# Ready Solutions AI

> Claude-first AI integration consulting. Advisory sessions and premium
> implementation for technical teams.

## About

- [About the Founder](https://readysolutions.ai/about/): Mitchel Lairscey. 20 years in software engineering, 3.5 years of LLM work, 18 months deploying Claude daily. Independent consultant, sole practitioner.

## Services

- [All Services](https://readysolutions.ai/services/): Two tiers. Advisory consultations and workshops for AI strategy. Premium implementation and build work for Claude Code, MCP servers, and agentic workflows.

## Tools

- [AI Readiness Assessment](https://readysolutions.ai/assess/): Free 10-minute AI-powered assessment scoring organizations across five dimensions with a personalized action plan.

## Guides

- [Guides](https://readysolutions.ai/guides/): Production references for running Claude Code, MCP servers, hooks, skills, Claude API patterns, reliability, and governance.
- [Running Claude Code as a Production Engineering Practice](https://readysolutions.ai/guides/running-claude-code-as-a-production-engineering-practice/): The core operating model for disciplined Claude Code use.
- [MCP Servers in Production](https://readysolutions.ai/guides/mcp-servers-in-production/): How to run MCP servers with tool access, auth boundaries, and maintenance in mind.
- [Claude Code Skills in Production](https://readysolutions.ai/guides/claude-code-skills-in-production/): How skills encode team standards and reusable Claude Code workflows.
- [Refusal Handling and Model Fallback in Production](https://readysolutions.ai/guides/refusal-handling-and-model-fallback-in-production/): Detecting stop_reason refusal, choosing a fallback path per platform, and the observability contract for Claude Fable 5 classifier refusals.

## Glossary

- [Agentic Development Glossary](https://readysolutions.ai/glossary/): Topic-led definitions for Claude Code, MCP, context engineering, agent reliability, governance, and AI adoption.
- [MCP server](https://readysolutions.ai/glossary/mcp-server/): The external-tool integration boundary for Model Context Protocol.
- [Claude Code hook](https://readysolutions.ai/glossary/claude-code-hook/): Runtime gates that fire before or after Claude Code actions.
- [Subagent orchestration](https://readysolutions.ai/glossary/subagent-orchestration/): Coordinating separate agents around bounded work.

## Engineering methodology

- [Engineering methodology](https://readysolutions.ai/engineering/): Machine-oriented methodology page for how Ready Solutions AI structures validators, provenance, and authority surfaces.

## Blog

- [Why Claude Code Changes How SaaS Teams Ship](https://readysolutions.ai/blog/2026-03-28-why-claude-code-changes-how-saas-teams-ship/): Claude Code workflows in CI/CD, code review, and sprint planning for engineering teams.
- [Connecting Claude to Jira, GitHub, and Confluence via MCP](https://readysolutions.ai/blog/2026-03-29-connecting-claude-jira-github-confluence-mcp/): Step-by-step MCP server setup guide for the Atlassian and GitHub stack.
- [Claude Code Custom Skills: Encoding Team Standards](https://readysolutions.ai/blog/2026-03-29-claude-code-custom-skills-team-standards/): Turning team coding standards into AI-enforced instructions via custom skills.
- [Your PRs Are Missing Context. Claude Code Fixes That.](https://readysolutions.ai/blog/2026-03-29-your-prs-are-missing-context-claude-code-fixes-that/): AI-generated PR descriptions from codebase and commit history.
- [The State of AI-Assisted Development in 2026](https://readysolutions.ai/blog/2026-03-29-state-of-ai-assisted-development-2026/): Data-driven analysis of AI tool adoption and developer productivity.
- [Running Post-Mortems on Disaster Scenarios with AI](https://readysolutions.ai/blog/2026-03-29-ai-disaster-postmortems/): Using Claude to reconstruct timelines, detect patterns, and improve incident response.
- [AI Is Coming for Your Research Business](https://readysolutions.ai/blog/2026-03-31-ai-is-coming-for-your-research-business/): Strategy for independent researchers to use AI as a growth tool.
- [Is AI Worth It for Your Small Business?](https://readysolutions.ai/blog/2026-04-01-is-ai-worth-it-for-your-small-business/): ROI analysis for small business AI adoption with savings benchmarks.
- [What Should a Solopreneur Automate With AI First?](https://readysolutions.ai/blog/2026-04-01-what-should-a-solopreneur-automate-with-ai-first/): Prioritization framework for solo business AI automation.
- [The Three-Question Filter for AI Tool Decisions](https://readysolutions.ai/blog/2026-04-01-three-question-filter-ai-tool-decisions/): Framework to evaluate AI tools and cut through vendor noise.
- [Turn Research Reports into Social Content with AI](https://readysolutions.ai/blog/2026-04-01-ai-social-content-market-researchers/): AI workflow for repurposing research into LinkedIn posts and data visuals.
- [AI Is Not Taking Your Job. It's Rewriting the Job Description.](https://readysolutions.ai/blog/2026-04-05-ai-entry-level-jobs-new-graduates/): How AI reshapes entry-level roles and strategies for new graduates.
- [The Agentic Development Starter Guide](https://readysolutions.ai/blog/2026-04-08-agentic-development-starter-guide/): Plan-audit-implement-verify cycle for disciplined AI-assisted coding.
- [AI in the SDLC for SaaS LMS Teams](https://readysolutions.ai/blog/2026-04-08-ai-sdlc-saas-lms/): Phase-by-phase guide to AI-driven SDLC changes with ROI data for edtech.
- [Your Claude API Integration Is Probably a Wrapper Around a String Function](https://readysolutions.ai/blog/2026-04-09-claude-api-integration-mistakes/): Common Claude API integration mistakes that cost money, add latency, and leave the strongest features unused.
- [The AI Compliance Stack: Five Layers Every Organization Faces in 2026](https://readysolutions.ai/blog/2026-04-09-ai-compliance-2026-by-industry/): Visual guide to AI compliance across healthcare, fintech, and government with deadlines, penalties, and a governance checklist.
- [Your Team Bought AI Licenses Three Months Ago. Why Aren't They Using Them?](https://readysolutions.ai/blog/2026-04-10-ai-adoption-stalls-workflow-not-training/): Why AI adoption stalls after license purchase and why workflow redesign beats training as the fix.
- [Beyond the Wrapper: Five Claude API Patterns That Separate Prototypes from Production](https://readysolutions.ai/blog/2026-04-15-claude-api-production-patterns/): Implementation guide for tool use, prompt caching, streaming, extended thinking, and structured outputs in production Claude API integrations.
- [The Engineering Manager's Guide to Governing Agentic Development](https://readysolutions.ai/blog/2026-04-15-engineering-managers-guide-agentic-development-governance/): Three-tier governance framework for standardizing AI development workflows without micromanaging engineers.
- [What 20-Turn Conversations Taught Me About the Claude API](https://readysolutions.ai/blog/2026-04-16-claude-api-advanced-production-patterns/): Nine production-tested Claude API patterns from building a real multi-turn assessment agent on Cloudflare Workers.
- [Claude Opus 4.7 Is a Split, Not an Upgrade](https://readysolutions.ai/blog/2026-04-16-claude-opus-4-7-is-a-split-not-an-upgrade/): Predictive take on the Opus 4.7 release as a deliberate product split between long-horizon agent operators and fast-turn assistants, and what that means for Opus-vs-Sonnet routing.
- [Stop Asking Which AI Tool to Buy. Ask This Instead.](https://readysolutions.ai/blog/2026-04-16-stop-asking-which-ai-tool-to-buy/): Contrarian quick-format argument that the procurement-first framing of AI adoption causes the shelfware and pilot-failure outcomes leaders are trying to avoid. The better first question is workflow friction.
- [Claude Opus 4.7: Your Eval Harness Can't See What Just Changed](https://readysolutions.ai/blog/2026-04-17-claude-opus-4-7-deep-dive/): Deep-dive operator's guide arguing that Opus 4.7's real upgrade lives in four dimensions that single-turn evals can't measure (trace length, tool-call density, instruction literalism, checkpoint integrity), with a harness rebuild methodology and a concrete Opus-vs-Sonnet routing framework.
- [The Four Sub-Agent Orchestration Patterns That Cover 90% of Production Claude Workloads](https://readysolutions.ai/blog/2026-04-18-sub-agent-orchestration-patterns-claude/): Deep-dive synthesis naming four observable patterns (parallel fan-out, sequential review chains, adversarial dual-analysis, hierarchical planner-executor) with Anthropic-sourced cost math, latency tradeoffs, and failure modes. Fills the gap where Anthropic publishes primitives but no canonical reference architecture.
- [Claude vs Claude Code: How to Pick the Right One (and When to Use Both)](https://readysolutions.ai/blog/2026-04-20-claude-vs-claude-code-when-to-use-each/): Orientation deep-dive comparing Claude.ai and Claude Code capability surfaces, with a decision tree and a three-stage pairing pattern (design in Claude.ai, plan in Claude Code Plan Mode or Superpowers, execute in Claude Code).
- [Opus 4.7 vs 4.6 in Claude Code: Where the Upgrade Shows Up, and When xhigh Is Worth the Tokens](https://readysolutions.ai/blog/2026-04-24-opus-4-7-claude-code-effort-ladder/): Practitioner field guide on the five-rung effort ladder (low/medium/high/xhigh/max) in Claude Code on Opus 4.7. Argues for stepping down from Claude Code's new xhigh default to high for most agentic work, with tokenizer inflation as the steel-man.
- [Claude Design: The Handoff Is the Feature](https://readysolutions.ai/blog/2026-04-24-claude-design-handoff-not-canvas/): Contrarian deep-dive on Anthropic's April 2026 Claude Design launch. Argues the real leverage isn't designs but the design-to-code handoff bundle that transfers prototypes to Claude Code with one instruction, because producer and consumer share a model family for the first time.
- [Where Does That Rule Go? A Decision Tree for CLAUDE.md, Settings, Skills, and Hooks in Claude Code](https://readysolutions.ai/blog/2026-04-26-claude-code-rule-routing-decision-tree/): Practitioner-focused deep-dive routing Claude Code directives across the five customization layers (CLAUDE.md, settings rules, path-scoped rules, skills, hooks) with a decision tree, five worked examples, and a steel-man on the limits of hook enforcement.
- [The free Claude Code skill that audits your CLAUDE.md, hooks, and subagents for Opus 4.7 breaking changes](https://readysolutions.ai/blog/2026-04-26-opus-4-7-compatibility-scanner-claude-code/): Free downloadable Claude Code skill (opus-compatibility-scanner) auditing CLAUDE.md, AGENTS.md, settings.json, hooks, and SDK call sites against 70 Opus 4.6 / 4.7 compatibility patterns. Catches the configuration debt Anthropic's migration guide does not cover.
- [How to Create a Claude or Claude Code Skill: A Hands-On Guide for First-Timers and Optimizers](https://readysolutions.ai/blog/2026-04-29-how-to-create-claude-or-claude-code-skill-hands-on-guide/): Demystifies Claude Skills creation across all five Claude surfaces (claude.ai, Desktop, Code, API, Agent SDK) plus portability to Codex, Cursor, Windsurf. Three creation paths (vanilla Claude, Anthropic /skill-creator, Superpowers /writing-skills), an author-original RED-DRAFT-GREEN-REFACTOR-OPTIMIZE-SHIP workflow, manual SKILL.md frontmatter reference, and an honest steel-man on when a skill is the wrong layer.
- [The AI Coding Spectrum: Vibe Coding, Agentic Development, and Who Owns the Verification Loop](https://readysolutions.ai/blog/2026-04-30-vibe-coding-vs-agentic-development-ai-coding/): Synthesis-with-contrarian-edge framing of vibe coding and true agentic development as positions on the AI Coding discipline spectrum, with the verification-loop ownership test as the diagnostic. Backed by 2025-2026 production data (Veracode 45% OWASP fail, METR 19% slowdown RCT, Opsera 4.6x review wait, GitClear 4x clone growth, arXiv 2510.00328 grey-lit QA breakdown), Anthropic vendor docs on verification primacy, the Backslash three-tier spectrum, and a 2-vs-24-person-week experiential anchor from a solo agentic delivery on an unfamiliar codebase.
- [90 Minutes With 400 Engineers, PMs, and Ops on Claude Code: Here's What They Wanted to Know](https://readysolutions.ai/blog/2026-05-02-six-questions-400-employees-asked-about-claude-code/): Field-observation deep-dive on the six questions that recurred across a 90-minute Claude Code training to 400+ employees at an enterprise SaaS org. Diagnoses the dominant rollout-failure pattern as a flat "one tool, one config" mental model that conflates CLAUDE.md, skills, hooks, plugins, and the marketplace. Each question gets the diagnostic answer with sourced Anthropic docs, plus a closing five-question triage the post argues should replace the "go read the docs" reflex.
- [Half Your Team Is on Opus 4.6, Half Is on 4.7. The Problem Isn't the Model.](https://readysolutions.ai/blog/2026-05-02-engineering-manager-mixed-fleet-opus-4-6-4-7/): Engineering manager's deep-dive on managing a Claude Code team split between Opus 4.6 and 4.7. Reframes the version-uniformity question as a configuration-audit problem, not a model-version problem. Walks the nine documented behavioral changes that propagate through a shared CLAUDE.md, the three breaking API changes that hard-fail half the fleet, the hidden mixed-fleet from the `opus` alias resolving differently across providers, and the five-layer audit checklist EMs should run before scheduling any model-update training.
- [How to Onboard Engineers to Unfamiliar Codebases in Week One With Claude Code](https://readysolutions.ai/blog/2026-05-03-claude-code-engineering-onboarding-unfamiliar-codebases/): Engineering manager's deep-dive arguing the 30-day ramp was always a measurement of senior-engineer translation capacity, not new-engineer reading speed, and that the deployment mistake teams keep making is treating Claude Code as a pairing accelerator for senior engineers instead of putting it directly in the new engineer's hands as their primary investigative interface. Anchored in DX 91-to-49-to-33-day onboarding-time data, Faros AI Productivity Paradox, Anthropic's skill-formation RCT (17% comprehension gap with conceptual-vs-delegation deployment-mode nuance), and a first-sprint-multiple-tickets observation from the author's day-job team.
- [When to Reach for Claude Code's /batch (and When claude-squad, Agent Teams, or Superpowers Wins)](https://readysolutions.ai/blog/2026-05-04-claude-code-batch-when-to-use/): Decision-framework reference for senior engineers evaluating /batch on real work. Covers the bundled-skill mechanics (5-30 worktrees, plan-approval gate, auto-PRs), suitability and anti-patterns from BaristaLabs and Batsov, a five-row comparison table against claude-squad, Agent Teams, the Anthropic Agent SDK, and Superpowers v5, the Feb-April 2026 quality-regression timeline (Fortune + Anthropic April 23 postmortem + GitHub Issues #41411 and #42796 with the 122x daily-cost figure), and a seven-question pre-flight checklist.
- [How to Re-Engage Engineers Who Already Tried AI and Wrote It Off](https://readysolutions.ai/blog/2026-05-06-burned-by-copilot-re-engaging-skeptical-teams/): Contrarian engineering-leadership argument that the team burned by an earlier AI wave is your easiest re-engagement, not your hardest, because their objection is tool-and-era specific (Copilot 2023-2024, raw chat LLMs) and dissolves under a clean diagnostic plus one working agentic-work session on their own code. Backed by Stack Overflow 2025 trust trend (43% to 33%), SWE-bench progression (49% to 77.2%), JetBrains 2026 Claude Code CSAT 91% / NPS 54, METR 19% slowdown RCT addressed head-on, and a four-step recovery playbook with triage heuristics for the three failure modes (hardened anti-AI, platform debt, change fatigue).
- [Stop Loading UX Research Reports Into Claude One at a Time](https://readysolutions.ai/blog/2026-05-06-ux-research-corpus-not-reports/): The UX research disconnect is a retrieval problem, not a reports problem. The fix is structuring research as a shared, queryable corpus any team member's Claude can pull from on demand, exposed via MCP and interpreted through a conventions skill.
- [Claude Routing Has Two Knobs, Not One: A Model x Effort Matrix for 2026](https://readysolutions.ai/blog/2026-05-08-claude-model-effort-matrix/): Visual reference card for technical leaders making Claude routing decisions. Frames model selection as a two-axis decision (model x effort) and walks the structural asymmetries of the matrix: xhigh effort is limited to Fable 5 and the Opus flagships, Haiku 4.5 does not support the effort parameter at all, and Fast Mode pricing forked at Opus 4.8 ($10/$50 vs $30/$150 on 4.6/4.7; no Fast Mode for Fable 5). Quantifies the actual base-rate cost spread (5x output between Opus 4.7 and Haiku 4.5, 10x for Fable 5, 60x against Opus 4.6/4.7 Fast Mode) against the unsourceable "100x" folklore. Engages the prompt-caching counter-argument directly: cached Opus 4.7 input ($0.50/MTok) costs half an uncached Haiku 4.5 read ($1.00/MTok). Production routing example draws from the Ready Solutions Assessment Worker turn-typed architecture (Haiku classify → Sonnet 1-19 → Opus xhigh synthesis → Sonnet fallback). Includes a seven-mode routing failure-pattern grid and the Fable 5 retention gate: Covered Models (Fable 5, Mythos 5) carry a mandatory 30-day minimum retention on the Claude API, zero data retention is unavailable, and ZDR-bound calls fail with a 400.
- [Claude Code iPhone: The Practitioner Playbook for Shipping Without a Laptop](https://readysolutions.ai/blog/2026-05-09-claude-code-iphone-full-stack-mobile-only/): Deep-dive playbook for full-stack development entirely from the Claude iOS app, opened on the out-to-dinner P0 scenario (CTO message lands mid-meal; with the right pipeline, the engineer steps outside, dispatches the session, returns to the table in five minutes with the change merged and deployed before appetizers arrive). Three workflow patterns plus a five-phase end-to-end loop from "I have an idea" to "here's a finished product" in a weekend to a week. The pipeline carrying the verification load is the unlock; the device that started the session matters less than the discipline of the system around it.
- [The AI Diligence Operating System: Schema First, Subagents Second](https://readysolutions.ai/blog/2026-05-12-ai-diligence-operating-system-memo-schema/): Executive-tier argument for boutique private-equity, commercial diligence, and strategy firms hiring a Claude or MCP architect. A Claude-native diligence operating system is a memo schema with subagents attached, not the reverse — the artifact graph determines what each subagent can produce. The six pitched diligence functions (ingestion, synthesis, memo generation, red-flag detection, source verification, follow-up questioning) are downstream of one decision: what does your memo look like, structurally, before any subagent runs. Includes a five-test framework for what your architect should be able to show in the first thirty days and a re-sequenced 30/60/90 plan.
- [What Does 'Critical' Actually Mean in an AI Code Review?](https://readysolutions.ai/blog/2026-05-14-claude-severity-on-a-curve/): Technical-leader deep-dive on AI code review severity calibration. Claude grades severity on a curve — the curve is whatever fills the current context window — so an audit that finds three nits will still rank them Critical / Major / Minor. Names the mechanism in three forces and one complication, distinguishes Anthropic's <1% incorrect-finding accuracy data from severity-label portability, and ships five concrete patterns that flatten the curve plus a two-run diagnostic teams can put on their own pipeline this week.
- [How to Train AI on Your Company's Knowledge: A Field Guide for Non-Engineers](https://readysolutions.ai/blog/2026-05-14-train-ai-company-knowledge-field-guide-non-engineers/): General-audience deep-dive on equipping AI agents with proprietary business knowledge. The thesis: train the AI on how your best people decide, not just every document your team has ever written. Names the document-dump failure mode, walks the eight methods practitioners actually use (system prompt, few-shot examples, tool definitions, memory, retrieval, knowledge base, feedback loops, fine-tuning), maps eight agent archetypes (support, legal, ops, finance, HR, sales, marketing, executive), engages the "just improve retrieval" steel-man honestly, and ends with the three-step starter sequence anyone can run in an afternoon.
- [How to Audit an MCP Server Before Its Upstream API Breaks It](https://readysolutions.ai/blog/2026-05-16-mcp-server-audit-playbook/): Practitioner deep-dive arguing an MCP server is not one-time plumbing but a maintained dependency on someone else's evolving API plus a standing token/security tradeoff. Anchored on the Atlassian HTTP+SSE June 30 2026 sunset and the Aug 2025 Jira REST shutdown, it covers spec/auth drift since early 2025, the official-vs-community resilience-vs-token-economy tradeoff, the MCP security surface (tool poisoning, STDIO RCE, 53% static-credential repos), and ships a six-step re-audit playbook with a quarterly-plus-on-change cadence.
- [The Seat License Is the Cheapest Part of Your AI Coding Tool](https://readysolutions.ai/blog/2026-05-16-ai-coding-tool-true-cost-tco/): Infographic for technical leaders on the true cost of AI coding tools in 2026. The seat is the smallest line: Anthropic's own data puts enterprise Claude Code at $150-250 per developer per month against a $20 sticker, the verification/review tax compounds (Faros telemetry, MIT Sloan/GitClear), and most teams misprice by benchmarking the cheap tier against light usage. Argues per-role provisioning ($100 daily-use, $200 Max for power users, entry tier for light/non-technical) and engages the METR 19%-slower RCT as the steel-man.
- [Claude Code Can Auto-Fix Your Pull Requests Unattended. Who Owns the Review Bar Now?](https://readysolutions.ai/blog/2026-05-17-claude-code-autofix-pr-review-bar/): Technical-leader deep-dive on Claude Code's /autofix-pr command, which lets a cloud agent watch a PR and resolve CI failures and review comments unattended. The thesis: this relocates the merge gate from "is the code right" to "did the agent satisfy the checks," so the review bar has to become an artifact the team owns deliberately rather than an instinct in whoever reviewed the PR. Covers the mechanics and clear-fix/ask-first/no-op decision boundary, the automation-complacency evidence (Parasuraman & Manzey; Perry et al. CCS 2023), the GitHub-identity-replies and issue_comment/Atlantis footguns, and a five-step governance checklist.
- [IDE-Optional Is Earned, Not Granted: Who Owns the Verification Loop](https://readysolutions.ai/blog/2026-05-22-ide-optional-verification-loop/): Deep-dive for engineering leaders reframing Gartner's May 2026 prediction that 65% of teams will treat IDEs as optional by 2027. The headline buries the real shift: control, governance, and verification move out of the editor into a layer you build (skills, quality gates, automation with human oversight). IDE-optional is earned through workflow maturity, not granted by adopting an agent, and usage-based pricing is the forcing function that exposes teams who never decided who owns the verification loop.
- [Claude's New Agent SDK Credit Pool Isn't a Price Hike for Most of You](https://readysolutions.ai/blog/2026-05-18-claude-agent-sdk-credit-pool/): Technical-leader post on Anthropic's June 15 2026 split of programmatic/Agent SDK usage into a separate monthly credit pool ($20 Pro / $100 Max-5x / $200 Max-20x, API-rate metered, non-rollover). Contrarian thesis: the "25x cut" panic generalizes a heavy-user reality — for Pro and most Max-5x it's pure visibility (burn was always under the cap), for Max-20x it's a genuine 15-30x reduction. Either way the response is the same: treat programmatic Claude as metered compute and budget per workload, while designing around the three gaps the cloud-spend analogy misses (non-rollover, no team pooling, agents don't self-throttle).
- [GEO Is Two Jobs, and Your Marketing Team Can Only Do One](https://readysolutions.ai/blog/2026-05-19-geo-engineering-ai-answer-engines/): Executive deep-dive on generative engine optimization arguing GEO is two jobs, not one. Citation selection (earned media, statistics, quotable expertise) is genuinely a content/PR discipline marketing should own; retrieval eligibility (crawl access, machine-readable rendering, entity disambiguation) is an engineering problem marketing structurally can't fix. Two-front contrarian: against handing all of GEO to marketing, and against the vendor playbook that sells schema markup and llms.txt, both of which the evidence (Princeton GEO paper, an Evil Martians JSON-LD test, Google's Gary Illyes on llms.txt, Muck Rack's 82%-earned-media finding, Ahrefs across 75k brands) shows barely move citation. Anchored on the AI-referral conversion premium (~11x) against a ~1% traffic share, the Cloudflare July 2025 default-block, and the author's own on-site @id authority engine plus the adversarial audit that found its citation claims over-promised.
- [The AI Productivity Paradox: What the 2026 Data Shows](https://readysolutions.ai/blog/2026-05-20-ai-productivity-paradox/): Infographic for engineering leaders on the gap between felt and measured AI productivity. Adoption is near-universal (84% use or plan to use AI) but median PR throughput rose about 8% on a 65% usage increase, the METR RCT measured experienced developers 19% slower while they felt 20% faster, and 45% of AI-generated code fails a security review. The gap is a workflow problem, not an AI ceiling: workflow-equipped teams hold a consistent 2-3x while un-equipped trust in the agent's first output produces the defect-and-review tax. Built on METR (including the Feb and May 2026 updates), DX Q1 2026, Veracode, Sonar, DORA, Faros, GitClear, and a Microsoft Research 2026 study.
- [Claude Code /advisor: When to Call It and Where It's Blind](https://readysolutions.ai/blog/2026-05-23-claude-code-advisor-when-to-call-and-where-blind/): Practitioner deep-dive on Claude Code's /advisor slash command. After 159 invocations across 73 sessions in 17 days, the empirical decision rule beats the official guidance: pre-work calls do nearly all the value, final-check calls are mostly cosmetic. Disambiguates /advisor from /agents, /skills, /ultrareview, and the Agent tool with primary-source binary extraction of the slash command from v2.1.150. Names five structural blind spots: cannot read files, cannot replace a structural gate, decisions invisible to per-task reviewers, sycophancy via interaction context, TTL bug class. 80% Claude Code usage, 20% API.
- [When to Orchestrate Claude Code Subagents: A Four-Gate Decision Map](https://readysolutions.ai/blog/2026-05-24-when-to-orchestrate-subagents/): Visual Brief for engineering leaders on the orchestrate-or-not decision. Orchestration is a scope-of-work test, not a parallelism/speed play: does the task split into distinct scopes (research, code, QA, design) that each get more trustworthy in a dedicated agent? Refined by three guardrails (firsthand-knowledge coupling, context blast radius, write safety), and the same test that says skip-it also picks the pattern (stay flat, parallel fan-out, orchestrator-worker, sequential review, planner-executor). Built on Anthropic's 15x-token / 90.2% multi-agent numbers, the MAST failure taxonomy, the equal-token-budget single-agent result (arXiv:2604.02460, which excludes tool use), Gartner's over-40%-canceled forecast, and a first-party 12-subagent pipeline. Back-dated to 2026-05-24.
- [Build Your First MCP Server: The Hard Part Isn't the Code](https://readysolutions.ai/blog/2026-05-26-build-your-first-mcp-server/): Practitioner deep-dive tutorial on building a custom MCP (Model Context Protocol) server from scratch in TypeScript. For a local stdio server the protocol is a thin wrapper over functions you already have, and the genuinely hard part is tool-description design, not the transport or a framework. Builds a ~30-line server, connects it to Claude Desktop and Claude Code, and shows why the description decides routing (72% vs 20% selection; only 0.5% of 41,902 servers earn an A grade). Steel-mans when to adopt a maintained connector instead, and scopes the thesis to local-first with stdio chosen over Streamable HTTP. Back-dated to 2026-05-26.
- [Claude Code Can Run a Thousand Subagents. The Verification Contract Is the Part You Design First.](https://readysolutions.ai/blog/2026-05-28-claude-code-dynamic-workflows-verification-contracts/): Deep-dive for engineering leaders on Claude Code dynamic workflows (v2.1.154, 2026-05-28), which fan out up to 1,000 background subagents per run. At fleet scale no human reads every diff, so the verification contract (acceptance criteria, adversarial checks, deterministic gates, stop conditions) designed before the fleet runs becomes the primary design artifact, not the prompt and not the agent. Verification ships as a composable pattern, not a default; teams that treat dynamic workflows as more-agents-same-review inherit a green light that means less than it used to.
- [I Read the Files Claude Code Wrote About Me: How Auto-Memory Actually Works](https://readysolutions.ai/blog/2026-05-29-claude-code-auto-memory-how-it-works/): Mechanism deep-dive on Claude Code's auto-memory: where the files live on disk, what triggers a save, the four documented memory types, and how to read, steer, and disable what Claude records about you.
- [The Most Important Number in Claude Opus 4.8 Isn't a Benchmark](https://readysolutions.ai/blog/2026-05-29-claude-opus-4-8-honesty/): Deep-dive for technical leaders on Claude Opus 4.8's headline change: honesty, not raw capability. Anthropic reports the model is around four times less likely to let flaws in code it wrote pass unremarked. The post argues the real tax on running agents unsupervised was never intelligence but re-verifying a confident model's false "done," explains why no public benchmark can see the gain, and makes the contrarian case that an honesty claim you cannot verify from outside still leaves you owning the verification loop.
- [Claude Opus 4.8 in Claude Code: I Couldn't Trust What It Said About Its Own Tools.](https://readysolutions.ai/blog/2026-05-31-opus-4-8-tool-state-regression/): Deep-dive for practitioners and technical leaders on a launch-week tool-state failure. In the author's Claude Code loops, Opus 4.8 re-fires identical reads and emits throwaway echo/flush probes for output that is not stuck; public bug reports show a darker fabrication branch (declaring a build verified it never ran). First-party measurement across 28,995 messages finds wide tool-call batches about seven times more common on 4.8 than 4.7. The model-vs-harness cause is unresolvable from outside, because Claude Code v2.1.154 shipped the model and a streaming-tool-execution change together. The fix is to instrument your own verification loop and keep known-good models on critical paths until you have receipts.
- [How to Instrument Observability for Production Claude Agents](https://readysolutions.ai/blog/2026-06-01-agent-observability-when-self-report-isnt-enough/): Self-report and a binary busy/idle status are not observability. Maps the four native Claude surfaces that shipped or matured this spring (Agent View, Claude Code CLI OpenTelemetry, Agent SDK traces, Managed Agents), narrows the residual gap to an always-on operator-state glance over attached terminal sessions, and reframes "build it yourself" as owning a telemetry contract for agents you cannot watch. Target audience: technical leaders running Claude agents in production.
- [How to Control Claude Code Token Spend Without Sacrificing Output](https://readysolutions.ai/blog/2026-06-02-claude-code-token-spend/): Deep-dive for engineers running Claude Code agentically. Token spend is set by the session's shape, not the prompt, and three structural levers (default model + effort, context hygiene, subagent fan-out) govern both cost and output quality, so the goal is the cheapest setup that still clears your quality bar, not minimum spend. Covers opusplan and /effort, /clear and /compact, the agent-teams ~7x multiplier, /usage (with /cost and /stats aliases) and OpenTelemetry, the ANTHROPIC_API_KEY surprise-bill trap, and when the right move is to spend more.
- [Is Your AI Rollout Actually Working? The Metrics That Matter at Month 3](https://readysolutions.ai/blog/2026-06-03-measure-ai-rollout-success-month-3/): Field note for engineering leaders on measuring AI coding-tool adoption. Seat activation and usage volume are vanity metrics; the real signal a rollout worked is whether the team's workflow changed (task-category spread, depth of use, the direction the senior support burden moved). Month three is the first honest checkpoint, not the final grade. Covers the vanity-vs-signal metric split, leading workflow-change indicators, the senior-burden-shift diagnostic, and the lagging DORA and ROI metrics that confirm it, plus the anti-metrics to retire.
- [Opus 4.8 vs 4.7, One Week Later: The Upgrade Call I Couldn't Make on Day One](https://readysolutions.ai/blog/2026-06-05-opus-4-8-one-week-later/): Field note for engineering leaders weighing a 4.7-to-4.8 upgrade. A week after a rocky launch, Opus 4.8 became the author's default Claude Code model: the launch-week tool-state bug got a dated changelog fix (Claude Code v2.1.161, 2 June, so a failed Bash call no longer cancels its parallel batch), the honesty gain held on instrumented code work, and the verdict is to treat a model launch as a measurement window, not a day-one verdict. Covers what the fix did and did not cover (harness cancellation vs model-side over-trust), why the honesty pays off in front of a deterministic check but not against an adversary (Andon Labs Vending-Bench), and when to move versus stay on 4.7.
- [I Ran Seven Claude Code Sessions on One Repo. Three Spiraled. Git Worktrees Were Only Half the Fix.](https://readysolutions.ai/blog/2026-06-06-parallel-claude-code-sessions-one-repo/): Practitioner deep-dive on running 4-8 parallel Claude Code sessions against one repository. Opens from the failure mode (three of seven sessions in a context death spiral as adjacent files changed mid-run), then the one-hour structural fix: git worktrees as the session default, a PreToolUse write-path guard hook, and a serialized merge gate. Honest about what worktrees don't isolate (shared .git internals, ~/.claude.json, runtime ports/databases/disk) and why partitioning quality, not session count, is the binding constraint (AgenticFlict: 27.67% of 142k+ parallel agentic PRs hit merge conflicts; similarity predicts collision). Ends with a six-step playbook.
- [Claude Code Hooks, Visualized: The Lifecycle Map, the Deny Gate, and the Enforcement Ladder](https://readysolutions.ai/blog/2026-06-07-claude-code-hooks-visualized/): Visual brief for practitioners on Claude Code hooks: a bespoke session-lifecycle map of where hooks fire (SessionStart through PreToolUse to Stop), the PreToolUse allow/ask/deny/defer decision gate with payload anatomy, and the enforcement ladder explaining why CLAUDE.md rules slip (instruction-following reliability drops 18.3% under rephrasing even for frontier models) while hooks execute every time. Honest about the gaps: the subagent dispatch hole, settings self-edit, Bash route-arounds, and CVE-2025-59536.
- [Claude Data Retention Isn't a Mystery. Your Approval Process Is.](https://readysolutions.ai/blog/2026-06-08-claude-data-handling-due-diligence/): The four data-handling questions that stall Claude approval (training use, retention windows, zero data retention, certifications) have published vendor answers. An afternoon of reading converts an open-ended security debate into a short list of scoped contract steps, and the post is honest about where an afternoon isn't enough: the ZDR safety carve-out, BAA contracting time, EU transfer impact assessments, and the personal-account shadow-IT trap. Target audience: executives and leaders whose Claude security review has stalled.
- [Claude Fable 5 Is 'Mostly Drop-In.' The Word Doing the Work Is 'Mostly.'](https://readysolutions.ai/blog/2026-06-09-claude-fable-5-mostly-drop-in/): Launch-day decode of the Opus 4.8 to Claude Fable 5 migration for engineering leaders. The request shape holds still while the operating contract moves: always-on adaptive thinking at $10/$50 per million tokens (2x Opus 4.8), a refusal path that arrives as HTTP 200 with per-surface fallback asymmetry (consumer apps auto-switch to Opus 4.8 with a notice; the API needs the opt-in beta fallbacks parameter or SDK middleware, absent on Bedrock/Vertex/Foundry/Batches), and a 30-day retention rule that hard-400s ZDR workspaces on the Claude API. Ends with a five-step contract preflight: max_tokens audit, refusal instrumentation, retention review, effort re-baseline at high, and a conversation-replay audit.
- [Claude Fable 5's Silent Degradation: The Safety Tier You Can't See, Log, or Turn Off](https://readysolutions.ai/blog/2026-06-10-claude-fable-5-silent-degradation/): Deep-dive field note for engineering leaders and senior practitioners mapping Claude Fable 5's safety surface as three distinct tiers. The loud tier is visible refusal classifiers (HTTP 200 with stop_reason refusal and a stop_details.category, loggable and routable to Opus 4.8). The silent tier degrades rather than refuses on a narrow band of frontier-ML topics (~0.03% of traffic) with no API signal and no opt-out, disclosed only in the system card. The third is the model's own documented diligence failures, which Anthropic files under alignment. Argues the loud tier is your day-to-day cost, the silent tier is the precedent worth watching, and the diligence tier needs human oversight no classifier supplies. Covers Terminal-Bench's 20.9%-of-trials safety-refusal rate, server-side-fallback detection via the usage.iterations fallback_message entry, and a canary regression smoke test for the unobservable tier. Cross-model reviewed by a 4-lens Codex swarm before publish.
- [Claude Code Subagents vs Skills: One Teaches the Session, the Other Staffs It](https://readysolutions.ai/blog/2026-06-11-claude-code-custom-subagents-vs-skills/): Deep-dive practitioner tutorial on creating custom Claude Code subagents (.claude/agents file anatomy, description-as-routing-rule, tool allowlists, model selection) and the context-architecture distinction that decides when a subagent beats a skill. A skill loads procedural knowledge into the session that invoked it; a subagent staffs the work out to a separate context window with its own tool boundaries. Covers the four routing signals that justify a subagent (bulky throwaway output, parallel lanes, enforced tool boundaries, cold-eyes verification), the token bill that comes with them, the skill-catalog routing-decay research, and the composition case (skills preloaded inside subagents) with the documented context: fork gap.
- [The Changelog Says 5 Levels. My Probe Went 9 Deep. Inside Claude Code's Nested Subagents.](https://readysolutions.ai/blog/2026-06-11-claude-code-nested-subagents/): Deep-dive companion to the subagents-vs-skills post, on the v2.1.172 nesting capability (file-based subagents spawning their own subagents, June 10 2026). Thesis: depth is a budget, not headroom; every delegation boundary strips context on the way down and information on the way up, so nesting earns its keep only when the work itself is recursive, and the common failure is mirroring an org chart into an agent tree. Opens with a first-party probe that ran a recursive subagent chain nine levels deep with zero errors, disk-trace verified, past the documented five-level cap. Maps the four delegation surfaces (only CLI file subagents nest; SDK, Agent Teams, and the Managed Agents API stay flat), the live docs-vs-changelog contradiction, the recursive-shape test, the token/error-amplification/visibility bill, a RecursiveMAS-and-Osmani steel man, and a six-step production discipline. Target audience: practitioners running multi-agent Claude Code workloads.
- [The Approval Budget: A Model for Spending Human Attention in Agentic Workflows](https://readysolutions.ai/blog/2026-06-12-approval-fatigue-attention-budget/): Approval fatigue treated as a budget problem, not a discipline problem. Anthropic's own telemetry shows users approve 93% of Claude Code permission prompts; each prompt an operator can't genuinely evaluate trains the next reflexive yes, so past a threshold every added gate subtracts safety. Maps where a tool call should be settled before it costs attention (deny rules, allow rules, PreToolUse hooks, OS sandbox, permission modes), steel-mans the structural-bypass catalog (Cursor denylist mathematics, Claude Code sandbox escapes, Semantic Kernel CVEs, EU AI Act Article 14), and closes with a reallocation playbook: three properties that earn a prompt, and the approval-rate diagnostic. Target audience: technical leaders running agentic rollouts.
- [When to Trust an LLM Judge (and When to Pull It Off the Gate)](https://readysolutions.ai/blog/2026-06-13-when-to-trust-an-llm-judge/): An LLM-as-a-judge earns trust as a prioritizer by default and as a merge-gate only under two measured conditions (agreement with ground truth on the exact task class, plus a bounded blast radius), because an uncalibrated verdict shares the generator's blind spots and is ordinal at best. Covers the three-layer verification stack (deterministic floor where the model gets no vote, the LLM judge band that prioritizes, the human arbiter that decides), quantified failure modes (35% verdict-flip on answer-order swap, >90% verbosity bias, a nine-judge panel collapsing to roughly two independent votes, ~1-in-5 production defect recall), calibration via measured true-positive and true-negative rates, triangulation that verifies against outcomes rather than stated reasoning, and the RLAIF / Constitutional AI steel-man (training-time aggregate signals that meet both conditions). Target audience: practitioners building AI verification pipelines.
- [MCP Authorization Scope: Least Privilege Before You Connect](https://readysolutions.ai/blog/2026-06-14-mcp-authorization-scope-least-privilege/): Deep-dive field note for technical leaders on why the hard part of enterprise MCP is the authorization scope you decide before connecting, not the connection itself. An over-broad scope turns a connector into a lethal-trifecta blast radius (private data plus untrusted content plus a write channel on one connection); the fix is a five-step pre-connection scope review (enumerate resources, classify verbs, map egress, rate the blast radius, name an owner) whose every row binds to a concrete enforcement point or is logged as an accepted exception. Covers Claude Code's oauth.scopes pin and deny-then-ask-then-allow tool rules, Atlassian Rovo's admin Read/Write/Search permission groups that default to the user's full permissions, the MCP spec's mandatory RFC 8707 resource indicators and token-passthrough ban, and the steel-man that most deployed MCP servers (38.7% with no auth) lack OAuth scopes at all. Target audience: senior and staff engineers, platform leads, and security-adjacent engineering managers.
- [Model Availability Is a Production Dependency: Build the Fallback Ladder Before the Next Model Vanishes](https://readysolutions.ai/blog/2026-06-15-model-availability-production-dependency/): Standard field note for technical leaders treating model availability as a production dependency rather than a vendor guarantee. Anchored on the June 12 2026 worldwide suspension of Claude Fable 5 by a US government export-control directive, three days after launch. Thesis: most teams wired their continuity plan to the wrong failure. Anthropic's fallbacks parameter fires only on safety-classifier refusals ("a rate limit, overload, or server error on the requested model is returned to you as-is") and the Claude Code fallbackModel chain is IDE-only, so neither native path covers a whole-model suspension. The fix is a pre-built fallback ladder (same-provider alias, capability-class sibling, tested cross-provider standby) with the model id behind config so a swap is a configuration change not a redeploy, a canary baseline to detect changes, and shallow coupling until a model earns trust. First-person anchor: the author's revert to Opus 4.8 was a zero-config non-event because the fallback was already vetted over 8 days and nothing was coupled to Fable 5 in its 3-day life. Steel-mans the native-tooling-is-enough objection and frames planned deprecation (60-day notice; Sonnet 4 and Opus 4 retired June 15) as the chronic version of the same risk. Target audience: technical leaders running production Claude integrations.
- [The Lethal Trifecta Has Three Legs. Least Privilege Covers One.](https://readysolutions.ai/blog/2026-06-16-agent-egress-least-privilege/): Deep-dive field note for senior engineers and technical leads on why least-privilege permission scoping is necessary but not sufficient for production AI agents. Scope governs only the access leg of the lethal trifecta; the breach happens on the outbound path, through the untrusted-input and egress legs the scope review never touches. Frames a three-leg-plus-orthogonal-blast-radius model (scoped explicitly to prompt-injection exfiltration, with adjacent threats like connector supply-chain and credential compromise named as out of frame), then maps controls: tool-result quarantine for untrusted input, destination-provenance egress control (why a domain allowlist alone fails, per Anthropic's own allowlisted-endpoint bypass), and human gates plus deterministic limits for blast radius. Anchored on the Notion AI exfiltration attack (permitted tools only) and the author's daily MCP-connected agent footprint. Steel-mans the "you can't egress-proof everything, gate it all with a human" objection. Target audience: senior and staff engineers and technical leads deploying or reviewing production Claude agents.

## Case Studies

- [A Full-Stack App, Built Without a Keyboard](https://readysolutions.ai/case-studies/2026-04-10-vibecheckme-full-stack-ai-app-from-mobile/): Production AI app powered by Claude API, built in a weekend from a phone. React, Cloudflare Workers, streaming Claude API integration. 253 commits, zero hand-written code.
- [59/60 Voice Fidelity: Multi-Agent AI Persona Profiler](https://readysolutions.ai/case-studies/2026-04-11-ai-persona-profiler/): Multi-agent pipeline interviewing real people, running adversarial dual-analysis, and producing persona files scoring 59/60 on quantitative voice fidelity testing. 10+ coordinated Claude Opus instances, 8.5k LOC orchestration, built in 8 days.
- [Claude Code Opus 4.7 Migration Tool: 5-Layer Compatibility Scanner](https://readysolutions.ai/case-studies/2026-05-06-opus-compatibility-scanner/): MIT-licensed Claude Code skill that audits CLAUDE.md, AGENTS.md, settings.json plus hooks, skill bodies, and Anthropic SDK call sites (with package manifests scanned alongside) against 70 catalogued Opus 4.6 / 4.7 compatibility pattern IDs. Three Critical and four Warning findings surfaced on the author's own most-curated setup. Per-finding apply, Tier-1 citation enforcement, never proposes a change that degrades either model.
- [Claude Code Case Study: A Five-Layer AI Writing Pipeline](https://readysolutions.ai/case-studies/2026-05-08-ai-authoring-pipeline/): A five-layer Claude Code authoring pipeline catching seven recurring failure modes of AI-assisted writing before they reach a reader. Hub-and-spoke orchestration with twelve specialist subagents doing judgment work, sixteen deterministic validators backing them, and a parent skill owning every disk write. Companion 14-page technical white paper "Engineering trust into AI-authored content" available as a download.

## Contact

- Email: hello@readysolutions.ai
- Booking: https://calendly.com/hello-readysolutions/30min