →Temperature, Top-p, Top-k
Randomness parameters. When to crank, when to zero out.
Read article →Prompt Template Libraries
Reusable prompt templates. Variables. Composition. Trade-offs.
Read article →Prompt Versioning
Version-control prompts. Track eval metrics per version. Prevent silent drift.
Read article →ReAct
Model alternates Thought → Action → Observation. Foundation of agentic prompting.
Read article →Reflexion
Agent learns across attempts via natural language memory of past mistakes.
Read article →RAG
Retrieve relevant docs from vector DB, feed to LLM. Foundation of KB Q&A.
Read article →Role Prompting
'You are a senior security engineer…' Real gain or superstition?
Read article →Self-Consistency
Sample N CoT solutions, take majority answer. Boosts accuracy 5-15%.
Read article →Self-Refine
Model generates → model critiques → model rewrites. 1-3 iterations.
Read article →Structured Output
Force LLM output into machine-parseable format. Table stakes for production.
Read article →Summarization Prompts
Extractive vs abstractive. Length. Focus. Audience. Get precisely what you asked for.
Read article →Task Decomposition
Split hard problem into subproblems, solve smallest first, compose.
Read article →Tree of Thoughts
Explore multiple reasoning branches, prune, backtrack. RL-flavored prompting.
Read article →Zero-Shot Prompting
Ask the model directly, no examples. When it works, when it fails.
Read article →