14 lines
41 KiB
JavaScript
14 lines
41 KiB
JavaScript
import{c as e,Q as r,j as o,m as a}from"./chunks/framework.BPKcPtvA.js";const u=JSON.parse('{"title":"Competitive Analysis: Agentic Engineering Course Landscape","description":"","frontmatter":{},"headers":[],"relativePath":"modules/competitive-analysis.md","filePath":"modules/competitive-analysis.md","lastUpdated":1780488246000}'),d={name:"modules/competitive-analysis.md"};function n(s,t,i,l,h,c){return r(),o("div",null,[...t[0]||(t[0]=[a(`<h1 id="competitive-analysis-agentic-engineering-course-landscape" tabindex="-1">Competitive Analysis: Agentic Engineering Course Landscape <a class="header-anchor" href="#competitive-analysis-agentic-engineering-course-landscape" aria-label="Permalink to "Competitive Analysis: Agentic Engineering Course Landscape""></a></h1><h2 id="live-comparison-walkinglabs-github-io-course" tabindex="-1">Live Comparison: walkinglabs.github.io Course <a class="header-anchor" href="#live-comparison-walkinglabs-github-io-course" aria-label="Permalink to "Live Comparison: walkinglabs.github.io Course""></a></h2><p>I fetched the actual Anthropic Learn Harness Engineering website to compare directly.</p><h3 id="their-structure-12-lectures-6-projects" tabindex="-1">Their Structure (12 lectures, 6 projects) <a class="header-anchor" href="#their-structure-12-lectures-6-projects" aria-label="Permalink to "Their Structure (12 lectures, 6 projects)""></a></h3><table tabindex="0"><thead><tr><th>#</th><th>Lecture Topic</th><th>Projects</th></tr></thead><tbody><tr><td>1</td><td>Why Capable Agents Still Fail (50% SWE-bench, harness problem)</td><td>Project 01: Prompt-Only vs Rules-First</td></tr><tr><td>2</td><td>What a Harness Actually Is (5 subsystems: instructions, tools, environment, state, verification)</td><td>Project 02: Agent-Readable Workspace</td></tr><tr><td>3</td><td>Why the Repository Must Become the System of Record</td><td>Project 03: Multi-Session Continuity</td></tr><tr><td>4</td><td>Why One Giant Instruction File Fails</td><td>Project 04: Runtime Feedback & Scope Control</td></tr><tr><td>5</td><td>Why Long-Running Tasks Lose Continuity</td><td>Project 05: Self-Verification & Role Separation</td></tr><tr><td>6</td><td>Why Initialization Needs Its Own Phase</td><td>Project 06: Full Harness (observability + debugging)</td></tr><tr><td>7</td><td>Why Agents Overreach and Under-Finish</td><td></td></tr><tr><td>8</td><td>Why Feature Lists Are Harness Primitives</td><td></td></tr><tr><td>9</td><td>Why Agents Declare Victory Too Early</td><td></td></tr><tr><td>10</td><td>Why End-to-End Testing Changes Results</td><td></td></tr><tr><td>11</td><td>Why Observability Belongs Inside the Harness</td><td></td></tr><tr><td>12</td><td>Why Every Session Must Leave a Clean State</td><td></td></tr></tbody></table><h3 id="key-differences-your-course-vs-theirs" tabindex="-1">Key Differences: Your Course vs Theirs <a class="header-anchor" href="#key-differences-your-course-vs-theirs" aria-label="Permalink to "Key Differences: Your Course vs Theirs""></a></h3><p><strong>Their strength</strong>: Deep philosophical focus on the "harness" concept (instructions, tools, environment, state, verification). Each lecture is a "why" — explaining root causes of agent failure. Clean, professional presentation with VitePress.</p><p><strong>Their gaps</strong> (what they DON'T cover that you do):</p><ul><li>❌ <strong>Zero security content</strong> — no bash security ladder, no damage control, no hook architecture</li><li>❌ <strong>Zero multi-agent orchestration</strong> — no teams, chains, delegation, P2P, CEO Board</li><li>❌ <strong>Zero production deployment</strong> — no CI/CD, shadow deploys, rollback, monitoring</li><li>❌ <strong>Zero model economics</strong> — no pricing comparison, cascade routing, cost optimization</li><li>❌ <strong>Zero autoresearch</strong> — no self-improving agents</li><li>❌ <strong>Zero meta-agents</strong> — no agents that build agents</li><li>❌ <strong>Zero non-technical frameworks</strong> — no decision frameworks for managers</li></ul><p><strong>Their unique content</strong> (what they cover that you could add):</p><ul><li>✓ Harness = 5 subsystems model (instructions, tools, environment, state, verification) — Your "harness" framing could adopt this language</li><li>✓ The "repo IS the spec" philosophy — Stronger emphasis on repo-as-source-of-truth</li><li>✓ Agent-readable workspace design patterns</li><li>✓ Session continuity across long-running tasks</li><li>✓ Initialization as a distinct phase</li><li>✓ Feature lists as harness primitives (structured feature tagging)</li><li>✓ Clean state between sessions</li><li>✓ Professionally designed VitePress site with multi-language support (11 languages)</li><li>✓ "Try Harness" button linking to templates and a live agentic session</li></ul><hr><h2 id="live-comparison-claudefast-claudefa-st" tabindex="-1">Live Comparison: ClaudeFAST (claudefa.st) <a class="header-anchor" href="#live-comparison-claudefast-claudefa-st" aria-label="Permalink to "Live Comparison: ClaudeFAST (claudefa.st)""></a></h2><p>ClaudeFAST is a commercial product/blog selling Claude Code kits ($89-$299). Their blog reveals what content they think is valuable.</p><h3 id="claudefast-blog-topics" tabindex="-1">ClaudeFAST Blog Topics <a class="header-anchor" href="#claudefast-blog-topics" aria-label="Permalink to "ClaudeFAST Blog Topics""></a></h3><table tabindex="0"><thead><tr><th>Topic</th><th>Price Point</th><th>Our Coverage</th></tr></thead><tbody><tr><td>Agent Manager role (job scope, DRI, 90-day playbook)</td><td><strong>Free blog</strong></td><td>M5 — Production patterns covers deployment, but not organizational roles</td></tr><tr><td>Large codebases (monorepos, legacy, 8 strategies)</td><td><strong>Free blog</strong></td><td>M2 — Architecture covers context management, not specifically monorepo</td></tr><tr><td>Path-scoped skills (file-path-based skill loading)</td><td><strong>Free blog</strong></td><td>M4 briefly covers skills; we don't have a dedicated skills lesson</td></tr><tr><td>Subdirectory CLAUDE.md (tree-walking context)</td><td><strong>Free blog</strong></td><td>Not covered — interesting pattern for large projects</td></tr><tr><td>AI Layer thesis (harness > model)</td><td><strong>Free blog</strong></td><td>M1 — We cover this extensively</td></tr><tr><td>Self-improving CLAUDE.md (stop hook reviews changes)</td><td><strong>Free blog</strong></td><td>M3 — Hooks architecture covers this pattern</td></tr><tr><td>LSP MCP Server (symbol-level search)</td><td><strong>Free blog</strong></td><td>Not covered — MCP server for code intelligence</td></tr><tr><td>Plugins distribution (shareable bundles)</td><td><strong>Free blog</strong></td><td>Not covered — packaging/sharing agent configs</td></tr></tbody></table><h3 id="their-product-kits" tabindex="-1">Their Product Kits <a class="header-anchor" href="#their-product-kits" aria-label="Permalink to "Their Product Kits""></a></h3><table tabindex="0"><thead><tr><th>Kit</th><th>Price</th><th>What It Is</th><th>Gap in Our Course</th></tr></thead><tbody><tr><td>Code Kit</td><td><strong>$89</strong></td><td>AI Development Framework</td><td>We have this code-wise, not packaged</td></tr><tr><td>Growth Kit</td><td><strong>$89</strong></td><td>Sales, Marketing Research</td><td>We don't cover agent use cases for sales/marketing</td></tr><tr><td>Complete Kit</td><td><strong>$149</strong></td><td>Code + Growth Kits</td><td>Bundled value</td></tr><tr><td>Shopify Kit</td><td><strong>$199</strong></td><td>Ecom Consulting for Claude</td><td>Domain-specific (Shopify)</td></tr><tr><td>Ultimate Kit</td><td><strong>$299</strong></td><td>Complete + Shopify</td><td>Premium bundle</td></tr></tbody></table><h3 id="what-their-16-agents-280-skills-tell-us" tabindex="-1">What Their 16 Agents + 280 Skills Tell Us <a class="header-anchor" href="#what-their-16-agents-280-skills-tell-us" aria-label="Permalink to "What Their 16 Agents + 280 Skills Tell Us""></a></h3><p>They claim "16 agents, 280 skill files, 90,000+ lines of workflows." This suggests:</p><ul><li><strong>Breadth over depth</strong>: 280 skills likely means many small, focused skill files</li><li><strong>Agent diversity</strong>: 16 agents implies specialized personas (not just coding)</li><li><strong>Production scale</strong>: 90K lines is substantial — they've invested heavily</li></ul><p><strong>Actionable gap</strong>: We should consider adding a "Skills System Deep Dive" lesson and a "Plugin/Kit Distribution" lesson showing how to package agent configs for sharing.</p><h3 id="what-you-should-borrow" tabindex="-1">What You Should Borrow <a class="header-anchor" href="#what-you-should-borrow" aria-label="Permalink to "What You Should Borrow""></a></h3><ol><li><strong>Adopt the "5 subsystems" framing</strong> for your harness definition — it's clean and teachable</li><li><strong>Your M3 Safety module</strong> maps to their "verification" subsystem — lead with that connection</li><li><strong>Add a "clean state" lesson</strong> — their Lecture 12 is a good pattern</li><li><strong>Add "repo IS the spec" to M1</strong> — it's a powerful mental model</li><li><strong>Consider VitePress for the final site</strong> — it's what Anthropic uses and it's excellent for docs-based courses</li></ol><h2 id="course-comparison-matrix" tabindex="-1">Course Comparison Matrix <a class="header-anchor" href="#course-comparison-matrix" aria-label="Permalink to "Course Comparison Matrix""></a></h2><table tabindex="0"><thead><tr><th>Dimension</th><th><strong>Your Course</strong></th><th><strong>IndyDevDan TAC</strong></th><th><strong>Anthropic Harness</strong></th><th><strong>DeepLearning.AI</strong></th><th><strong>LangChain Academy</strong></th></tr></thead><tbody><tr><td><strong>Price</strong></td><td>Not set</td><td>$97-147</td><td>Free</td><td>$45/mo sub</td><td>Free</td></tr><tr><td><strong>Format</strong></td><td>Self-paced docs + labs</td><td>Video + cohort</td><td>Docs only</td><td>Video + notebook</td><td>Docs + code</td></tr><tr><td><strong>Hours</strong></td><td>44-62</td><td>~20</td><td>~30</td><td>1-2 per course</td><td>~10</td></tr><tr><td><strong>Labs</strong></td><td>13 scaffolded</td><td>~5 demos</td><td>Many exercises</td><td>~3 per course</td><td>~8</td></tr><tr><td><strong>Code access</strong></td><td>Full source</td><td>Member assets</td><td>Open source</td><td>Notebooks</td><td>Open source</td></tr><tr><td><strong>Certificate</strong></td><td>Not yet</td><td>✓</td><td>✗</td><td>✓</td><td>✗</td></tr></tbody></table><hr><h2 id="module-by-module-comparison" tabindex="-1">Module-by-Module Comparison <a class="header-anchor" href="#module-by-module-comparison" aria-label="Permalink to "Module-by-Module Comparison""></a></h2><h3 id="foundations-what-is-an-agent-harness-vs-model-decision-frameworks" tabindex="-1">Foundations (What is an agent, harness vs model, decision frameworks) <a class="header-anchor" href="#foundations-what-is-an-agent-harness-vs-model-decision-frameworks" aria-label="Permalink to "Foundations (What is an agent, harness vs model, decision frameworks)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>Agent = LLM + Tools + Loop</td><td>✓ Covers</td><td>✓ Covers</td><td>✓ Covers</td><td>✓ Covers</td></tr><tr><td>Harness vs Model distinction</td><td><strong>DIFFERENTIATOR</strong></td><td>✗ Not covered</td><td>✗ Not covered</td><td>✗ Not covered</td></tr><tr><td>Decision framework (4-question filter)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗ Not covered</td><td>✗ Not covered</td><td>✗ Not covered</td></tr><tr><td>Vibe coding vs Agentic engineering (5 rules)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗ Not covered</td><td>✗ Not covered</td><td>✗ Not covered</td></tr></tbody></table><p><strong>Verdict</strong>: Your foundations module has 3 unique differentiators not found in any competitor course.</p><hr><h3 id="agent-architecture-tools-loops-context-memory" tabindex="-1">Agent Architecture (Tools, loops, context, memory) <a class="header-anchor" href="#agent-architecture-tools-loops-context-memory" aria-label="Permalink to "Agent Architecture (Tools, loops, context, memory)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>Four pillars (Tools/Loop/Context/Memory)</td><td>✓ Covered</td><td>✓ Partial</td><td>✓ Partial</td><td>✓ Partial</td></tr><tr><td>Tool design patterns (MCP/CLI/Script/Skill)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Agent loop variants (5 levels)</td><td>✓ Covered</td><td>✓ Partial</td><td>✓ Partial</td><td>✓ Basic</td></tr><tr><td>Context window management</td><td>✓ Covered</td><td>✓ Covered</td><td>✓ Covered</td><td>✓ Covered</td></tr><tr><td>The Reasoning Parameter</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Codebase architectures (4 patterns)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: Tool design channels and the reasoning parameter are unique. Codebase architectures (from your single-file-agents research) are not taught anywhere else.</p><hr><h3 id="safety-security-5-level-ladder-hooks-verifier" tabindex="-1">Safety & Security (5-level ladder, hooks, verifier) <a class="header-anchor" href="#safety-security-5-level-ladder-hooks-verifier" aria-label="Permalink to "Safety & Security (5-level ladder, hooks, verifier)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>5-level bash security ladder</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>L3 marquee break (agent writes cleanup.py)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Damage control (3 access levels)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Hook architecture (13 lifecycle events)</td><td>✓ Covered</td><td>✓ Partial</td><td>✓ Covered</td><td>✗</td></tr><tr><td>Verifier pattern (builder + verifier)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Defense-in-depth stacking</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Probability math (1% × 100 turns = 63%)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: This is your STRONGEST differentiator. No other course covers agent security at this depth. The bash-damage-from-within material is genuinely novel.</p><hr><h3 id="multi-agent-orchestration-teams-chains-p2p" tabindex="-1">Multi-Agent Orchestration (teams, chains, P2P) <a class="header-anchor" href="#multi-agent-orchestration-teams-chains-p2p" aria-label="Permalink to "Multi-Agent Orchestration (teams, chains, P2P)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>3 orchestration patterns (dispatch/pipeline/P2P)</td><td>✓ Covered</td><td>✓ Covered</td><td>✓ Basic</td><td>✓ Partial</td></tr><tr><td>Depth-2 delegation (orchestrator→lead→worker)</td><td><strong>DIFFERENTIATOR</strong></td><td>✓ Covered</td><td>✗</td><td>✗</td></tr><tr><td>Mental models (agent-owned expertise files)</td><td><strong>DIFFERENTIATOR</strong></td><td>✓ Covered</td><td>✗</td><td>✗</td></tr><tr><td>Domain locking (per-agent permissions)</td><td><strong>DIFFERENTIATOR</strong></td><td>✓ Covered</td><td>✗</td><td>✗</td></tr><tr><td>TillDone task discipline</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Agent chains ($INPUT/$ORIGINAL YAML)</td><td><strong>DIFFERENTIATOR</strong></td><td>✓ Covered</td><td>✗</td><td>✗</td></tr><tr><td>Pi-to-Pi peer-to-peer coms</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>CEO Board System (8 agents, adversarial debate)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>UI Agents (12 agents, brand→product→tree)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: You share depth-2 delegation and agent chains with TAC, but add 6 unique patterns (TillDone, P2P coms, CEO Board, UI Agents, mental model self-improve, domain locking).</p><hr><h3 id="production-patterns-deployment-ci-cd-monitoring" tabindex="-1">Production Patterns (deployment, CI/CD, monitoring) <a class="header-anchor" href="#production-patterns-deployment-ci-cd-monitoring" aria-label="Permalink to "Production Patterns (deployment, CI/CD, monitoring)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>CI/CD for agents (golden datasets)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✓ Partial</td><td>✗</td></tr><tr><td>Shadow deployments</td><td>✓ Covered</td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Rollback strategies (prompt + model + config)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Observability (SQLite tracing)</td><td>✓ Covered</td><td>✗</td><td>✓ Covered</td><td>✗</td></tr><tr><td>Agent-specific alerting</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Deployment modes (local/auth/public)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Budget control (warnings vs hard stops)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: Anthropic and LangChain cover basic observability, but no course covers agent-specific deployment patterns (shadow deploy, rollback, budgeting). This is a greenfield differentiator.</p><hr><h3 id="economics-evaluation-pricing-cost-optimization-evals" tabindex="-1">Economics & Evaluation (pricing, cost optimization, evals) <a class="header-anchor" href="#economics-evaluation-pricing-cost-optimization-evals" aria-label="Permalink to "Economics & Evaluation (pricing, cost optimization, evals)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>LLM pricing landscape (100x range)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Cascade routing (cheap/expensive model mix)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Cost per session math (3x rule)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>pass@k / pass^k evaluation</td><td>✓ Covered</td><td>✗</td><td>✓ Basic</td><td>✗</td></tr><tr><td>Golden datasets + regression testing</td><td>✓ Covered</td><td>✗</td><td>✓ Covered</td><td>✗</td></tr><tr><td>A/B testing agents (canary deploys)</td><td><strong>DIFFERENTIATOR</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: The 3x rule and practical cascade routing math are unique. Most courses skip cost optimization entirely.</p><hr><h3 id="advanced-topics-autoresearch-meta-agents" tabindex="-1">Advanced Topics (autoresearch, meta-agents) <a class="header-anchor" href="#advanced-topics-autoresearch-meta-agents" aria-label="Permalink to "Advanced Topics (autoresearch, meta-agents)""></a></h3><table tabindex="0"><thead><tr><th>Your Course</th><th>TAC</th><th>Anthropic</th><th>DL.AI</th><th>LangChain</th></tr></thead><tbody><tr><td>Autoresearch (self-improving agents)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Integrity guards (code hashing, grind detect)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Meta-agents (agents that build agents)</td><td><strong>DIFFERENTIATOR</strong></td><td>✓ Covered</td><td>✗</td><td>✗</td></tr><tr><td>Beyond MCP (context cost trade-off)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Mac Mini Agent (physical sandbox)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Always-on agents (voice→CLI bridge)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr><tr><td>Framework comparison (paperclip/multica/cabinet)</td><td><strong>UNIQUE</strong></td><td>✗</td><td>✗</td><td>✗</td></tr></tbody></table><p><strong>Verdict</strong>: Autoresearch with integrity guards (from your mythos-learnings research) is entirely unique. No course teaches agents that improve themselves.</p><hr><h2 id="differentiator-summary" tabindex="-1">Differentiator Summary <a class="header-anchor" href="#differentiator-summary" aria-label="Permalink to "Differentiator Summary""></a></h2><h3 id="content-only-your-course-has-unique" tabindex="-1">Content Only Your Course Has (UNIQUE) <a class="header-anchor" href="#content-only-your-course-has-unique" aria-label="Permalink to "Content Only Your Course Has (UNIQUE)""></a></h3><table tabindex="0"><thead><tr><th>Topic</th><th>Source Material</th><th>Why No One Else Has It</th></tr></thead><tbody><tr><td><strong>5-level bash security ladder</strong></td><td>bash-damage-from-within (your repo)</td><td>IndyDevDan created this, you have the source</td></tr><tr><td><strong>Verifier pattern (builder + verifier)</strong></td><td>the-verifier-agent (your repo)</td><td>You built the production system</td></tr><tr><td><strong>Autoresearch loop + integrity guards</strong></td><td>mythos-learnings.md, brand-monitor</td><td>Based on Anthropic's own 244-page Mythos paper</td></tr><tr><td><strong>3x cost rule</strong></td><td>Original analysis</td><td>Synthesized from your aiproxy benchmarks</td></tr><tr><td><strong>CEO Board System (8 agents)</strong></td><td>ceo-agents (your repo)</td><td>You built the working system</td></tr><tr><td><strong>UI Agents (12 teams, brand hierarchy)</strong></td><td>ui-agents (your repo)</td><td>You built the working system</td></tr><tr><td><strong>TillDone task discipline</strong></td><td>pi-vs-claude-code/tilldone.ts</td><td>Extension you studied and documented</td></tr><tr><td><strong>Pi-to-Pi peer-to-peer coms</strong></td><td>pi-vs-claude-code/coms.ts</td><td>Novel P2P agent architecture</td></tr><tr><td><strong>Beyond MCP context cost matrix</strong></td><td>beyond-mcp/ (your repo)</td><td>Novel research comparing 4 approaches</td></tr><tr><td><strong>Mac Mini Agent (physical sandbox)</strong></td><td>mac-mini-agent (your repo)</td><td>Physical + cloud isolation pattern</td></tr><tr><td><strong>Codebase architectures (4 patterns)</strong></td><td>single-file-agents</td><td>From your architectural research</td></tr><tr><td><strong>Probability math (1%×100turns=63%)</strong></td><td>bash-damage-from-within</td><td>Novel risk quantification</td></tr><tr><td><strong>Always-on voice→CLI bridge</strong></td><td>always-on-ai-assistant</td><td>Voice-operated agent pattern</td></tr></tbody></table><h3 id="content-shared-with-tac-only-differentiator-vs-everyone-else" tabindex="-1">Content Shared With TAC Only (DIFFERENTIATOR vs everyone else) <a class="header-anchor" href="#content-shared-with-tac-only-differentiator-vs-everyone-else" aria-label="Permalink to "Content Shared With TAC Only (DIFFERENTIATOR vs everyone else)""></a></h3><table tabindex="0"><thead><tr><th>Topic</th><th>Your Coverage</th><th>TAC Coverage</th></tr></thead><tbody><tr><td>Depth-2 delegation</td><td>Full with config examples</td><td>Teaches the pattern</td></tr><tr><td>Agent chains (YAML pipelines)</td><td>Full with $INPUT/$ORIGINAL</td><td>Teaches the pattern</td></tr><tr><td>Agent mental models</td><td>Full with self-improve commands</td><td>Teaches the concept</td></tr><tr><td>Domain locking</td><td>Full permission matrix</td><td>Teaches the concept</td></tr><tr><td>Hook architecture</td><td>13 lifecycle events</td><td>Focuses on practical hooks</td></tr><tr><td>Meta-agents</td><td>Includes research expert pattern</td><td>Teaches the concept</td></tr></tbody></table><h3 id="content-commoditized-everyone-covers-it" tabindex="-1">Content Commoditized (Everyone covers it) <a class="header-anchor" href="#content-commoditized-everyone-covers-it" aria-label="Permalink to "Content Commoditized (Everyone covers it)""></a></h3><table tabindex="0"><thead><tr><th>Topic</th><th>Your Angle</th><th>What's Different?</th></tr></thead><tbody><tr><td>Agent loops</td><td>5-level progression</td><td>Your levels match Anthropic's "Building Effective Agents"</td></tr><tr><td>Tool calling</td><td>Reasoning parameter</td><td>Novel addition not in other courses</td></tr><tr><td>Context management</td><td>Hybrid approach</td><td>Standard approach</td></tr><tr><td>pass@k evaluation</td><td>Working harness code</td><td>Implementation matters more than concept</td></tr></tbody></table><hr><h2 id="price-positioning" tabindex="-1">Price Positioning <a class="header-anchor" href="#price-positioning" aria-label="Permalink to "Price Positioning""></a></h2><table tabindex="0"><thead><tr><th>Competitor</th><th>Price</th><th>Your Relative Value</th></tr></thead><tbody><tr><td>Anthropic Learn Harness</td><td><strong>Free</strong></td><td>You offer more depth in security, production, and economics</td></tr><tr><td>LangChain Academy</td><td><strong>Free</strong></td><td>You are framework-agnostic (LangChain is framework-specific)</td></tr><tr><td>DeepLearning.AI</td><td><strong>$45/mo</strong> (1-2hr per course)</td><td>Your course is 20-30x more content</td></tr><tr><td>IndyDevDan TAC</td><td><strong>$97-147</strong></td><td>You have all their content + 15 unique topics not in TAC</td></tr><tr><td>Maven bootcamps</td><td><strong>$200-500</strong></td><td>Cohort format with mentorship</td></tr><tr><td>Certification tracks</td><td><strong>$500+</strong></td><td>You'd need certification infrastructure</td></tr></tbody></table><hr><h2 id="market-positioning-recommendation" tabindex="-1">Market Positioning Recommendation <a class="header-anchor" href="#market-positioning-recommendation" aria-label="Permalink to "Market Positioning Recommendation""></a></h2><p>Based on this analysis, here's where your course sits in the market:</p><div class="language- vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang"></span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span> FRAMEWORK-SPECIFIC</span></span>
|
||
<span class="line"><span> │</span></span>
|
||
<span class="line"><span> $0 ││ LangChain Academy │</span></span>
|
||
<span class="line"><span> ││ │</span></span>
|
||
<span class="line"><span> ││ DL.AI short courses │</span></span>
|
||
<span class="line"><span> ││ │</span></span>
|
||
<span class="line"><span> $100 ││ TAC / Your Course │</span></span>
|
||
<span class="line"><span> ││ │</span></span>
|
||
<span class="line"><span> ││ │</span></span>
|
||
<span class="line"><span> $500 ││ Maven bootcamps │</span></span>
|
||
<span class="line"><span> ││ │</span></span>
|
||
<span class="line"><span> └──────────────────────────────</span></span>
|
||
<span class="line"><span> FRAMEWORK-AGNOSTIC</span></span></code></pre></div><p><strong>Your competitive moat</strong>: You are framework-agnostic (like Anthropic, unlike LangChain/DL.AI) AND you have hands-on labs (like TAC, unlike Anthropic). No other course occupies both quadrants.</p><p><strong>Unique selling points to emphasize</strong>:</p><ol><li>"From the engineer who reverse-engineered Anthropic's Mythos paper — learn what the frontier models CAN'T do"</li><li>"13 runnable labs with starter code AND solutions"</li><li>"The only course with a working verifier agent, bash security ladder, and autoresearch loop"</li><li>"Based on 25,000+ files of production agent systems — not tutorials, but battle scars"</li></ol><p><strong>Pricing recommendation</strong>: $97-147 positions you directly against TAC with MORE content. $197-247 positions you as premium (justified by original research + working systems source code).</p><hr><h2 id="validation-claudefast-articles-published-may-2026" tabindex="-1">Validation: ClaudeFAST Articles (Published May 2026) <a class="header-anchor" href="#validation-claudefast-articles-published-may-2026" aria-label="Permalink to "Validation: ClaudeFAST Articles (Published May 2026)""></a></h2><p>Fetched both articles to verify our course content against real published material.</p><h3 id="article-1-the-agent-manager-who-owns-claude-code" tabindex="-1">Article 1: "The Agent Manager: Who Owns Claude Code?" <a class="header-anchor" href="#article-1-the-agent-manager-who-owns-claude-code" aria-label="Permalink to "Article 1: "The Agent Manager: Who Owns Claude Code?"""></a></h3><p><strong>Published by</strong>: ClaudeFAST, citing Anthropic's May 2026 terminology<br><strong>Thesis</strong>: The Agent Manager role exists because enterprises need someone to own the harness.</p><p><strong>Their 3 failure modes without an agent manager</strong>:</p><ol><li><strong>Tribal knowledge</strong> — Every dev evolves a personal AI layer; nothing is shared</li><li><strong>Inconsistent results</strong> — Same model, same codebase, different output quality</li><li><strong>Security drift</strong> — Permissions configured ad hoc, MCP servers with unbounded scope</li></ol><p><strong>Their 5 areas of ownership</strong> (maps to the 5 harness subsystems):</p><ul><li>Maps to <strong>Instructions</strong> (CLAUDE.md, skills standardization)</li><li>Maps to <strong>Tools</strong> (MCP servers, plugin curation)</li><li>Maps to <strong>Environment</strong> (init scripts, workspace consistency)</li><li>Maps to <strong>State</strong> (session continuity, shared config)</li><li>Maps to <strong>Verification</strong> (hooks, quality gates)</li></ul><p><strong>Our coverage in M5</strong>: We added the Agent Manager role lesson (5.2c) with the 90-day playbook, plus the 5-tool production stack case study (5.2b). Both align exactly with ClaudeFAST's framing. ✅</p><p><strong>Grade</strong>: A — Our coverage matches the published standard.</p><h3 id="article-2-thread-based-engineering-scale-claude-code-sessions" tabindex="-1">Article 2: "Thread-Based Engineering: Scale Claude Code Sessions" <a class="header-anchor" href="#article-2-thread-based-engineering-scale-claude-code-sessions" aria-label="Permalink to "Article 2: "Thread-Based Engineering: Scale Claude Code Sessions"""></a></h3><p><strong>Published by</strong>: ClaudeFAST<br><strong>Thesis</strong>: 6 fundamental thread patterns that scale AI-assisted engineering work.</p><p><strong>Their 6 thread types vs our coverage</strong>:</p><table tabindex="0"><thead><tr><th>Thread</th><th>ClaudeFAST defines it as</th><th style="text-align:center;">Our Coverage</th><th style="text-align:center;">Grade</th></tr></thead><tbody><tr><td><strong>Base Thread</strong></td><td>Prompt → Tool Calls → Review</td><td style="text-align:center;">M1 agent loop</td><td style="text-align:center;">✅ A</td></tr><tr><td><strong>P-Thread</strong></td><td>Parallel instances (Boris runs 15)</td><td style="text-align:center;">M4, added this session</td><td style="text-align:center;">✅ A</td></tr><tr><td><strong>L-Thread</strong></td><td>Long-running, hours+</td><td style="text-align:center;">M7 always-on</td><td style="text-align:center;">✅ A</td></tr><tr><td><strong>B-Thread</strong></td><td>Agents managing agents</td><td style="text-align:center;">M4 delegation</td><td style="text-align:center;">✅ A</td></tr><tr><td><strong>F-Thread</strong></td><td>Fusion, N agents 1 winner</td><td style="text-align:center;">M4, added this session</td><td style="text-align:center;">✅ A</td></tr><tr><td><strong>C-Thread</strong></td><td>Checkpoint gates</td><td style="text-align:center;">M5 HITL</td><td style="text-align:center;">✅ A</td></tr></tbody></table><p>ClaudeFAST explicitly names Boris Cherny (creator of Claude Code) running 5 tmux tabs + 5-10 web instances = 10-15 parallel P-threads. This matches HypeMan's psmux + mprocs stack exactly.</p><p><strong>Our coverage in M4</strong>: All 6 thread types covered. ✅</p><h3 id="what-this-validates" tabindex="-1">What This Validates <a class="header-anchor" href="#what-this-validates" aria-label="Permalink to "What This Validates""></a></h3><ol><li><strong>REFERENCE-STACK.md is production-accurate</strong> — Your stack mirrors Boris Cherny's own setup and ClaudeFAST's documented patterns.</li><li><strong>Agent Manager role is real</strong> — Anthropic published the terminology May 2026. Our M5 lesson matches.</li><li><strong>Thread-based engineering is the standard</strong> — Both ClaudeFAST and IndyDevDan teach it. We cover all 6 types.</li><li><strong>Your stack is ahead of the course</strong> — HypeMan runs exactly what ClaudeFAST teaches. REFERENCE-STACK.md documents it.</li></ol><hr><h2 id="live-comparison-indydevdan-s-published-content-agenticengineer-com" tabindex="-1">Live Comparison: IndyDevDan's Published Content (agenticengineer.com) <a class="header-anchor" href="#live-comparison-indydevdan-s-published-content-agenticengineer-com" aria-label="Permalink to "Live Comparison: IndyDevDan's Published Content (agenticengineer.com)""></a></h2><p>I fetched 5 pages from his site. Here's his published intellectual property and what it means for your course.</p><h3 id="page-1-the-only-claude-code-competitor" tabindex="-1">Page 1: "The Only Claude Code Competitor" <a class="header-anchor" href="#page-1-the-only-claude-code-competitor" aria-label="Permalink to "Page 1: "The Only Claude Code Competitor"""></a></h3><p><strong>Core framework</strong>: 4 dimensions of agent control — <strong>context, model, prompt, tools</strong><strong>3-tier customization ladder</strong>:</p><ul><li>Tier 1: Basic harness customization (settings, skills)</li><li>Tier 2: Hooks, programmatic control, distribution</li><li>Tier 3: Multi-agent orchestration</li></ul><p><strong>Thesis</strong>: "Claude Code is the starter pack. Pi is the endgame." He positions Claude Code for beginners (first 100 hours) and Pi for advanced users who need harness control.</p><p><strong>Course gap analysis</strong>: Your M1-M4 already cover all 3 tiers. His 4-dimensions framing (context, model, prompt, tools) is cleaner than your current 4 pillars. <strong>Consider adopting his framing language.</strong></p><h3 id="page-2-top-2-agentic-engineering-—-10-bets-for-2026" tabindex="-1">Page 2: "Top 2% Agentic Engineering" — 10 Bets for 2026 <a class="header-anchor" href="#page-2-top-2-agentic-engineering-—-10-bets-for-2026" aria-label="Permalink to "Page 2: "Top 2% Agentic Engineering" — 10 Bets for 2026""></a></h3><table tabindex="0"><thead><tr><th>Bet #</th><th>Topic</th><th style="text-align:center;">Covered in Your Course?</th></tr></thead><tbody><tr><td>1</td><td>Anthropic becomes a monster (ecosystem moat)</td><td style="text-align:center;">Not covered as a thesis</td></tr><tr><td>2</td><td>Tool calling is the opportunity</td><td style="text-align:center;">Covered in M2</td></tr><tr><td>3</td><td>Custom agents above all</td><td style="text-align:center;">Covered in M2, M4</td></tr><tr><td>4</td><td>Multi-agent orchestration</td><td style="text-align:center;">Covered in M4</td></tr><tr><td>5</td><td>Agent sandboxes</td><td style="text-align:center;">Covered in M7</td></tr><tr><td>6</td><td>In-loop vs out-loop agentic coding</td><td style="text-align:center;">Not framed this way</td></tr><tr><td>7</td><td>Agentic Coding 2.0 (agents conducting agents)</td><td style="text-align:center;">Covered in M4, M7</td></tr><tr><td>8</td><td>The benchmark breakdown (skepticism)</td><td style="text-align:center;">Not covered</td></tr><tr><td>9</td><td>Agents eating software (market trend)</td><td style="text-align:center;">Not covered</td></tr><tr><td>10</td><td>Agent-native architecture</td><td style="text-align:center;">Partial in M4</td></tr></tbody></table><p><strong>Gap</strong>: Bets 1, 6, 8, 9 are market/intellectual framing you don't address. They're opinion/positioning pieces.</p><h3 id="page-3-thinking-in-threads-—-his-signature-framework" tabindex="-1">Page 3: "Thinking in Threads" — His Signature Framework <a class="header-anchor" href="#page-3-thinking-in-threads-—-his-signature-framework" aria-label="Permalink to "Page 3: "Thinking in Threads" — His Signature Framework""></a></h3><table tabindex="0"><thead><tr><th>Thread</th><th>What It Is</th><th style="text-align:center;">In Your Course?</th></tr></thead><tbody><tr><td><strong>Base Thread</strong></td><td>Prompt → Tool Calls → Review</td><td style="text-align:center;">M1 agent loop</td></tr><tr><td><strong>P-Thread</strong></td><td>Run N agents in parallel</td><td style="text-align:center;">Not covered</td></tr><tr><td><strong>C-Thread</strong></td><td>Checkpoints, human gates</td><td style="text-align:center;">M5 (HITL)</td></tr><tr><td><strong>F-Thread</strong></td><td>Fusion: N agents, 1 winner</td><td style="text-align:center;">Not covered</td></tr><tr><td><strong>B-Thread</strong></td><td>Branch: agents manage agents</td><td style="text-align:center;">M4 delegation</td></tr><tr><td><strong>L-Thread</strong></td><td>Long-running (hours/days)</td><td style="text-align:center;">M7 always-on</td></tr><tr><td><strong>Z-Thread</strong></td><td>atomic prompt→tools→ship</td><td style="text-align:center;">M1 basic loop</td></tr></tbody></table><p><strong>Gap</strong>: P-threads (parallel) and F-threads (fusion) are missing. These are his most distinctive frameworks.</p><h3 id="page-4-engineering-with-exponentials" tabindex="-1">Page 4: "Engineering with Exponentials" <a class="header-anchor" href="#page-4-engineering-with-exponentials" aria-label="Permalink to "Page 4: "Engineering with Exponentials"""></a></h3><p><strong>Thesis</strong>: "The prompt is the new fundamental unit of knowledge work programming." <strong>3-part essay series</strong>: (1) Engineering with Exponentials, (2) AI Coding is Transitory, (3) Agentic Coding is the Endgame. <strong>No direct technical gaps</strong>. This is thought-leadership positioning.</p><h3 id="page-5-compute-advantage-equation" tabindex="-1">Page 5: "Compute Advantage Equation" <a class="header-anchor" href="#page-5-compute-advantage-equation" aria-label="Permalink to "Page 5: "Compute Advantage Equation"""></a></h3><p><strong>Formula</strong>: <code>(Compute Scaling × Autonomy) ÷ (Time + Effort + Monetary Cost)</code><strong>Purpose</strong>: Interactive calculator to compare AI coding tools. <strong>Gap</strong>: You don't have a unified "value equation" for agentic engineering.</p><h3 id="summary-his-ip-vs-your-coverage" tabindex="-1">Summary: His IP vs Your Coverage <a class="header-anchor" href="#summary-his-ip-vs-your-coverage" aria-label="Permalink to "Summary: His IP vs Your Coverage""></a></h3><p><strong>His unique IP you should reference or adopt</strong>:</p><ol><li><strong>4 dimensions of control</strong> (context, model, prompt, tools) — cleaner framing for M1</li><li><strong>Thread framework</strong> (P-thread, F-thread, C-thread) — add as orchestration patterns in M4</li><li><strong>Compute Advantage Equation</strong> — add a version in M6 (economics)</li><li><strong>"Do you trust your agents?"</strong> — use this framing hook in M1</li></ol><p><strong>His gaps that you fill (your competitive moat)</strong>:</p><ul><li>He has NO security module (your M3 crushes him)</li><li>He has NO production deployment module (your M5)</li><li>He has NO model economics module (your M6)</li><li>He has NO autoresearch module (your M7)</li><li>He has NO scaffolded labs with solutions (your 13 labs)</li><li>He has NO non-technical track (your NON-TECHNICAL.md)</li><li>He has NO assessment/quizzes (your ASSESSMENTS.md)</li><li>He has NO verifier agent pattern (your M3)</li><li>He has NO CEO Board System (your M4)</li><li>He has NO bash security ladder (your M3)</li></ul><p><strong>Verdict</strong>: Dan is a <strong>thought leader</strong> with strong mental models (threads, 4 dimensions, compute advantage). You are a <strong>curriculum builder</strong> with more comprehensive technical coverage. The ideal course combines both — his mental models + your technical depth.</p>`,117)])])}const p=e(d,[["render",n]]);export{u as __pageData,p as default};
|