The Build Log

Read them all

AI Agent Orchestration: The 3-Lane Chain I Run Daily

AI agent orchestration that holds up in production: the plan, execute and check lanes, why the check lane must be able to veto, and where it stops paying.

23 Sept 20269 min readai-agent-orchestrationmulti-agentqa-gatesubagentsaccountability

Read it →

Best AI Coding Assistant 2026: Routed, Not Ranked

The best AI coding assistant in 2026, routed by task from daily production use: Claude Code, Codex, Gemini CLI and Cursor, plus dated pricing for each.

23 Sept 202613 min readai-coding-assistantclaude-codecodexcursorgemini-climodel-routing

Read it →

Claude Code Hooks: The Rules an Agent Can't Talk Past

Claude Code hooks explained: the lifecycle events you can hook, how to wire one up, and the four gates that block a bad agent commit.

23 Sept 20269 min readclaude-codehookslifecycle-eventspre-commitguardrails

Read it →

Claude Code MCP Integrations After the 2026 Protocol Break

Claude Code MCP integrations explained: connecting your first server, scoping what it can touch, which integrations earn their place, and how to debug.

23 Sept 20269 min readclaude-codemcpmodel-context-protocolintegrationsstateless-protocol

Read it →

Claude Code Usage Limit: The Two Meters Nobody Explains

Hit the Claude Code usage limit? What the 5 hour window really is, what happens when you hit it, and the caching habits that stop it recurring.

23 Sept 202612 min readclaude-codeusage-limitrate-limitprompt-cachingmax-plan

Read it →

Claude Code Memory: What Survives a Session Restart

Does Claude Code keep memory between sessions? Where it writes, what the memory skill does, and the write-on-pause rule that makes it survive.

23 Sept 20269 min readclaude-codeagent-memorymemory-mdsession-continuityhindsight

Read it →

Claude Code Subagents: The Brief, the Wave, the Handback

Claude Code subagents in practice: when to spawn one, how to run several in parallel, what to name them, and the jobs they are the wrong tool for.

23 Sept 20269 min readclaude-codesubagentsparallel-agentsorchestrationagent-naming

Read it →

Claude Code vs Cursor: Which Task Goes Where

Claude Code vs Cursor from a terminal-first production fleet: which task goes where, whether they work together, and what running both actually costs.

23 Sept 20269 min readclaude-codecursorai-coding-assistantagentic-codingpricing

Read it →

Context Engineering AI: How I Run It in Production

Context engineering explained from production use: the layers, how it differs from prompt engineering, and how retrieval, memory and caching fit together.

23 Sept 20269 min readcontext-engineeringprompt-cachingretrievalagent-memoryclaude-code

Read it →

DeepSeek Local Setup: Running V4 Inside an Agent Harness

Running DeepSeek locally and wiring it into an agent harness: offline use, self-hosting, the safety questions, and when a local model earns its slot.

23 Sept 20269 min readdeepseekdeepseek-v4local-llmagent-harnessquantization

Read it →

Grok AI Review 2026: The 200K Price Cliff Agents Hit

A Grok review from someone running a production agent fleet: is it good, how it compares to ChatGPT, what it costs, and where it would fit.

23 Sept 202610 min readgrokgrok-4-6xaigrok-5api-pricing

Read it →

MCP Server Setup in Hermes: The Key Everyone Misses

Setting up an MCP server and registering it in a Hermes runtime, including the config key that makes people think Hermes has no MCP support at all.

23 Sept 202612 min readmcpmodel-context-protocolhermesmcp-server-setupagent-runtime

Read it →

Claude Skills: How to Build One That Survives Production

How to build a Claude skill that survives production: the SKILL.md anatomy, the build standard every skill has to pass, and what to build first.

23 Sept 202614 min readclaude-skillsskill-mdclaude-codeagent-toolingbuild-standard

Read it →

Hermes AI Agent Reliability: The Failures I Actually Hit

Why AI agents fail, and how to make a Hermes agent reliable: the four common failure modes, runtime drift repair, and proof before you trust output.

23 Sept 202610 min readhermesagent-reliabilityruntime-driftfailure-modesopenclaw

Read it →

How to Use Claude Code: The Workflow My Agent Fleet Runs

How to use Claude Code the way a production agent fleet runs it: install, first real task, free tier limits, and the workflow that scales past day one.

23 Sept 202613 min readclaude-codeclaude-code-tutorialagentic-codinggit-worktreegetting-started

Read it →

Free ChatGPT Alternative: What Open Source Really Costs

Open source AI alternatives to paid subscriptions: what is genuinely free, what the free option quietly costs instead, and how to decide build vs pay.

23 Sept 20269 min readfree-chatgpt-alternativeopen-source-aideepseekself-hosted-aisubscription-cost

Read it →

What GPT-6 Astra Can Do, and How I'd Test It

What GPT-6 Astra can do, why it shipped restricted, and the three jobs it would have to win before it gets a lane in my agent fleet.

23 Sept 20268 min readgpt-6gpt-6-astraopenaichatgptmodel-evaluation

Read it →

Why Token Cost Control Is Harder Than It Seems

The honest tension between flat-rate max plans that hide token cost, and the caching discipline an agent fleet still needs to stay efficient.

2 Sept 20264 min readtoken-cost-controlprompt-cachingagent-fleetmax-plan

Read it →

The four gates that stop a bad agent commit before it lands

My enforced-template repo runs a plan gate, a tech stack enforcer, a spec gate, and an SEO validator, built on a plan-first principle I wrote down and kept.

26 Aug 20264 min readclaude-codehooksenforcementcode-review

Read it →

Why I Track Map Pack Rank on a Grid, Not a Single Point

A single rank check hides how a business performs across its service area. Inside the geo grid rank tracker I built on DataForSEO and Google Maps.

25 Aug 20264 min readgmbgeogridmap-packrank-trackinggmb-tools

Read it →

Rebranding a Template From One Config File

Inside the brand.md to rippr-engine pipeline that turns one config file into N branded sites across Next.js, static HTML, and WordPress.

25 Aug 20264 min readripprrippr-enginebrand-jsontemplate-factory

Read it →

What I Actually Teach About Token Spend

The real curriculum I run on prompt caching and context pruning, the cache mechanics I verify instead of assume, and why caching is the lever that cuts spend.

18 Aug 20264 min readtoken-cost-controlprompt-cachingclaudetraining

Read it →

How Client Work Actually Moves Through Our Agency

Discovery, Build, Review, Handoff, with a checkpoint at every stage. Why the Review step, backed by a real tracker cadence, keeps work from failing.

18 Aug 20264 min readclient-deliveryreview-cadenceagency-operationsproof-of-work

Read it →

The GMB Suspension Recovery Playbook I Actually Run

A suspended Google Business Profile needs its own process, not resubmit-and-hope. How I sequence recovery using the GMB Sniper checklist and a recovery skill.

11 Aug 20264 min readgmbsuspension-recoverygmb-sniper-optimizationreview-responseplaybook

Read it →

Routing Work to the Right Lead Agent

Inside the Oliver to Carlos to lead to Queen chain that keeps 40-plus named agents accountable instead of grabbing whoever is free.

4 Aug 20264 min readagent-fleetorchestrationqa-gateagent-routing

Read it →

Running Five Coding Agents on One Repo, No Collisions

A shipped template built on IndyDevDan's Infinite Agentic Loop runs parallel agent waves, and a standing routing rule keeps that fan-out from becoming chaos.

4 Aug 20264 min readclaude-codeworktreesmulti-agentparallel-agents

Read it →

Why Spielberg Owns Video and Routes Avatars to Creatify

A done-for-you avatar video only scales with one named lead and one fixed default tool. Spielberg's routing rule and Creatify skill make it repeatable.

3 Aug 20264 min readcreatifyspielbergdfy-videoavatar-videovideo-pipeline

Read it →

The Ranking Signals You Can Actually Move, Pillar by Pillar

Local SEO is not one lever, it is five. The five pillars in The Merlino Spell, the tools built to track each one, and why one score hides the problem.

28 Jul 20264 min readlocal-seothe-merlino-spellranking-signalsgmb-toolsgreen-grid-goblins

Read it →

How I Decide What to Build vs. What to Fork

What breaks when your agent fleet ships in public: skipping the registry, skipping eval-before-you-build, and losing control of who narrates your stack.

28 Jul 20264 min readbuild-in-publicregistryopen-sourcedecision-making

Read it →

Inside the Avatar and Lipsync Pipeline We Actually Run

Script to finished clip runs through a versioned chain: avatar-v3, lipsyncs-v2, Creatify, ElevenLabs, Remotion, ffmpeg. The version numbers are proof it works.

21 Jul 20264 min readcreatifyelevenlabsremotionlipsyncavatar-pipelineranking-reels

Read it →

Why my agents never forget what they did last session

Every agent keeps a MEMORY.md pushed on every /wrap. I even ran a pilot backfilling one agent's memory from old Claude conversations so it never started cold.

15 Jul 20264 min readagent-memorysession-persistencewrap-pause-resumehindsight

Read it →

How My Agents Run a Local SEO Audit From My Own Framework

The Merlino Spell is a five-pillar framework I wrote myself. Here is how agents run real audits from it, and the infrastructure that supports that work.

14 Jul 20265 min readlocal-seothe-merlino-spellauditsagent-fleetgmb

Read it →

Naming Every Agent a Role, Not a Tool

Why my fleet has 40-plus named specialists instead of a folder of scripts, and why semantic naming is doctrine, not branding.

14 Jul 20264 min readagent-fleetagent-namingorchestrationfleet-doctrine

Read it →

The Dashboards I Actually Run merlinoai.com On

A real walkthrough of the live dashboard family under merlinoai.com, the public tools that skip the login wall, and why I ship to a real URL, not a demo.

14 Jul 20264 min readbuild-in-publicdashboardsmerlinoaiagent-fleet

Read it →

Running an AI-Native Agency, Day to Day

A fleet of specialist agents only works if the humans running it stay more disciplined than the tools chasing their attention. Why I keep my team on one stack.

2 Jul 20264 min readagent-fleettooling-disciplineagency-operationshiring-economics

Read it →

The memory store every agent in my fleet writes to

I picked Hindsight plus Master Brain after testing and killing an alternative, built it so every agent's memory injects automatically, and proved it works.

24 Jun 20264 min readagent-memorymulti-agenthindsightmaster-brain

Read it →

Cloning a Site Design Into a Template

How Rippr turned a one-off site clone into a banked, rebrandable template, and why aurora-ai-saas is the proof it holds up across three stacks.

20 Jun 20264 min readripprsite-cloningtemplate-factoryaurora-ai-saas

Read it →