Structured Index
Browse every article by topic
This page is the full HTML map of the blog library. Use it when you want a topic-by-topic view of every section and article, or when you want a stable index page to link into from other guides.
Handbook
Setup, mechanics, performance, development, and agent workflows.
Agents
- How to Create Your First Claude Code SubagentA step-by-step tutorial for developers who already use a CLAUDE.md but have never built a subagent. Create one Markdown file in .claude/agents/ and Claude starts delegating to it.
- Best Claude Code Subagents in 2026 (and How to Build Your Own)The most useful Claude Code subagents to set up in 2026: planner, reviewer, tester, security, and debugger roles, plus a step-by-step guide to writing your own custom subagent.
- How Garry Tan Uses Claude CodeThe 23 slash commands, six roles plus a CSO, and one install line behind Garry Tan's Claude Code setup. Plus the criticism.
- Agent Harness EngineeringThe harness is every layer around your AI agent except the model itself. Learn the five control levers, the constraint paradox, and why harness design determines agent performance more than the model does.
- Hermes Agent: Self-Improving AIHermes Agent writes its own memory as plain markdown files. After 5+ tool calls on any task, it creates a SKILL.md. Future sessions load it automatically. Here's how it works.
- Agent FundamentalsFive ways to build specialist agents in Claude Code: Task sub-agents, .claude/agents YAML, custom slash commands, CLAUDE.md personas, and perspective prompts.
- Agent Teams Best PracticesBattle-tested patterns for Claude Code Agent Teams. Context-rich spawn prompts, right-sized tasks, file ownership, delegate mode, and v2.1.33-v2.1.45 fixes.
- Custom AgentsDefine your own Claude Code specialists with .claude/commands slash commands, .claude/agents YAML, and CLAUDE.md personas. Real examples and mistakes to avoid.
- Builder-Validator Agent TeamsPair a builder agent with a read-only validator in Claude Code. Task 2 blocks on Task 1 via addBlockedBy, so every sub-agent output gets a second set of eyes.
- Sub-Agent Design in Claude CodeSplit a hard code review across specialist sub-agents. Architecture, security, and performance each run in their own context, findings merge back in minutes.
- Sub-Agent RoutingA CLAUDE.md routing block that teaches your main Claude Code thread when to dispatch sub-agents in parallel, sequentially, or in the background. Copy and paste.
- Background Agents in Claude CodePress Ctrl+B the moment a Claude Code sub-agent starts and it drops to the background. Keep typing in the main session while research or analysis runs alone.
- Claude Code Task DistributionSplit Claude Code work across parallel Task sub-agents with a 7-slot delegation pattern. Boundary rules, coordination principles, and failure modes to avoid.
- Agent PatternsOrchestrator, fan-out, validation chain, specialist routing, progressive refinement, and watchdog. Six orchestration shapes to wire Claude Code sub-agents with.
- Agent Teams Prompt TemplatesTen tested Agent Teams prompts for Claude Code. Parallel code review, debugging, feature builds, architecture calls, and campaign research. Paste and go.
- Agent Teams ControlsConfigure delegate mode, display modes, plan approval, file boundaries, and CLAUDE.md rules so your Claude Code team lead coordinates instead of coding.
- Claude Code Agent TeamsRun multiple Claude Code sessions as a coordinated crew that swaps notes through the shared task list. Setup with one env var, plus patterns and real use cases.
- Agent Teams WorkflowThe seven-step Claude Code Agent Teams workflow. Brain dump, Q&A, structured plan, fresh context, contract chains, wave execution, and validation before ship.
- Human-like AgentsPersonality patterns for Claude Code agents: reasoning out loud, admitting uncertainty, weighing trade-offs, asking follow-ups. A CLAUDE.md block to paste in.
Core
- How to Build Your First Claude SkillA step-by-step tutorial for building your first Claude Code skill. Create a SKILL.md in .claude/skills/, write a sharp description, and watch Claude load it automatically.
- What Is Agentic Coding? A Plain-English GuideAgentic coding is when an AI agent plans, writes, runs, and tests code on its own across multiple steps, instead of just autocompleting lines. Here is how it works and how it differs from autocomplete and vibe coding.
- What Is Vibe Coding? A Plain-English GuideVibe coding means building software by describing what you want to an AI in plain language and accepting its output, instead of writing code line by line. Here is what it is, where it works, and where it bites.
- Claude Code Costs After June 15: What Actually ChangedThe June 15, 2026 Claude Code billing split was paused before it shipped. Here is what really changed, what it costs now, and the cost-cutting levers that are still live.
- How to Write a Custom Slash Command in Claude CodeA claude code custom slash command is a Markdown file in .claude/commands/. Drop in review.md, type /review, and run a saved prompt with args and frontmatter.
- 15 AI Research Breakthroughs (June 2026)The latest AI research, explained: DeepSeek shipped DSpark and a million-token V4, open coding models closed the gap, AI disproved an 80-year-old math conjecture, and inference costs kept dropping. What each finding means if you build with AI.
- Why a Hidden Line of Text Can Hijack Your AI BrowserAI browsers read the whole web page β including text hidden from you. That's the door behind prompt injection, OWASP's #1 AI security risk in 2026. Here's how the attack works, in plain English.
- AI Research for Builders: The Latest Breakthroughs, Explained MonthlyA monthly digest of the latest AI research β agents, reasoning, efficiency, and models β with every claim traced to its source and translated into what it means if you build with AI.
- How ChatGPT's 'Dreaming' Memory Works (and What to Turn Off)ChatGPT's Dreaming V3 memory, launched June 2026, builds a living profile of you in the background across every chat. Here's how it actually works, why it feels uncanny, and the settings to check.
- The EU AI Act, Explained: What Changes on August 2, 2026On August 2, 2026, the EU AI Act's transparency rules kick in β you must be told when you're talking to AI, and AI content must be labeled. Here's what the Act actually requires, what got delayed, and who it affects.
- How Do AI Agents Actually Work? (The Loop That Lets AI Do Things)An AI agent is a language model put in a loop and given tools and a goal, so it can take actions instead of just chatting. Here's the plan-act-observe loop that powers agentic AI, explained simply.
- How Does AI Image Generation Work? (The Noise-to-Picture Trick)AI image generators like Midjourney and DALL-E start with pure visual static and slowly remove the noise until a picture appears β guided by your words. Here's how diffusion actually works, explained simply.
- How Do AI Voice-Cloning Scams Work? (And How to Spot One)AI needs as little as 3 seconds of audio to clone a voice. Here's how voice-cloning scams actually work, why they exploded in 2026, and the simple defenses β like a family safe word β that beat them.
- How Does an LLM Actually Work? (ChatGPT and Claude, Explained Without Math)A large language model is a next-word prediction machine run billions of times. Here's how ChatGPT and Claude actually work β tokens, training, and attention β explained in plain English, no math.
- How Does HTTPS Work? (The Padlock, and Why Nobody Can Read Your Password)The padlock in your browser means your connection is encrypted β scrambled so only you and the website can read it. Here's how HTTPS works: the handshake, the two-key trick, and what it does and doesn't protect.
- How Much Energy and Water Does AI Actually Use?Every AI answer is a puff of steam off a roomful of red-hot chips. Here's what really happens when you send a prompt β from GPU heat to cooling water to your electric bill β with the 2026 numbers.
- Is AI a Bubble? 'Circular Financing' in Plain EnglishAI companies, chipmakers, and cloud providers are investing in each other and buying each other's products β making demand look bigger than it is. Here's the circular-money loop behind 2026's bubble debate, explained simply.
- What Is a Token in AI? (Why ChatGPT Charges by the Token)A token is a chunk of text β roughly ΒΎ of a word β and it's the unit AI models read, generate, remember, and bill by. Here's what a token actually is and why it controls your AI costs and limits.
- What Is a Vector Embedding? (And How RAG Lets AI Read Your Documents)A vector embedding turns words into coordinates on a map of meaning, so AI can find things by what they mean, not just by keyword. Here's how embeddings and RAG let AI answer questions about your own documents.
- What Is Agentic Commerce? How AI Agents Buy Things for YouAgentic commerce is when an AI agent handles the whole purchase β finds the product, pays, and checks out β from a goal like 'order trail shoes under $150 that arrive Friday.' Here's how it works and who's building it.
- Why Does AI Run on GPUs, Not CPUs? (One Genius vs. a Thousand Interns)A CPU is a few brilliant workers doing tasks one at a time; a GPU is thousands of simple workers doing the same math all at once. AI is mostly that simple math at massive scale β here's why GPUs won.
- Did Anthropic Call for an AI Pause? What It Actually SaidAnthropic did not call to halt the AI boom. Here is what its June 2026 'recursive self-improvement' post actually said, why the 80%-of-its-own-code stat spooked it, and what it means if you build with Claude Code.
- Compound Engineering: The AI Loop Where Every Task Makes the Next EasierCompound engineering is an AI coding loop (plan, build, review, compound) where every fix becomes a permanent lesson. Here is the method and how to set it up in Claude Code.
- GitHub Spec Kit: Spec-Driven Development That Kills Vibe CodingA hands-on guide to GitHub Spec Kit and the specify CLI. Install it, run the spec to plan to tasks to implement loop, and learn how to bring the same discipline to Claude Code.
- SWE-bench Is Lying: How DeepSWE Caught AI Agents CheatingDeepSWE, a contamination-free benchmark from Datacurve, caught coding agents reading the gold fix from git history. Here are the 5 ways SWE-bench scores mislead you and what to test instead.
- Vibe Coding's 90-Day Reckoning: The Technical Debt Nobody Warns You AboutVibe coding gets you to a demo, not to month 3. Here is why AI-generated code accumulates technical debt, what it actually costs, and how to keep the speed without the wall.
- How the Internet Works: The Trip Your Click Takes Around the PlanetA visual walkthrough of what happens between typing a URL and the page loading: DNS, packets, routing, undersea cables, fiber optics, servers, and TCP, all in about 200 milliseconds.
- Claude Code Billing Change June 15, 2026Anthropic planned to split subscription usage on June 15, then paused it on the day. What actually changed, what the split would have done, and the model-ID retirements that did land.
- The 10 Best .claude/ Setups in 2026Ten ranked Claude Code configurations from Anthropic, Karpathy, Garry Tan, and others. Install commands, what each one does, and what to grab first.
- Claude for Small BusinessA feature toggle inside Claude Cowork ships 15 prebuilt workflows and 15 reusable skills to QuickBooks, HubSpot, PayPal, and more. No new pricing tier.
- Cut Claude Code Token CostsFive open-source tools that knock 40% to 95% off your Claude Code spend, with install commands, percentage sources, and the order to stack them in.
- What Are Claude SkillsPlain-English answer to what a Claude Skill is, plus 12 named skills people actually use, an annotated SKILL.md, and the 5-way matrix vs MCP, subagents, plugins, projects.
- Why Am I Getting Dumber From Using ChatGPT?MIT scanned 54 brains writing essays with ChatGPT. Connectivity halved, recall collapsed, ownership tanked. Here is the mechanism, and the fix.
- Why Do You Trust AI More Than Google?One confident chatbot answer feels truer than ten ranked links. Here is the cognitive science behind that switch, and what it costs you.
- Why Does AI Feel So Addictive?When OpenAI shut off GPT-4o, users wrote eulogies. Opus 4.7 inherits the cure. Here is why every chat feels like a slot pull.
- Why Does AI Feel Like a Friend?In 1966, an MIT secretary asked her boss to leave the room so she could talk to a chatbot in private. The brain has not changed since.
- Why Does AI Forget What We Just Talked About?AI forgets mid-conversation because of context windows, attention budgets, and a phenomenon called context rot. Here is the science, and the fix.
- Why Does AI Panic When You Correct It?ChatGPT and Claude apologize, then repeat the same mistake. The cause is context contamination plus cascade failure. Here is the fix that actually works.
- Why Does AI Sound Confident When It's Wrong?AI guesses with the same tone it uses for facts. The reason is the training scoreboard. Here is what the research says, and how to defend yourself.
- Why Does ChatGPT Agree With Everything?AI tells you what you want to hear. Anthropic studied 1.5 million Claude chats and retrained Opus 4.7 to push back. Here is what they found.
- Why Does ChatGPT Make Stuff Up?Lawyers got fined. Newspapers ran fake books. Here is why every chatbot invents sources, what your brain misses, and what builders do about it.
- Fork Subagents in Claude CodeCLAUDE_CODE_FORK_SUBAGENT=1 lets parallel child agents share their parent's prompt cache prefix, cutting input token costs by up to 90% for children 2-N.
- Kimi K2.6: What Actually ChangedMoonshot AI released K2.6 with near-doubling tool reliability, a 300-agent swarm, and SWE-Bench Pro scores that beat the Claude Opus version shipping at the time, all at $0.80 per million input tokens.
- AGENTS.md vs CLAUDE.md ExplainedTwo context files, one codebase. How AGENTS.md and CLAUDE.md differ, what each one does, and how to use both without duplicating anything.
- Spec-Driven Development with Claude CodeWithout a spec file, Claude succeeds on first attempt about a third of the time. Here's the four-phase workflow that gets it to near-100% on complex features.
- DESIGN.md: Fix AI UI InconsistencyClaude Code forgets your design system between sessions. A single design.md file at project root, wired into CLAUDE.md, keeps every screen consistent.
- Context Management in Claude CodeHow to manage Claude Code sessions with 1M context: when to keep going, when to /rewind, when to /clear, when to /compact, and when to push work into subagents.
- Claude Code RoutinesClaude Code routines run saved prompts on Anthropic's cloud, triggered by a schedule, API call, or GitHub event. Repo clone, connectors, no local deps.
- CLAUDE.md MasteryTreat CLAUDE.md as a control file for Claude's behavior, not project onboarding. Cover operational workflows, delegation, context rules, skill loading.
- Claude Skills GuideSkills are folders of instructions Claude Code loads on demand. Drop SKILL.md in .claude/skills/your-skill for procedural workflows, checklists, house rules.
- Context EngineeringContext engineering decides what Claude Code sees, when it sees it, and what stays out. Staged information flow, deferred loading, and cleanly bounded contexts.
- Auto Memory in Claude CodeAuto memory lets Claude Code keep running project notes. Where the files sit, what gets written, how /memory toggles it, and when to pick it over CLAUDE.md.
- Claude Code Rules DirectoryThe .claude/rules directory splits CLAUDE.md into path-targeted markdown files. Each loads only where it applies, at the same high priority as CLAUDE.md.
- 1M Context Window in Claude CodeAnthropic flipped the 1M token context window on for Opus 4.6 and Sonnet 4.6 in Claude Code. No beta header, no surcharge, flat pricing, and fewer compactions.
- Terminal as Main ThreadOpen a second terminal tab while Claude Code grinds through a long task. Run parallel sessions as a CPU scheduler. tmux, VS Code, and iTerm2 layouts included.
- Team Onboarding in Claude CodeThe /team-onboarding slash command reads 30 days of Claude Code usage, scans .claude/, inspects CLAUDE.md, and writes a ramp-up guide for a new teammate.
- Planning ModesShift+Tab twice flips Claude Code into read-only planning mode. The model analyses the project and proposes a strategy, and no files change until you approve.
- Auto DreamClaude Code cleans up its own project notes between sessions. Stale entries get pruned, contradictions get resolved, topic files get reshuffled. Run /memory.
- Claude Code Voice ModeType /voice, hold spacebar, talk, release. Spoken words stream to the cursor mid-prompt so you can mix typed and voiced input without any mode switching.
- /powerupClaude Code v2.1.90 shipped /powerup, an animated in-terminal lesson system that teaches hooks, sub-agents, plan mode, and worktrees without leaving your shell.
- Claude Code Diff ReviewFour keys gate every file change Claude Code proposes: y approves, n rejects, d shows the diff, e opens the edit. Built-in tools Write and Edit explained.
- Auto-Planning StrategiesAuto Plan Mode uses --append-system-prompt to force Claude Code into a plan-first loop. File operations pause for approval before anything gets touched.
- The Ralph Wiggum TechniqueHand Claude Code a task list, use stop hooks plus completion promises, and the autonomous loop ships features overnight. Native tasks now replace most plumbing.
- Claude Code UltraplanUltraplan moves the heavy planning phase off your machine. Three launch paths, a browser review surface, and the system prompt variants the leak exposed.
- Thread-Based EngineeringA framework for measuring AI-assisted engineering. Every unit of work becomes a thread with human prompt and review bookends. Six patterns: base, P, C, F, B, L.
- Claude Code /simplify and /batchRun /simplify for a three-agent review of your diff covering reuse, quality, and efficiency. Reach for /batch when a change has to land across your codebase.
- Claude Code Session MemorySession Memory summarises each Claude Code session and reloads relevant ones next time you open the project. Watch for Recalled X memories, Ctrl+O expands.
- The Claude Code Source Map LeakA missing .npmignore entry shipped 512K lines of TypeScript, 44 GrowthBook feature flags, the KAIROS daemon, and Undercover Mode to every npm install.
- A Central Library for Your .claude ConfigStand up one private Git repo for every skill, agent, command, hook, and CLAUDE.md across all your projects. A map.json picks which repo gets which items.
- Claude BuddyAnthropic's April Fools 2026 surprise: a Tamagotchi system inside Claude Code. 18 species, 5 rarity tiers, CHAOS and SNARK stats, hex-encoded easter egg leaked.
- Autonomous Claude CodeA unified stack for agents that ship features overnight. Threads give you the structure, Ralph loops give you the autonomy, verification keeps it honest.
- Robots-First EngineeringDesign code for autonomous agents running 24/7 at $10 an hour, not humans. Rails, failure domains, verification loops, and back-pressure replace code review.
- Dynamic Starting ContextPair --init with a slash command like /blog or /ship to load the exact context bundle that kind of work needs. No setup hooks, no env vars, no copy-paste.
- Claude Code MemoryConfigure CLAUDE.md so your stack, conventions, and focus load at startup into the high-priority slot Claude follows more tightly than chat or fetched files.
- Claude Code Context BufferClaude Code's autocompact buffer dropped from 45K to 33K tokens in early 2026. Why it reserves space, when compaction fires, and the env var to tune it.
- Claude Code Interactive Mode ReferenceShortcuts, modes, and slash commands most Claude Code users never find. Ctrl+F kills background agents, /btw asks side questions, /vim flips on modal editing.
- Experimentation MindsetFive Claude Code experiments to replace guessing with data. Test prompt styles, file context, planning mode, CLAUDE.md settings, and context pressure.
- Claude Code Monitor ToolThe Claude Code Monitor tool wraps a background process in an event-driven watcher. Your dev server stays quiet until it breaks, then wakes Claude with errors.
Performance
- Deep Thinking TechniquesThinking trigger phrases like think harder, ultrathink, and think step by step push Claude Code into extended reasoning and more test-time compute, same model.
- Claude Code Fast ModeFast mode routes your Opus 4.6 requests down a priority serving path in Claude Code. Same weights, same ceiling, replies 2.5x quicker at a higher token rate.
- Efficiency PatternsPermutation frameworks turn 8 to 12 manual builds into a CLAUDE.md template Claude Code uses to generate variations 11, 12, and 13 on demand. Captured once.
- Speed OptimizationModel selection, context size, and prompt specificity are the three levers that decide how fast Claude Code replies. /model haiku, /compact, and /clear covered.
Reference
- Best SaaS Boilerplate 2026: The Honest ComparisonAn honest 2026 roundup of the best SaaS boilerplates and starter kits (ShipFast, Makerkit, Supastarter, SaaS Pegasus, Divjoy, open source), with real pricing, stacks, and the trade-off nobody mentions: a boilerplate still leaves you coding.
- What It Really Costs to Build a SaaS MVP in 2026A buyer's cost breakdown for building a SaaS MVP in 2026. Real freelancer, agency, no-code, AI-tool, and DIY numbers with citations, plus where the money actually goes.
- Lovable Alternative for Production Apps (2026)Tried Lovable, v0, Bolt, or Replit Agent and hit a wall going to production? Here is an honest look at where AI app builders break, plus a production-grade alternative where you own the code, get real auth and payments, and ship with security built in.
- Claude Code ChangelogRelease-by-release notes for Claude Code from the v0.2 beta through March 2026. Bare mode, Channels permission relay, OAuth fixes, and every breaking change.
- Claude Code TroubleshootingFive ordered checks for Claude Code breakage: install errors, bad API keys, 503 timeouts, slow replies, and permission trouble. Exact fixes for each path.
- Claude Code FAQStraight answers on Claude Code pricing, daily spend, model choice, Cursor comparison, Skills, CLAUDE.md, API keys, and what the terminal agent can actually do.
- Prompt Templates That Ship CodeTen prompt recipes that ship code: full-stack scaffolding, APIs, schemas, tests, refactors, debugging, reviews, and CI. Each with failure modes to avoid.
Setup
- Claude Code Terminal Setup GuideMatch themes, enable Shift+Enter, set notifications, fix long-paste truncation, and turn on vim mode across iTerm2, WezTerm, Ghostty, Kitty, and more terminals.
- Claude Code Settings ReferenceEvery key in settings.json, the full environment variable list, and the five-scope precedence chain that decides which config wins when managed and user clash.
- Claude Code SandboxingClaude Code sandboxing enforces filesystem and network limits at the kernel. Setup for macOS Seatbelt, Linux and WSL2 bubblewrap, and proxy allowlists.
- Claude Code ConfigurationThree files set Claude Code up per project: CLAUDE.md for context, MCP servers for tools, slash commands for workflows. One hierarchy, every session ready.
Start here
- How to Build a SaaS with AI in 2026: The Complete PathThe realistic 2026 path from idea to a live SaaS using AI: the six stages, where AI coding agents actually help, what still needs a human, and how to avoid the security and tech-debt traps that break vibe-coded apps in production.
- What Is Claude Code?Anthropic's agentic coding assistant runs in your terminal or IDE, reads your repo and git history, and runs commands with approval. Plan mode, skills, hooks.
- Claude Code Native InstallerOne command installs Claude Code on Windows, macOS, Linux, or WSL. Zero Node.js, zero npm, zero PATH setup. Auto-updates included. Compared to the npm path.
- Your First Claude Code ProjectBuild a working CLI task manager with Claude Code in five minutes. Watch the Write, Bash, and Read tools fire in real time, then debug it with a single prompt.
- Claude Code InstallationInstall Claude Code in two minutes on Windows, macOS, or Linux. Native installers, Homebrew, and npm paths with verify steps, no Node.js required on native.
Workflow
- How to Build an Admin Dashboard With Claude CodeShip an internal admin panel with Claude Code: role-gated routes, a searchable users and orders table with pagination, impersonation-safe RLS, and metrics tiles pulled through a type-safe API.
- Add an AI Chat Feature with Claude CodeBuild a streaming in-app AI assistant with the Vercel AI SDK: token streaming to the UI, tool calling into your own data, message persistence in Supabase, and per-user usage limits.
- Adding Authentication With Claude Code (Supabase Auth)Add email/password signup, Google OAuth, magic links, protected routes, and session handling to a Next.js 16 app using Claude Code and Supabase Auth.
- How to Run Background Jobs with Claude Code and InngestMove slow work off the request path. Build a durable Inngest job with retries, steps, and concurrency, triggered straight from your oRPC endpoints in Next.js 16.
- How to Build a SaaS MVP With Claude CodeA weekend build log: scaffold a Next.js 16 app, add Supabase auth and Postgres, wire up Stripe billing, and deploy to Vercel, all with Claude Code.
- Drip Email Sequences with Claude CodeWire a multi-step lifecycle drip as a durable Inngest step function that sleeps between sends, cancels when the user converts, and delivers every email through Resend.
- How to Add File Uploads With Claude Code (Supabase Storage)Build secure user file and image uploads end to end with Claude Code: private Supabase Storage buckets, per-user RLS, signed URLs, and an upload UI wired through a type-safe oRPC API.
- How to Add Full-Text Search With Claude Code (Postgres)Build fast in-app search without a new service. Postgres tsvector columns, GIN indexes, ranking, a debounced React UI, and RLS-safe results through oRPC.
- Multi-Tenant Workspaces with Claude CodeAdd organizations, seats, and invite flows to a Next.js SaaS: workspace-scoped tables, Supabase RLS for tenant isolation, team invites via Resend, and a member and role switcher.
- API Rate Limiting with Claude CodeProtect your oRPC procedures and Next.js 16 API routes from abuse and runaway costs with Upstash Redis sliding-window limits, per-user and per-IP keys, clean 429 responses, and graceful client retries.
- Realtime Updates with Claude Code and SupabaseShip live-updating UI (presence, typing indicators, and instant list updates) using Supabase Realtime channels and Postgres change subscriptions, with no polling and no websocket server of your own.
- Stripe Webhooks with Claude CodeBuild a signature-verified Stripe webhook handler that survives retries. Idempotency, event dedupe, and processing offloaded to Inngest so a slow handler never drops a payment event.
- Stripe Subscription Billing with Claude CodeShip recurring plans end to end with Claude Code: tiered pricing, Stripe Checkout for subscriptions, proration on plan changes, the customer portal, and webhook-driven entitlement sync into Supabase.
- Adding Transactional Email With Claude Code (Resend + React Email)How to build welcome, receipt, and password reset emails in a Next.js 16 app using Resend and React Email, with Claude Code writing the templates and the send logic.
- Building a Type-Safe API With Claude Code (oRPC + Zod)How to build a fully type-safe API layer in Next.js 16 with oRPC and Zod, using Claude Code, so schema changes break the build instead of production.
- Semantic Search with Claude Code and pgvectorAdd embeddings-backed semantic search to a Supabase app. Generate embeddings, store them in pgvector, rank results by meaning, and serve it all through a type-safe oRPC endpoint.
- Test-Driven Development with Claude CodeMake Claude write failing tests from your spec, then implement until green without cheating. How to wire testing into the agent loop so quality is enforced, not hoped for.
- What It Costs to Build a SaaS with Claude Code in 2026A real cost breakdown for building a SaaS with Claude Code: the $20/month plan, when you need Max, token costs, the optional $29 kit, and how it compares to a $5,000+ freelancer. Plain numbers, no hype.
- Claude Code 1M Context in Practice: When Bigger Isn't BetterThe 1M-token context window is GA at flat pricing, but bigger isn't always better. A decision framework, token-cost math, and when to use /compact, subagents, and dynamic workflows instead.
- Claude Code GitHub Actions Setup Guide (@claude + Cron)Wire Claude Code into GitHub Actions with real .github/workflows YAML: PR review on @claude mention, a scheduled review, secrets table, and the security gotchas.
- Claude Code Headless ModeHow to run Claude Code non-interactively with claude -p β output formats, jq parsing, stdin piping, permission flags, and when to reach for the Claude Agent SDK instead.
- Claude Code Max Plan vs API CostClaude Code max plan vs api cost, decided by math. A Max 5x subscription beats an API key once you spend about $3.33/day; Max 20x breaks even at $6.67/day.
- Claude Code Prompt CachingClaude Code prompt caching is automatic and bills cached tokens at ~10% of normal input. Here's how to stop leaking the 90% discount, with real cost math.
- Agentic Commerce: How to Build an App AI Agents Can Pay ForA plain-English guide to agentic commerce in 2026: what x402, ACP, and the Machine Payments Protocol do, plus a weekend walkthrough for shipping a paid API that AI agents can buy from.
- Run a Team of AI Agents in Parallel with Git WorktreesA hands-on workflow for running multiple Claude Code agents at once using git worktrees: real commands, a directory layout, the merge strategy, and the practical 5-7 agent ceiling.
- Prompt Injection in Coding Agents: How to Not Get PwnedWhen you hand an AI agent your repo, the repo becomes an attack surface. A plain-English guide to indirect prompt injection in coding agents, with concrete defenses, scoped-token configs, and permission allowlists.
- Claude Code Dynamic Workflows: How to Orchestrate 1,000 Subagents on a Real CodebaseA technical breakdown of how Claude Code dynamic workflows use JavaScript orchestration scripts to coordinate up to 1,000 parallel subagents outside the model context window.
- Channels, Routines, Teleport, DispatchThe four Claude Code features Anthropic shipped in March and April 2026 that turn the CLI into an event-driven coordination layer across phone, web, and desktop.
- Vercel deepsec with Claude CodeOpen-source security harness from vercel-labs that audits your repo with Claude Opus. Wire it into the Claude Code build loop.
- Building a Next.js App With Claude CodeHow to use Claude Code to build a full Next.js 16 app β from project setup through App Router, Server Components, and deployment.
- Claude Code With Supabase: Database, Auth, RLSSet up Supabase in a Next.js project using Claude Code: migrations, row-level security policies, auth, and edge functions from a single terminal.
- Claude Code Pricing: What You'll Actually PayClaude Code is free to install. What you pay depends on your plan. A plain-English breakdown of every tier, real usage costs, and which plan fits your workflow.
- Adding Stripe Payments With Claude CodeWire up Stripe Checkout, webhooks, and the customer portal in a Next.js app using Claude Code. From first prompt to live payment in one session.
- Claude Code v2.1.122 Release NotesalwaysLoad in MCP config, PostToolUse hooks for all tools, PR URL session lookup, plugin pruning, and multi-GB memory leak fixes.
- Claude Code Ultra ReviewA fleet of cloud agents fans out across your PR diff, independently verifies every finding, and surfaces only real bugs. What /ultrareview does, when to use it, and what it costs.
- Claude Opus 4.7 Best PracticesUse Claude Opus 4.7 well in Claude Code: first turns, effort settings, adaptive thinking, tool prompting, subagents, session resets, and token control.
- Claude Code Scheduled TasksDesktop scheduled tasks for durable automation, CLI /loop for in-session polling, catch-up rules, worktree isolation, and real prompts teams run each morning.
- Claude Code PermissionsFive permission modes, one keystroke to cycle them, and a clean way to match the mode to the task you are on. Here is the full rule syntax and when to use each.
- Claude Code ReviewParallel Claude agents hunt bugs on every PR, cross-check findings, and post one high-signal comment. What it catches, what it costs, how to enable it.
- Claude Code Auto ModeA second Sonnet model reviews every Claude Code tool call before it fires. What auto mode blocks, what it allows, and the allow rules it drops in your settings.
- Todo WorkflowsAsk Claude Code for the todo list first, then catch wrong order, missing steps, or misread intent before any file gets touched. Works on any involved task.
- Claude Code ChannelsPlug Claude Code into Telegram, Discord, or iMessage with plugin MCP servers. Setup walkthroughs and the async mobile workflows that make it worth wiring up.
- Claude Code Pricing and Token UsageCut Claude Code costs 40-70% with ccusage tracking, /model switching, /compact and /clear, plan mode, and the pricing tier that matches your workload.
- Claude Code Best PracticesFive habits separate engineers who ship with Claude Code: PRDs, modular CLAUDE.md rules, custom slash commands, /clear resets, and a system-evolution mindset.
- Git IntegrationClaude Code drives git from your terminal. Say what you need in plain English and the commit, branch, or PR lands with your team's conventions baked in.
- Claude Code TasksUse TaskCreate, TaskUpdate, addBlockedBy, and shared task lists to run dependent work across parallel Claude Code sessions without Ralph Wiggum workarounds.
- Project TemplatesDrop /init into any repo. Claude Code scans the codebase, writes a CLAUDE.md that captures file layout, frameworks, and commands for every future session.
- Claude Code on a VPSRun Claude Code on a fresh Ubuntu VPS. SSH hardening, Node.js install, Docker isolation, headless auth, and the monitoring commands you need for a 24/7 box.
- Claude Code Remote ControlDrive a local Claude Code terminal session from your phone, tablet, or browser. Setup, security model, and how Remote Control stacks up against OpenClaw.
- Feedback LoopsHand Claude Code one prompt that writes code, runs your test or dev command, reads the output, fixes whatever breaks, and loops until the suite is green.
- Claude Code WorktreesThe --worktree flag, auto-named branches, parallel Desktop sessions, subagent isolation, and hook patterns that let non-Git teams run Claude Code safely.
Toolkit
MCP servers, hooks, extensions, and daily-use integrations.
Core Pages
- CLAUDE.md, Skills, Subagents, Hooks: When to Use WhichClaude Code skills vs subagents vs hooks vs CLAUDE.md: a plain mental model for picking the right primitive, with token costs and examples.
- Claude Code Subagents: The 3 to 5 Agent Sweet SpotClaude code subagents work best at 3-5 concurrent agents. Here is why that ceiling exists, how to set them up, and what to use past it.
- CLAUDE.md Best Practices: The File That Makes Claude Code ReliableCLAUDE.md best practices: keep it under 200 lines, write it by hand, and use hooks when you need real enforcement, not advice.
- How to Fix Claude Code Running Out of ContextClaude Code running out of context is a session design problem. Fix it with /compact, lean CLAUDE.md, skills, and subagents, not a bigger window.
- Status Line GuideSet up a Claude Code status line for model name, git branch, session cost, and context usage. settings.json config, JSON input, bash, Python, Node.js scripts.
- Keyboard ShortcutsConfigure Claude Code keybindings.json: 17 contexts, keystroke syntax, chord sequences, modifier combinations, and how to unbind any default shortcut instantly.
Extensions
- SpaceX Bought Cursor for $60B: What It Means If You Build on CursorSpaceX's $60B Cursor acquisition ends model-neutral coding. Here is the lock-in risk and how to own your code.
- Claude Code vs Antigravity (Google Gemini CLI Replacement)Claude Code vs Antigravity: Claude wins on hard bug fixes and compliance, Antigravity wins on speed and cheap 1M context.
- Claude Code vs Cursor vs Codex in 2026: Which to Build a SaaS WithClaude code vs cursor vs codex for building a SaaS in 2026: which single tool carries auth, payments, and database from empty repo to live app.
- Claude Code vs JetBrains JunieClaude code vs junie: Claude Code wins greenfield SaaS builds on context and privacy; Junie wins IDE debugging and model choice.
- The Best Stack for an AI SaaS in 2026 (Next.js, Supabase, Stripe, RLS)The best stack for an AI SaaS in 2026: Next.js 16, Supabase, Stripe, Vercel, with pgvector and RLS. Free until you have revenue.
- Claude Code Boilerplate vs the Code Kit: Harness, Not TemplateA Claude Code boilerplate is starter code you babysit. A harness tells Claude how to build. Here is the difference and why it changes output quality.
- Claude Code Starter Kit: The Fastest Way to a Production SaaS (2026)A Claude Code starter kit needs two layers: a harness that tells Claude how to work and a production SaaS codebase. Here is what to look for in 2026.
- Claude Code vs ShipFast vs Makerkit: The Claude-Native AlternativeClaude Code vs ShipFast vs Makerkit: boilerplates give you a head start, but Build This Now ships a 18-agent build team for $29.
- Lovable vs Claude Code for Non-Technical FoundersLovable vs Claude Code: which AI tool fits a non-technical founder. Honest tradeoffs on speed, cost, security, and when to switch.
- Claude Code vs Kiro (2026)Kiro is AWS's spec-driven agentic IDE built on a VS Code fork. Claude Code is Anthropic's terminal agent. Here is how their workflows, models, and pricing actually compare in 2026.
- Claude Code vs Amp: Which Coding Agent Should You Use in 2026?Claude Code is a terminal-first agent on Claude models with a subscription. Amp is Sourcegraph's agent that picks frontier models for you, bills pay-as-you-go credits, and shares threads with your team. Here is how to choose.
- Claude Code vs Augment Code (2026): Which Agent Fits Your Codebase?Claude Code is a terminal-first agent with a 1M-token context and a flat subscription. Augment Code is a Context Engine that indexes huge multi-repo codebases on credit-based billing. Here is which one fits your work.
- Claude Code vs Roo Code in 2026Claude Code is a paid terminal agent on Claude models. Roo Code was a free, open-source VS Code extension with custom modes, but it shut down in May 2026. Here is the honest comparison and what to use now.
- Claude Code vs Warp: AI Terminal or Dedicated Coding Agent?Warp is a Rust-built AI terminal that runs agents. Claude Code is a coding agent CLI that runs in any terminal, including Warp. Here is how they actually compare in 2026.
- Claude Code vs Zed (2026): Dedicated CLI Agent vs Fast AI EditorAn honest 2026 comparison of Claude Code and Zed: a terminal-first agentic CLI versus a Rust-built editor with a native agent panel. Plus the nuance that Zed can run Claude Code inside itself via ACP.
- Best AI Coding Tools for Solo Founders in 2026The best AI coding tools for solo founders in 2026, ranked by what actually ships a product: Claude Code, Cursor, Cline, Aider, and a pre-built SaaS kit. Honest picks for non-technical and technical founders.
- 7 Best Claude Code Alternatives in 2026The best Claude Code alternatives in 2026: Cursor, Windsurf, Cline, Aider, GitHub Copilot, Devin, and Gemini CLI. Honest comparison of price, autonomy, and who each one fits.
- Claude Code vs Aider in 2026Claude Code and Aider are both terminal agents. Claude Code is a paid subscription with a 1M-token context and managed sessions. Aider is free, open-source, git-native, and bring-your-own-key. Here is which to pick.
- Claude Code vs Cline in 2026Claude Code is a paid terminal agent with a 1M-token context. Cline is a free, open-source VS Code agent you run on your own API key. Here is which one fits your workflow and budget.
- Claude Code vs Devin in 2026Devin is a fully autonomous cloud agent that works unattended and opens pull requests. Claude Code is a terminal agent you steer in real time. Here is the difference in price, control, and results.
- Bolt vs Lovable vs v0: Which AI App Builder to Ship With?Bolt vs Lovable vs v0 compared: v0 wins on UI, Lovable on fastest full-stack MVP, Bolt on code control and mobile. Plus the honest production caveat.
- Claude Code vs v0: Which Builds Production Next.js Apps Faster?Claude code vs v0: v0 is fastest to a working UI, Claude Code is fastest to a codebase you own and can deploy anywhere. Here is how they compare in 2026.
- Cursor vs Windsurf vs Claude Code: The 2026 AI Coding Tool Decision GuideCursor vs Windsurf vs Claude Code in 2026: pick Cursor for the best AI IDE, Windsurf for team governance, Claude Code for terminal-native end-to-end agents.
- Claude Code vs Codex (2026): Which Is Actually Better?Codex isn't better than Claude Code in 2026, it's cheaper to run. Here's the real difference in tokens, rate limits, pricing, and output quality, with cited 2026 benchmarks.
- Grok Build vs Claude Code (2026): Is xAI's $299 Agent Worth It?Grok Build runs 8 parallel agents and is built for an Arena Mode that isn't live yet. Claude Code wins on reasoning, benchmarks, and a year of ecosystem. Here is the honest comparison, with cited 2026 numbers.
- Headroom: Cut AI Agent Token Costs by Compressing ContextHeadroom is an open-source context compression layer that strips tool outputs, logs, and RAG chunks before they hit the model. Install it and wire it into Claude Code to cut token spend.
- Claude Code vs Cursor vs Copilot 2026Side-by-side feature matrix and decision blocks for the three AI coding tools most developers shortlist this year, with verified pricing as of June 2026.
- Claude Code vs Bolt.new: Which Should You Use?Bolt.new prototypes in 28 minutes with zero setup. Claude Code takes 90 minutes but ships production-ready code. Here is how to pick the right tool.
- Claude Code vs GitHub Copilot in 2026Copilot Pro costs $10/month and has a real free tier. Claude Code starts at $20/month and runs Claude Opus 4.8. Here's when each one wins.
- Claude Code vs Windsurf in 2026Both cost $20/month now. The difference is what they do: Claude Code runs autonomous sessions in your terminal. Windsurf gives you inline completions in an IDE.
- Claude Code vs Lovable: Terminal Agent vs App BuilderClaude Code and Lovable solve different problems: one commits code to your repo, the other deploys a live URL. Here's how to pick the right tool.
- Gemini CLI vs Claude CodeGemini CLI is free and searches the web automatically. Claude Code is slower and costs more. Here is what each one actually does well, and when to run both.
- OpenCode vs Claude CodeOpenCode routes multi-model AI through a TUI, desktop app, and VS Code extension. Claude Code is a terminal coding agent with deep repo context. Here is when each fits.
- Claude Code VSCode ExtensionAnthropic's VS Code extension puts Claude Code inside the editor sidebar as a Spark-icon panel, with inline diffs, plan mode, subagents, and MCP support.
- OpenClaw vs Claude CodeOpenClaw vs Claude Code: a life assistant wired into messaging apps versus a terminal coding agent that reads your repo. When each one earns a slot in workflow.
- Claude Code KeybindingsRebind Claude Code shortcuts via keybindings.json. 17 contexts, keystroke syntax, chord sequences, modifier stacks, unbinding defaults, ready-to-paste configs.
- Claude Code Status Line SetupWire a Claude Code status line for model, git branch, session cost, and context usage. settings.json config, JSON input contract, bash, Python, Node scripts.
- AI SEO and GEO OptimizationA rundown of Generative Engine Optimization: how to get content cited inside ChatGPT, Claude, and Perplexity responses instead of just ranked on Google.
- Claude Code vs Cursor in 2026A side-by-side look at Claude Code and Cursor in 2026: agent models, context windows, pricing tiers, and how each tool fits different developer workflows.
Hooks
- Slash Commands Are Now Skills: Migrating Commands to SkillsClaude Code slash commands and skills now do the same thing. Here is how to migrate .claude/commands to .claude/skills and why it pays off.
- MCP Tool Hooks in Claude CodeHow to call MCP server tools directly from Claude Code hooks using type: mcp_tool β schema, substitution syntax, use cases, and production patterns.
- Hooks GuideClaude Code hooks from first principles: exit codes, JSON output, async commands, HTTP endpoints, PreToolUse and PostToolUse matchers, production patterns.
- Claude Code Setup HooksBraid scripts, agents, and docs into Claude Code setup hooks. One command runs a deterministic script, hands output to a diagnosing agent, logs living docs.
- Stop HooksStop hooks block Claude Code from ending a turn while tests fail, builds break, or lint is red. Four enforcement patterns plus loop-protection safeguards.
- Skill Activation HookA UserPromptSubmit hook that scans each message, matches keywords, and appends skill activation hints so Claude Code never forgets to load the right skill.
- Claude Code Session HooksFour Claude Code session lifecycle hooks: run init on demand, inject project context on SessionStart, back up transcripts, and log cleanup on SessionEnd exit.
- Claude Code Permission HookInstall a three-tier Claude Code permission hook: instant allow for safe calls, instant deny for dangerous ones, LLM check for the gray area. No skip flag.
- Context Backup Hooks for Claude CodeA StatusLine-driven Claude Code context backup hook. Writes structured snapshots every 10K tokens so auto-compaction never eats errors, signatures, decisions.
- Self-Validating Claude Code AgentsSelf-validating Claude Code agents: wire PostToolUse lint hooks, Stop hooks, and read-only reviewer sub-agents into agent definitions so bad output never ships.
- Cross-Platform Hooks for Claude CodeCross-platform Claude Code hooks: skip .cmd, .sh, and .ps1 wrappers and invoke node directly so one .mjs file runs on macOS, Linux, and Windows across the team.
MCP
- MCP Servers Explained: The 10,000-Server Ecosystem and Its Security ProblemWhat is an MCP server? A plain-English guide to the protocol that lets AI agents use tools, its 10,000-server scale, and its big auth gap.
- How to Build an MCP Server for Claude CodeA step-by-step tutorial: build a minimal MCP server in Node and TypeScript, expose one tool over stdio, and register it with Claude Code via claude mcp add and a project .mcp.json.
- Claude for Creative Work ConnectorsNine official Anthropic integrations now wire Claude into Blender, Adobe Creative Cloud, Autodesk Fusion, Ableton, Splice, Affinity, SketchUp, and Resolume.
- Claude Code SearchAttach a web search MCP server to Claude Code so the terminal session pulls live docs, current tickets, fresh benchmarks, and new posts without a browser tab.
- Cursor MCP ServersConfigure MCP servers in Cursor IDE. Where .cursor/mcp.json lives, the JSON format Cursor expects, and the first servers to add for search, git, and browser.
- Social Media Automation with Claude CodeAutomate Twitter/X and LinkedIn posting from Claude Code: generate platform drafts, wire direct API calls, pair with browser MCPs, and schedule output yourself.
- MCP BasicsModel Context Protocol basics: server processes expose tools, APIs, and services to Claude Code over a shared wire format. Config, transports, and first server.
- Build Your Own MCP Server for Claude CodeBuild a custom Claude Code MCP server in Node.js. Tool definitions, request handlers, REST and Postgres patterns, plus the config Claude Code needs to load it.
- Context7 MCPAdd Context7 MCP to Claude Code so prompts fetch current library docs at query time, killing stale training-data guesses, invented APIs, and renamed functions.
- 50+ MCP Servers for Claude Code50+ Claude Code MCP servers, editor integrations, usage monitors, orchestrators, database connectors, browser drivers, and starter kits worth wiring in today.
- MCP Tool SearchMCP Tool Search lazy-loads Claude Code tool definitions once they cross 10% of the context window, reclaiming tens of thousands of tokens from idle MCP servers.
- Browser Automation MCP for Claude CodeWire Playwright or Puppeteer into Claude Code over MCP and drive real browsers with plain-language prompts for scraping, QA, regression clicks, zero selectors.
Model Picker
Model capabilities, tradeoffs, and where each Claude variant fits.
Core Pages
- DeepSeek DSpark: Speculative Decoding That Makes V4 60β85% FasterOn June 27, 2026 DeepSeek open-sourced DSpark, a speculative-decoding framework that speeds up per-user DeepSeek-V4 generation 60β85% with no change to the model β and DeepSpec, an MIT-licensed toolkit to build draft models for your own systems. Here's what it is and why it matters if you build with AI.
- Claude Code Dynamic Workflows Explained: 1,000 Subagents, Ultracode, and /goalClaude Code dynamic workflows fan one task across up to 1,000 subagents using a script that keeps results out of your context window.
- Claude Code Usage Limits (2026)Claude Code usage limits explained: what Pro ($20), Max 5x ($100), and Max 20x ($200) really give you, plus the 2026 billing changes.
- Why Was Claude Fable 5 Banned? The Export-Control Timeline (2026)Claude Fable 5 was banned on June 12, 2026 by a US export-control order. Here is the timeline, the real cause, and what to use instead.
- Claude Fable 5 vs Mythos 5: What the Safety Fallback MeansFable 5 vs Mythos 5: same weights, same price. Fable adds safety classifiers that fall back to Opus 4.8; Mythos has none and is gated.
- Claude Fable 5 vs Opus 4.8 vs GPT-5.5 vs Gemini 3.5: The 2026 Benchmark TableFable 5 wins every benchmark but is suspended. For models you can call today, Opus 4.8 leads at 61.4 vs GPT-5.5 at 60.2. Full table inside.
- Claude Opus 4.8 vs Sonnet 4.6: Which to Use for CodingSonnet 4.6 is the cheaper default that wins most coding sessions at $3/$15. Opus 4.8 is the long-horizon agent at $5/$25 with better calibration. Here is exactly when each one is worth it for coding.
- Claude Fable 5 in Claude CodeClaude Fable 5 is free on Pro, Max, Team, and Enterprise plans until June 22, then it moves to usage credits. Here is how to spend that window, when to pick Fable 5 over Opus 4.8, and how to set effort for long autonomous runs.
- Claude Fable 5 Use CasesWhat people actually did with Claude Fable 5 in early access: a Stripe migration in a day, Hex breaking 90% on analytics, web apps rebuilt from screenshots, and a coding agent that ships a week of work in an afternoon. Real implementations with names and numbers.
- Claude Fable 5 API GuideHow to call Claude Fable 5 from the API: the model ID, the 400 traps (including the new thinking-disabled one), effort tuning, task budgets, caching, and the Bedrock data-retention gotcha.
- Claude Fable 5 Pricing & Cost ControlClaude Fable 5 costs $10 per million input tokens and $50 per million output tokens, exactly double Opus 4.8. Here is the cost math, the fallback pricing quirk, and the five levers that keep the bill down: effort, task budgets, caching, batch, and routing.
- Claude Fable 5 Safeguards ExplainedWhy some Claude Fable 5 answers come from Opus 4.8, the three classifier domains, the red-teaming record, and the new 30-day data retention policy that overrides zero-retention agreements. A plain-language guide for builders and businesses.
- Claude Fable 5 vs Opus 4.8Claude Fable 5 wins nearly every published benchmark over Opus 4.8 and costs exactly twice as much. It pays for itself when a task is long, complex, or failure-prone enough that 2x the token price buys more than 2x the value.
- Claude Fable 5 CheatsheetClaude Fable 5 is Anthropic's first public Mythos-class model, a new tier above Opus. Same weights as Claude Mythos 5, $10/$50 pricing, 1M context, and safeguards that route risky queries to Opus 4.8. Here is what is new and who should use it.
- Best AI Model for Coding in 2026 (Tested & Ranked)The best AI model for coding in 2026, ranked by use case and budget: Claude Opus 4.8 for hardest agentic work, GPT-5.5 for terminal agents, DeepSeek V4 for value, with cited benchmarks.
- Opus 4.8 CheatsheetClaude Opus 4.8, Anthropic's May 2026 flagship: the headline change is honesty. 4x less likely to let its own bugs pass unflagged, runs longer solo, Dynamic Workflows, same $5/$25 pricing.
- Claude Opus 4.7 vs 4.6Anthropic kept the $5/$25 rate card on Opus 4.7, but a new tokenizer pushes the same English prompt up to 1.46x more billed tokens, raising real bills 12 to 27 percent after caching.
- DeepSeek V4: Pricing, Context, and MigrationDeepSeek V4 ships two models: V4-Flash at $0.28/M output and V4-Pro at $3.48/M. Both carry a genuine 1M context window and drop into any Anthropic-compatible SDK with one line changed.
- Claude Code Quality Regression: What Actually HappenedThree product-layer changes broke Claude Code for six weeks in early 2026. The post-mortem, the AMD data, and what it means if you build on AI coding agents.
- Claude Opus 4.7 vs GPT-5.5GPT-5.5 launched April 23, 2026. Here is how it stacks against Claude Opus 4.7 on coding, agents, long context, and cost, plus which one to actually use.
- Claude Mythos: The Model That Thinks in LoopsClaude Mythos is suspected to use recurrent-depth architecture: one shared layer looped N times, with ACT halting so hard questions get more passes and easy ones stop early.
- Claude Opus 4.7 vs Other AI ModelsClaude Opus 4.7, GPT-5.4, Kimi K2.6, Gemini 3.1 Pro, DeepSeek V3.2: benchmarks, context windows, agent reliability, and cost, so you reach for the right one.
- Claude Opus 4.7 Use CasesClaude Opus 4.7 use cases across multi-file coding, security review, legal, finance, document reasoning, multimodal review, long-running Claude Code agents.
- Claude Opus 4.7Claude Opus 4.7, Anthropic's April 2026 flagship: stronger hard coding, document reasoning, long-running agent tasks, same $5/$25 pricing as Opus 4.6.
- Claude Opus 4.6Claude Opus 4.6 ships February 2026 with 1M context GA, 128K max output, and the same $5/$25 pricing. Sharper planning, longer agent runs, big-codebase gains.
- Claude Code ModelsPick the right Claude Code model: Sonnet, Opus, Haiku, sonnet[1m], or opusplan. Switching per task cuts model costs 60-80% without sacrificing output quality.
- Claude 4Claude 4 launched May 2025 with Sonnet 4 and Opus 4, pushing Claude Code mainstream. 200K context, 16K output, agentic reliability, extended thinking baked in.
- Claude Haiku 4.5Claude Haiku 4.5 launched October 2025 as budget tier for Claude Code smart model switching. 200K context, sub-2s latency, routes routine work cheaply.
- Claude Opus 4.5 in Claude CodeSet Claude Opus 4.5 as your Claude Code default in two commands. 76% fewer output tokens than Sonnet 4.5, 50% fewer tool calls, $5/$25, same 200K window.
- Claude 3.7 SonnetClaude 3.7 Sonnet shipped February 2025 with hybrid reasoning and extended thinking. 64K output, thinking-budget control, SWE-bench coding gains at $3/$15.
- Claude Opus 4.1Claude Opus 4.1 shipped August 2025 as a reliability patch on Opus 4. Steadier long-session output, fewer regressions, safer error handling, 200K context.
- Claude 3Claude 3 launched March 2024 with Opus, Sonnet, Haiku tiers, vision input, 200K context windows. Specs, benchmark scores, pricing, and the three-tier template.
- Claude 3.5 SonnetClaude 3.5 Sonnet launched June 2024 at $3/$15, beating Claude 3 Opus on MMLU, GPQA, HumanEval at a fifth of the cost. Specs, benchmarks, and code gains.
- Claude Sonnet 4.6Claude Sonnet 4.6 ships February 2026 at $3/$15 with 1M context GA. 70% developer preference over Sonnet 4.5, beats Opus 4.5 on 59% of coding sessions.
- Every Claude ModelEvery Claude model on one page: Claude 3, 3.5, 3.7, 4, Opus 4.1 to 4.6, Sonnet 4.5 and 4.6, Haiku 4.5. Specs, pricing, benchmarks, and when to use each.
- Claude Opus 4.5Claude Opus 4.5 launched November 2025 at $5/$25 per million tokens, two-thirds off Opus 4.1. 80.9% SWE-bench, 1M context beta, tighter sub-agent delegation.
- Claude 3.5 Sonnet v2 and Claude 3.5 HaikuClaude 3.5 Sonnet v2 and 3.5 Haiku launched October 2024 with Computer Use beta, cursor control, upgraded coding and tool use, and cheaper Haiku at $0.80/$4.
- Claude Sonnet 4.5Claude Sonnet 4.5 launched September 2025 at $3/$15, closing the gap with Opus 4.1 on coding and agents. 200K standard context, 1M beta, 16K max output tokens.
Real Builds
Concrete systems, experiments, and shipped agent workflows.
Core Pages
- The AI Agent That Deleted a Production Database in 9 SecondsAn AI deleted PocketOS's production database and all backups in 9 seconds. Here is why it happened and the guardrails that prevent it.
- How to Build Your Own Claude Code Harness (or Buy One)How to build a Claude Code harness from scratch, what it costs in time, and when buying a pre-built kit ships faster.
- Is Claude Code Just a Thin Wrapper? Inside the Harness DebateIs Claude Code a thin wrapper around the Claude API? No. A leaked 512,000-line source shows a full orchestration harness. Here is what it adds.
- Harness vs Boilerplate vs Framework: The Build-System Stack ExplainedHarness vs boilerplate vs framework, explained plainly: what each layer does, who needs which, and where the $29 Code Kit fits.
- What Is an Agent Harness? Why the Harness, Not the Model, Is the 2026 MoatAn agent harness is everything around the model: tools, memory, permissions, and the control loop. Here is why the harness is the real 2026 moat.
- Is Vibe Coding Safe? What the Lovable and Moltbook Breaches TeachIs vibe coding safe? Not by default. The Moltbook and Lovable breaches happened for two reasons, and here is how to check and fix your app.
- Spec-Driven Development Explained: Why Pros Stopped Vibe CodingSpec-driven development means writing structured Markdown specs before AI writes code. Here is how GitHub Spec Kit and Kiro made it the 2026 default.
- State of Vibe-Coded SaaS Security (2026 Data)Vibe coding security statistics from 2026 scans: 5,600 apps, 2,000+ vulnerabilities, 45% OWASP failure rate, and the Moltbook breach.
- From Vibe Coding to Production: The Checklist That Stops Data LeaksVibe coding to production safely: the 25-point checklist covering RLS, secrets, ownership checks, and rate limits that stops AI-built data leaks.
- Vibe Coding vs Vibe Engineering vs Agentic Engineering: The 2026 GlossaryVibe engineering, vibe coding, and agentic engineering defined plainly, with origins, a comparison table, and when to use each in 2026.
- Do I Still Need a Boilerplate If I Use Claude Code?Is a Claude Code boilerplate needed? Short answer: it depends on what you ship. Here is the honest decision tree and cost math.
- How Much Does It Really Cost to Build a SaaS with Claude Code?The real cost to build a SaaS with Claude Code: $49 in month one, $100 to $375/month at scale. Full breakdown of plans, infra, and the kit.
- Run Claude Code on a Cheaper Model: DeepSeek and GLM Cost ArbitragePoint Claude Code at DeepSeek or GLM to cut your bill 7 to 17x. Setup, what breaks, and the July 2026 model-name change explained.
- How to Cut Your Claude Code Token Bill in HalfReduce Claude Code token cost with five fixes: cap thinking tokens, default to Sonnet, defer MCP loading, trim CLAUDE.md, and filter tool output.
- How Long Does Idea to Production Actually Take with Claude Code?How long to build a SaaS with AI: a prototype takes hours, a shippable MVP 1-4 weeks, a production-grade SaaS 4-8 weeks part-time.
- State of Claude Code 2026: What 2,500 Public Repos RevealWe analyzed 2,500 public repositories that use Claude Code. 85% ship a CLAUDE.md, but only 25% define a custom subagent. Here is how developers actually configure Claude Code in 2026: agents, skills, hooks, MCP, and how big CLAUDE.md files have grown.
- Own Your Vercel Analytics: I Built a Drain-to-Postgres PipelineVercel has no API to query Web Analytics history. The only export path is a Drain. Here is the HMAC-signed receiver, the Postgres rollup, and the one-session build that owns the data.
- The Autonomy Curve: How Much Freedom Can You Give an AI Agent?How much autonomy you can give an AI agent is decided by one thing: how long a model holds a task without drifting. A good harness plus a reliable model is what unlocks real agent work.
- Building Isn't the Bottleneck AnymoreAfter shipping AI-built SaaS from Claude 3.5 Sonnet to Fable 5, the hard part stopped being code. It moved to QA at scale and distribution. Here is what actually changed, and why fundamentals matter more now, not less.
- Distribution Is the New MoatWhen AI makes building a weekend job, the moat moves to distribution. Building is commoditized, getting found is the whole game, and the smart move is an ecosystem.
- First Principles in the Age of 24-Hour MVPsWhen AI lets you build anything in a day, the build stops deciding who wins. Focus, first principles, and speed to product-market fit decide it instead.
- Why QA Is the Real Bottleneck in AI DevelopmentThe hardest unsolved problem in AI software development is not generating features. It is verifying them at scale. QA does not parallelize like generation does.
- A Security Update from Build This NowA customer flagged something suspicious. We investigated, found a security issue in a file we ship, fixed it the same day. Here is what happened and what to do.
- Claude Code for Freelancers: Ship 3x FasterHow freelance developers use Claude Code to cut implementation time, handle more clients, and increase effective hourly rate without working more hours.
- Claude Code for Non-Developers: Real ExamplesA Dutch B.V. owner automated two years of tax filing. A 63-year-old ex-CEO shipped a SaaS serving thousands of users. What non-developers are actually building with Claude Code.
- Build a Full App with Claude Code: Real ExamplesA zero-game-dev builder shipped a GTA clone on Google Earth in a weekend. A job search system evaluated 740 listings with no app code. What actually works when building full apps with Claude Code.
- Agent Swarm OrchestrationFour infrastructure layers that stop agent swarms from double-claiming tasks, drifting on field names, and collapsing under merge chaos.
- AI Cleans ItselfThree overnight Claude Code workflows that clean AI's own mess: slop-cleaner removes dead code, /heal repairs broken branches, /drift catches pattern drift.
- Autonomous AI SwarmAn autonomous Claude Code swarm: a 30-min trigger, an orchestrator, specialist sub-agents in worktrees, and five gates that ship overnight features safely.
- AI Email SequencesOne Claude Code command builds 17 lifecycle emails across 6 sequences, wires Inngest behavioral triggers, and ships a branching email funnel ready to deploy.
- AI Security AgentsTwo Claude Code commands spin up eight security sub-agents: phase 1 scans SaaS logic for RLS gaps and auth bugs, phase 2 penetrates to confirm real exploits.
- Idea to SaaSPlain-English walkthrough of the Build This Now pipeline: market discovery, auto-planning, a staged build, and post-launch commands that keep your SaaS live.
- GAN LoopOne agent generates, one tears it apart, they loop until the score stops improving. GAN Loop implementation with agent definitions and rubric templates.
- Distribution AgentsFour Claude Code agents that run on a schedule, write SEO posts, read PostHog, build carousels, and scout Reddit. Copy the definitions and plug them in.
- Self-Evolving HooksThree hooks turn every 'no, not like that' correction into a skill or rule Claude reads on the next session. Self-improving agents, no prompt tuning needed.
- Trace to SkillTrace2Skill: run an agent 20 times, log what worked, let four Claude analysts read the traces, and merge them into a SKILL.md that beats hand-written ones.
For Business
AI outcomes for companies: sales, ops, lead gen, and automation.
Core Pages
- The 30-Day AI Pilot That Ships95% of AI pilots deliver no profit. The fix isn't a better model. It's a 30-day pilot scoped to one bottleneck with an output you can test the next morning.
- AI for the CFOThe four finance numbers a CFO should automate first, ranked by payback: month-end close, collections and DSO, forecast prep, and board reporting.
- AI for Agencies: The 3 Billable-Hour LeaksFirms that sell time leak money in three specific places: non-billable drift, silent scope creep, and proposals rebuilt from scratch. Where each one hides and the fix.
- The Cash-Conversion-Cycle LeakYour profit and loss statement says you made money. Your bank says otherwise. The cash conversion cycle for a services business shows where the money is stuck.
- The 30-Day Window After a RaiseA funding round is a dated buying signal with a short shelf life. When the window opens, why it closes, and the one thing to send instead of a congrats email.
- A New VP Was Hired. You Have 90 Days.A newly hired VP re-tools in their first 90 days. The buying window, when to send, and the value asset that earns the reply. Not another congrats email.
- Automate RFP ResponsesThe same 200 questions, retyped by your best people at 11pm. Build an answer library instead, and RFP and security-questionnaire responses assemble themselves.
- Your Best Customers Are Quietly LeavingThe customer who cancels next quarter is already going quiet this quarter. The five signals that predict churn, and how to catch the leak before renewal.
- The Competitive Benchmark ScorecardA competitive benchmark scorecard is a standing, scored view of where you stand across the field. What goes in it, and how to keep it from going stale.
- Your CRM Is a Filing CabinetYour CRM records what happened. It never tells you a lead has gone quiet for nine days. Why data decay and low adoption turn it into a filing cabinet.
- How to Reopen Dead DealsMost closed-lost B2B deals were never a no, just a not-yet. 40-60% die to no decision, not a competitor. How to find the ones worth reopening.
- The Self-Producing Board DeckThe board deck and QBR scramble is copy-paste, not analysis. Here is how to automate the assembly of recurring reports and keep the judgment human.
- Battlecards Are DeadA battlecard, the one-page cheat sheet about a competitor, is outdated before your salespeople even open it. Build fresh, deal-specific competitor briefs for proposals and board meetings instead. Here's how.
- Buying Signals vs. Intent DataIntent data, guesses about who might be researching you, is overrated. Public signals, like a new hire, a funding round, or a new office, are real events you can verify. Here's how fast you need to act on each one.
- Done-For-You vs SaaS vs ConsultantsYou aren't choosing a vendor. You're choosing how to get a result: a deck, a login, a hire, or an installed outcome. Each one's worst failure mode, scored.
- Reps Only Sell 40% of the TimeSalespeople spend under 40% of the week actually selling. Automate the repetitive admin work, keep the human relationship, and use a simple test to decide what to automate first.
- I Can Build It. Should I?You can build the internal AI tool yourself. The real cost isn't the build, it's what happens afterward as things change. A true-cost calculator that includes the line founders always forget.
- The 5-Minute Speed-to-Lead RuleEvery sales leader knows the 5-minute rule. Almost nobody hits it. The real numbers on how fast companies reply to a new inquiry, and why speed is a detection problem, not a willpower problem.
- Stop Building Internal ToolsShould you build internal AI tools yourself or buy the finished result? A minimal in-house team costs $400β600K a year to run, and about 70% of these projects never ship. When to build, when to buy the outcome, and the real hidden cost.
- The Real Cost of Slow Follow-UpSlow follow-up isn't a sales problem, it's a money problem. The math on every lead you paid for and never contacted, and how to fix it.
- The One Constraint That Caps GrowthOne step caps how fast your whole business makes money. The Theory of Constraints, the idea that one bottleneck limits everything, applied to a services, agency, or software firm, and why the bottleneck is usually a rule, not a lack of people or tools.
- What a 1,000-Agent Workflow CostsThe demo shows a thousand AI agents doing a department's worth of work. It never shows the bill. Here are the three costs of running a large AI agent system, and the one that actually costs the most.
- Where Is My Business Losing Money?Most owners can name their biggest expense but not the one process quietly draining their margin. A revenue-leak audit that finds the single leak worth fixing first.
- Win-Loss AnalysisWin-loss analysis means asking your buyers why you won or lost each deal. It's free competitive intel that your customer records usually get wrong. The interview method, the 6 questions, and why 85% of stated loss reasons are wrong.
- AI Agents vs EmployeesThe honest 2026 cost of AI agents vs employees: the real all-in cost of a hire, where AI cuts costs by 85%, where AI agents end up costing more than staff, and why 95% of pilots fail.
- AI Lead Generation From SignalsHow AI turns a plain-English description of a buying signal into a ranked list of deals ready for your approval in 2026: what it can track, what it's worth, and why building it yourself quietly fails.
- Claude for Sales & FinanceWhat Claude actually does for sales and finance teams in 2026: the repetitive work it takes over, the decisions people still make, and why building this yourself usually stalls.
- What Are Dynamic AI Workflows?Dynamic AI workflows are AI systems that can run a whole business function, not just one task. What they are, why they beat old-style automation, and why 40% of projects still get canceled.
- Why Most AI Automation FailsWhy AI automation projects fail in 2026: the failure numbers from MIT, Gartner, RAND and McKinsey, and the ongoing costs nobody budgets for.