Articles
In-depth articles on AI, web development, cloud native, and indie hacking.
Reasoning models such as o1, o3, and DeepSeek-R1 "think before answering." They excel at complex math, code, and multistep reasoning but are slower, costlier, and harder to control. This article explains where they help or hurt and presents a hybrid architecture that reserves them for genuinely hard reasoning.
Commit messages and PR descriptions are documentation developers hate writing but value most. AI can auto-generate high-quality versions from a diff. This article covers generating spec-compliant commits (Conventional Commits) and PR descriptions (motivation/changes/tests/risks), and integrating them into git hooks/CI so 'too lazy to write' is no longer an
Changing embedding models, revising chunking, and rebuilding indexes can cause downtime or degrade retrieval in production RAG. This article covers blue-green indexes, dual writes, shadow traffic, rebuild-and-switch workflows, and incremental updates that make these changes invisible to users.
Agents are non-deterministic, multi-step, and call tools — manual testing can't keep up. This article explains how to automate agent evaluation: build task sets + rubrics, batch-score with LLM-as-judge, do trace-driven regression, wire eval into CI, and mine bad cases from production traces for auto-intake — so every agent change gets an objective score.
AI products are harder to retain than traditional ones—novelty fades, free users free-ride without paying, and big platforms poach users. This article covers what's special about AI retention, key metrics (first value, habits, paid renewal), retention levers (continuous value, workflow embedding, data flywheel), and spotting and cutting churn.
MCP lets AI tools plug into your capabilities, but also exposes data and operations to every compatible client. This guide covers least-privilege tools, validation and injection defenses, remote authentication and rate limiting, local-vs.-remote trust boundaries, third-party supply-chain risk, and a prelaunch checklist.
LLM applications need SLIs and SLOs, but traditional Web targets do not transfer: latency varies widely, quality is hard to quantify, and cost is itself a metric. This article covers latency (p95/TTFT), availability (model API failures), quality (evaluation scores/bad-case rate), cost (tokens/request), alerting, and error budgets.
Take tons of notes but never read them? AI turns a "knowledge base" from a "warehouse" into a "queryable assistant." This article covers personal (private RAG over notes/docs) and team (shared base + AI retrieval + continuous capture) knowledge management—and the key point: AI doesn't replace organizing, it amplifies its value.
Streaming output defines how responsive an AI app feels, but many developers do not understand SSE, vendor data-stream formats, streaming tool calls, or cancellation. This article moves from HTTP SSE through the AI SDK data stream protocol to streaming tool calling and interruptible streams, explaining frontend parsing, backend emission, and common pitfalls.
Refactoring is the cure for technical debt but gets postponed because it's draining. AI makes untouchable legacy code touchable—but used wrong it introduces subtle bugs. This article shows the right way to do AI-assisted refactoring: pick scenarios, small steps with test guardrails, behavioral equivalence, and review priorities.
An Agent is a loop at its core, but whether that loop is well-designed decides if it is intelligent or just thrashing. This article covers the key design points of the Agent loop: stop conditions (when to end), self-reflection (self-assess after each step), branching and backtracking, max steps and budgets, and adding "human checkpoints" to intercept
AI startups shouldn't pile on infrastructure early, but some foundations must exist from day one—otherwise everything collapses once traffic arrives. A pragmatic checklist: the model call layer (routing/caching/fallback), observability, evaluation, data (vector DB/memory), and security (rate limiting/moderation), with what to build first vs. defer.