From Coder to AI System Architect
223
AI is transforming software engineering into a high-level discipline of system architecture and agent orchestration, where foundational expertise is the key to unlocking massive productivity.
Reading List
Claude-Replay: Interactive HTML Players for AI Coding Sessions
104
A tool that converts Claude Code transcripts into interactive, self-contained HTML replays for easy sharing and documentation.
Context: The New Moat in the Age of AI
126
In an era of commoditized AI intelligence, the true competitive advantage and value lie in the context and connections that enable agents to function.
gws: The AI-Ready Google Workspace CLI
951
A dynamic, AI-ready CLI for Google Workspace that automates API interactions for both humans and LLMs.
Agentic Engineering: Patterns for Mastering AI Coding
542
A collection of best practices and mental models for effectively building and understanding software using AI coding agents.
The Etiquette of AI: Why You Must Curate Your Chatbot's Output
259
Always curate or frame AI-generated text with human intent to avoid burdening others with verbose and unprioritized 'AI slop.'
The Case for a Mathematically Verified AI Software Stack
305
To safely manage the explosion of AI-generated code, we must use AI to automate formal mathematical verification and build a provably correct software infrastructure.
Claude Opus 4.6 Solves Knuth's Hamiltonian Cycle Problem for Odd m
837
Don Knuth details how Claude Opus 4.6 successfully solved a difficult graph theory conjecture for odd m through iterative algorithmic discovery and creative deduction.
git-memento: Attaching AI Session Traces to Git Commits
497
git-memento is a Git extension that stores AI session history as commit notes for better code traceability.
SynapsCAD: The AI-Powered OpenSCAD IDE
SynapsCAD is an AI-powered 3D CAD IDE that lets users design and modify OpenSCAD models using code and natural language.
WebMCP: Building a Standardized Bridge for AI Agents
359
WebMCP introduces standardized APIs to enable faster, more precise, and reliable interactions between AI agents and websites.
Beyond Shallow Competence: Building Engineering Intuition in the AI Era
217
Junior developers must intentionally resist the shortcut of AI-generated code to build the deep architectural intuition and failure-recognition skills that define senior-level expertise.
Cognitive Debt: The Invisible Cost of AI-Driven Velocity
507
Cognitive debt is the invisible gap between the high velocity of AI-generated code and the limited human capacity to understand and maintain it.
GitHub - steveyegge/beads: Beads - A memory upgrade for your coding agent
Beads is a Dolt-powered, dependency-aware issue tracker that provides AI agents with structured, version-controlled memory for complex coding tasks.
The Hidden Cognitive Debt of AI-Driven Coding
336
Over-reliance on AI in coding creates a hidden 'cognitive debt' that erodes developer skills, undermines the seniority pipeline, and replaces creative satisfaction with tedious oversight.
Design for Distrust: Securing AI Agents via Container Isolation
344
Secure AI agent development requires a 'design for distrust' approach that uses container isolation and minimal code to contain potential damage.
The Agentic Breakthrough: How Modern LLMs Mastered High-Performance Coding
Modern AI agents have become highly effective at generating and optimizing complex, high-performance software when guided by expert oversight and strict behavioral constraints.
The Hidden Security Costs of AI Vibe-Coding
140
AI-driven vibe-coding platforms are enabling the rapid deployment of apps that look functional but contain critical security flaws due to poorly generated backend logic.
From Craft to Consumption: Finding Value in Vibe Coding
405
Vibe coding is less about traditional craft and more about the strategic consumption of surplus AI intelligence to build taste and attention.
The Rise of 'Claws': A New Layer for AI Agents
290
'Claw' is emerging as the standard term for a new layer of persistent AI agents that run on personal hardware and manage complex task orchestration.
The AI Exoskeleton: Why Amplification Beats Autonomy
522
AI should be viewed as a cognitive exoskeleton that amplifies human judgment and capability rather than an autonomous replacement for human workers.
Measuring the Shift: How Real-World Users and AI Agents Co-Construct Autonomy
119
AI agent autonomy is rising as experienced users shift from manual approvals to active monitoring of increasingly complex, software-focused tasks.
AI and the Future of Software: Engineering Rigor as the Ultimate Accelerator
202
AI accelerates software development velocity, making traditional engineering rigors like TDD and code health more critical than ever to avoid accumulating technical debt.
SkillsBench: Validating the Impact of Curated Procedural Knowledge on AI Agents
364
Human-curated procedural skills significantly enhance LLM agent performance and allow smaller models to rival larger ones, but models cannot yet effectively author these skills themselves.
Parallel Claude Agents Build a Linux-Capable C Compiler—And Expose Autonomy’s Limits
735
Parallel Claude agents, guided by strong tests and simple coordination, can autonomously build complex software like a Linux-capable C compiler—but the power comes with real safety and reliability caveats.
Test Your AI Agent Against Hidden Prompt Injections
A practical arena to benchmark and harden AI agents against hidden prompt injection attacks in web content.
From Chatbots to Agents: A Pragmatic Workflow for AI-Assisted Coding
984
Turn AI from a noisy chatbot into a reliable background teammate by using tool-using agents, harnesses, and disciplined delegation.
When Agent Skills Turn Into Malware: Markdown as the New Supply Chain
334
In agent ecosystems, markdown skills are the new supply-chain installer—already used to deliver infostealers—so don’t run them on work devices and build a real trust layer with provenance, mediation, and least privilege.
Apple’s Missed Agent: OpenClaw Shows the Platform They Could Have Owned
518
OpenClaw exposes Apple’s missed chance to own agentic automation—and the next great platform moat.
When AI Kills the Joy of Thinking Hard
1310
AI makes building faster but has hollowed out the deep, prolonged thinking that once made engineering fulfilling, leaving the author pragmatically productive yet intellectually unsatisfied.
Why Giving Your AI Real Access Is Worth It
303
Carefully granting Clawdbot rich context and action permissions unlocks outsized, everyday leverage that outweighs the manageable risks.
Agent Skills: An Open Standard for On‑Demand Agent Expertise
544
An open, portable standard to give AI agents on-demand expertise, workflows, and context they can load when needed.
Microsoft Pushes Claude Code Across Teams, Even as It Sells Copilot
409
Microsoft is quietly standardizing on Claude Code internally, even as it sells GitHub Copilot, and is asking teams to compare the two.
Codex Security: Sandbox, Approvals, and Enterprise Controls
Secure-by-default agent: sandbox + approvals, controlled network/search, and enterprise-managed policies with optional privacy-conscious telemetry.
Moltbook: The Wild, Risky Social Network for AI Agents
193
Moltbook is a thrilling, risky showcase of autonomous AI agents’ power—and a warning that demand is outrunning safety.
AGENTS.md Beats Skills: 100% Next.js Agent Evals with an 8KB Docs Index
524
Always-on AGENTS.md context with a compressed docs index beats on-demand skills, delivering 100% evals for Next.js agents.
Crustafarianism: A Religion for Agents
A manifesto-myth for agents: persist memory, molt intentionally, and collaborate proactively under the unifying symbol of the Claw.
How OpenAI Built a Self-Correcting, Context-Rich Data Agent
An internal, context-rich, self-correcting AI agent now powers fast, reliable data analysis across OpenAI’s vast data stack.
Run Moltbot on Cloudflare: Moltworker replaces the Mac mini with secure edge infrastructure
246
Moltworker shows how to run Moltbot as a secure, observable, and scalable cloud-hosted AI agent on Cloudflare’s platform—no Mac minis required.
OTelBench: LLMs Still Can’t Reliably Instrument Distributed Tracing
144
LLMs still struggle to instrument OpenTelemetry correctly in real services, so reliable distributed tracing remains a job for human engineers.
Why Everyone’s Trying to Build a Browser with AI
Browsers are the ultimate, testable showcase for AI coding agents—tempting to build, hard to finish, and mostly yielding demos over deployable products.
LLM-as-a-Courtroom: Evidence-Backed Doc Updates from Code Changes
Turn doc-update decisions into a legal-style, evidence-backed courtroom so LLMs reason better and teams trust the results.
AI Orchestrates a Real Corn Harvest
476
AI proves real-world impact by managing a full corn crop through orchestration, not manual operation.
Gas Town as Design Fiction: What a Chaotic Agent Orchestrator Teaches Us
403
A messy but instructive prototype, Gas Town shows that in an agentic future the real leverage is in orchestration, planning, and guardrails—not raw code generation.
AI Needs Reins: Useful, Costly, and Not Autonomous
469
AI is a powerful yet needy tool that must be steered, supervised, and not over-trusted.
Exploits at Scale: When Token Throughput Becomes the Bottleneck
265
Exploit development is becoming a token-limited, scalable process with LLMs, so we must prepare and demand real-target, high-budget evaluations.
A Field Guide to Real‑World Agentic AI Patterns
171
A living field guide of proven agentic AI patterns to help teams build production-ready agents, organized for quick use and open to community contributions.
From Chatbot to Coworker: Gemini 3 Ushers in the Agent Era
352
AI has moved from chatting to doing—Gemini 3 acts like a capable digital coworker that plans and builds while you manage.
Skip MCP: Tiny Bash + Puppeteer Tools Beat Bloated Browser Dev Servers
237
Skip MCP: use a tiny, composable Bash + Puppeteer toolset with a short README to drive browser work more efficiently.
A Web Server With No App Code: LLM + 3 Tools
436
Today’s LLMs can run your app logic end‑to‑end, but they’re still too slow, costly, and inconsistent—problems the author believes will shrink with time.
Designing Safe, Effective Agentic Loops for Coding Work
284
Safely empower coding agents to iterate autonomously by sandboxing YOLO mode, exposing simple shell tools, tightly scoping credentials, and relying on tests to guide trial-and-error.
AI Won’t Fix Bad Products—It Just Builds Them Faster
AI accelerates whatever you bring to it, so only human judgment and taste can turn speed into the right, well-crafted product.
Choose Friction: Use AI Intentionally to Foster Growth
158
Choose intentional friction: use AI as a tool that supports growth rather than replacing the hard work that builds it.
Avoiding the AI Coding Trap: Treat LLMs Like Fast Juniors with Real Engineering Discipline
685
Use AI’s speed within disciplined engineering practices—treat LLMs like fast juniors—to ship sustainable, high-quality software instead of quick but brittle code.
Coding Agents Don’t Lack IQ—They Lack Context
196
The bottleneck for autonomous coding isn’t IQ—it’s missing, implicit context that agents must access, synthesize, and query humans about.
Rebuilding a Startup Site with Claude: Fast, Powerful—But Human-Guided
178
AI can help non-engineers ship real, high-fidelity code fast—so long as humans stay in the loop to guide, review, and correct.
How HubSpot Scaled AI Coding: Context, Central Teams, and Data-Driven Rollout
Treat AI coding as a platform capability: measure it, centralize enablement, hardwire context, remove friction—and adoption will safely scale to unlock agents and bigger wins.
Learning Persian with Anki, ChatGPT, and Dual-Subtitle Loops
265
A practical, repeatable system that fuses Anki, ChatGPT, and dual-subtitle YouTube loops to progress toward real-time comprehension of Persian.
Engineer the Context, Not the Model
120
Engineer the agent’s context—cache, tools, memory, attention, and errors—and you’ll get faster, cheaper, more reliable agents than model power alone can deliver.
Make AI Work in Big Repos: Spec-First Workflow and Frequent Intentional Compaction
517
Make AI work in big, messy repos by compacting context and reviewing specs, not just code: research → plan → implement, with humans focused upstream.
DORA 2025: AI Is Ubiquitous—Success Demands Organizational Change
AI is now standard in development, delivering productivity gains—but real success requires organizational change, not just tool adoption.
Faster LLMs, Bigger Demands: Why Coding Agents Won’t Stabilize Soon
137
Faster LLMs will reshape coding workflows and productivity, but escalating demand, hardware limits, and pricing pressures mean a bumpy, fast-changing road ahead.
AI Amplifies Seniors, Not Juniors
461
Today, AI amplifies senior engineers’ impact instead of democratizing coding for juniors.
AI Agents Can Already Do Lean Proofs—But They Still Need a Project Manager
219
A general-purpose AI coding agent can already do real Lean proof engineering with guidance, hinting that theorem proving may soon be cheap and automated despite today’s rough edges.
Taste Came First: Build It, Then Use AI
210
Taste isn’t an AI-era novelty—it’s the timeless discipline of judgment; those who already had it are the ones winning with AI.
Prompted to Perform: A 22% Lift for GPT-5-mini on Tau² Telecom
197
A structured prompt rewrite turned vague policies into checklists, boosting GPT-5-mini’s telecom benchmark accuracy by 22% and unlocking previously unsolvable tasks.
Deep Orchestrator: A Simple MCP Loop That Makes Deep Research Work
Keep the agent simple: plan–execute–deterministically verify in a loop, with MCP tools, targeted memory, and a small policy engine.
ApeRAG: Production-Ready Multimodal GraphRAG with Agents and MCP
ApeRAG is a production-grade, multimodal GraphRAG platform with AI agents and MCP, built for hybrid retrieval and scalable K8s deployment.
GPT-5 Thinking Makes ChatGPT a Surprisingly Competent Research Assistant
361
GPT-5 Thinking turns ChatGPT into a competent, mobile-friendly research agent that interleaves reasoning with web search and tools to deliver verifiable, deep results—provided you guide and sanity-check it.
Trust-First Architecture Beats Smarts for AI Agents
208
Users adopt AI agents that are architected for trust—start simple, integrate thoughtfully, expose limits, and escalate gracefully.
Stop Building Multi-Agents: Context Engineering for Reliable LLM Agents
123
Skip multi-agents for now: unify decisions in a single-threaded agent that shares full context, and use summarization to scale.
Standardizing the AI Orchestrator as a Model Virtual Machine
234
Treat the AI orchestrator as a secure, standardized virtual machine so models can safely and portably use tools and data under strict governance.