AI Made Me Smarter—and Slower to Meet Reality
AI can rapidly complete concepts, architectures, and plans while delaying falsifiable feedback. A 48-hour task shows how to let reality answer sooner.
Xinwei Xiong (cubxxw) · An Open Archive of Practice and Thought
Xinwei Xiong's open archive of AI workflows, real projects, engineering methods, systems thinking, personal growth, philosophy, and life.
Latest posts
Claude Tag Deep Dive: From Shared Slack AI to an Organizational Agent Runtime
Claude Tag turns Slack into a shared agent runtime with identity, memory, and tools. This deep dive covers its architecture, risks, Chinese peers, and future.
I Don't Have an Execution Problem. I Need a Battlefield That Can Say No
High output disproved my diagnosis of weak execution. What I lacked was an external result with the authority to reorder, stop, renew, or redirect the work.
Reading Hongyi in Quanzhou: Seeing Through, Then Taking Life Seriously
In Quanzhou, Master Hongyi’s Vinaya practice and a Buddhist temple’s Hindu pillars changed how I understand play, commitment, and the choice to keep building.
Common Sense Is Not the Mechanism: Rebuilding Industries Around Demand, Constraints, and Scarcity
A framework for testing AI industry ideas: start with durable demand, map constraints and scarcity, redesign the workflow, and validate it with real failures.
The Scoreboard You Choose Will Eventually Choose You: How Incentives Reshape an Organization
Metrics do more than measure organizations: they train attention and ethics. This essay shows founders how to audit incentives before they become culture.
The Opposite of Freedom Isn’t Constraint—It’s the Defaults That Live Your Day for You
Freedom is not the absence of schedules; defaults can quietly live the day for us. Three deliberate choice points help reclaim agency without rigid control.
A Second Brain Preserves Your Decision Function
A second brain preserves more than notes: over time it exposes the decision function behind what we protect, postpone, commit to, and let reality revise.
When Generation Becomes Infinite, Authentic Experiences Begin to Command Value
AI can generate endless content but cannot bear experience’s cost. Authentic creation therefore depends on embodiment, verification, and ethical boundaries.
Risk Preference Isn't About Enjoying Danger: What Truly Enables Taking a Bet
Cave danger and social silence are different risks. This essay reframes courage around ambiguity, exposure, and repeated bets so no rejection can define you.
The Cost of Observation: Why the More You Understand, the Harder It Is to Enter Life
Observation protects us through distance, but can keep love and action from beginning. A four-step rhythm reconnects clear judgment with lived contact.
Creators Don’t Lack Applause—They Lack an Accurate Witness
Creators need more than applause: they need precise evidence that a work arrived. Four kinds of readers reveal what praise, critique, and silence cannot.
Forgetting Is Harder Than Remembering: Why a Living System Must Know When to Let Go
Cheap storage makes forgetting a core AI capability. This essay designs auditable decay, archiving, and revocation for long-running AI memory systems.
Records Are Not Mirrors, But Scissors: What Tool You Use Ultimately Shapes Who You Become
Note-taking tools do not merely preserve experience; they edit attention and identity. A two-stage practice keeps facts separate from later interpretation.
Environment as the Hidden Author: Why Moving Cities Feels Like Becoming Someone Else
The same person changes across cities. This essay shows how environments shape behavior, while preserving responsibility and room to redesign daily life.
When Distance No Longer Changes You: What Keeps Life Moving After Exploration Fails
When novelty fades, travel must be judged by the relationships and obligations that still survive departure—not by distance, shock, or cities collected.
Every Path Has a Price
Every path charges in uncertainty, repetition, lost options, and identity. Freedom is not finding a free road, but choosing a price worth paying clearly.
After AI Solves a Problem, Where Does the Bottleneck Move?
AI makes production abundant, not unconstrained. Bottlenecks move to review, trust, distribution, and delivery; durable advantage follows the learning loop.
The Fastest Numbers Are Often Furthest from the Outcome
Views, clicks, and sales arrive quickly; refunds, trust, and cash flow take time. Fast metrics should guide learning, while slow outcomes judge the game.
Action Does Not Automatically Lead to Truth
“Think less, do more” can cure armchair thinking—or disguise haste as courage. Good action still tests rival explanations while keeping failure survivable.
The Escape Routes of Smart People: Applying Ability to the Wrong Problem
Procrastination can look like intense research and polished systems. Competence becomes an escape when it solves controllable problems instead of decisive ones.
What Does Profit Actually Prove?
Profit proves that exchange occurred and an organization may survive. It cannot prove who bore the hidden costs, who truly benefited, or what we become.
What Should I Do Next? A Debate About Money, Action, and Ideals
A dinner-table argument about profit forced me to examine how vision, research, and system-building can become elegant ways to avoid the verdict of reality.
Bear Quest — Play My Story
A small side-scrolling journey through Xinwei Xiong's life, travels, open-source work, products, and questions.
Open Design 0.16.1: A Design Harness for Coding Agents
A version-pinned guide to Open Design 0.16.1: its four-plane architecture, local and cloud paths, BYOK privacy boundaries, SSRF controls, and practical fit.
Claude Code Playbook: 10 Configurations for Reliable Agent Workflows
A practical Claude Code guide to verification, permissions, worktrees, subagents, workflows, goals, loops, and routines, with current limits and safe defaults.
Loop Engineering for Solo Builders: Verifiers, State, and Safe Automation
Loop engineering turns repeated prompting into a verifiable system. A field guide to state, layered checks, permissions, and clear stopping rules for solo work.
You Think You're the Player — But "Player" Is Also a Role: Rereading the Samādhi of Play
'Treat life as a game you're playing.' I wrote that line almost two years ago. Then I went looking for where the phrase 游戏三昧 — the samādhi of play — actually comes from, and found I may have had it backwards. The Zen 游 is glossed as unobstructed freedom and traces to Zhuangzi; modern gamification traces to the Skinner box. Worse: the thought 'I am the player' quietly reinstalls a watcher behind…
A Solo Creator's AI Video Editing Pipeline That Actually Works
A practical AI video editing pipeline for solo creators: structure shoots, automate rough cuts, keep judgment human, and publish safely across platforms.
When Anyone Can Build the Thing, "He Built It" Becomes the Signal
When the cost of building approaches zero, having built something no longer proves anything, so people go look at who built it. This essay covers the first layer of the gear stack — reputation: why trust is the one thing AI cannot flatten, the three carriers it lives on, why it will never appear on any tool list, and its three most uncomfortable truths. It closes by answering the question the…
Distribution for Solo Builders: Your Readers Can't Be Copied
A distribution strategy for solo builders: turn product work into durable content, choose channels deliberately, and measure the audience you actually own.
AI Writes Requirements for Free, and That's the Most Dangerous Part
When execution cost approaches zero, the price of building the wrong thing doesn't shrink — it becomes payable in full. AI generates requirements at near-zero cost, so your queue fills with tasks that look reasonable and deliver little, and agents will faithfully complete every one. This essay takes apart the three jobs of the judgment layer: how to guard the requirement gate, how to write…
I Ran Ten Agents Overnight, Woke Up to Ten PRs, and Then I Got Stuck
Part two of 'The Super Individual's Gear Stack,' on the bottom layer — production. Every advance in this layer is a public good, handed to everyone at the same moment, so once it's good enough you should stop investing. What actually blocks you isn't execution bandwidth but review bandwidth, which cannot be scaled the same way. The only lever that works is raising first-pass correctness —…
The Four-Layer AI Stack for Super Individuals: What Actually Compounds
A four-layer AI stack for super individuals: separate production tools from judgment, distribution, and reputation assets that compound over time in practice.
Designing devbox-doctor: A Safer Mac Toolchain Audit
A design for auditing Mac developer tools with read-only evidence, bounded AI judgments, privacy safeguards, graceful fallbacks, and false-positive tests.
Agent Skill Design: What a Dangerous SKILL.md Taught Me
Agent Skill and SKILL.md design through a storage-cleanup case study: architecture, permission boundaries, prompt contracts, real-machine tests, and lessons.
Building an AI Knowledge Base: From Notes to Executable Workflows
Build an evidence-driven AI knowledge base with executable workflows, method cards, voice rules, review gates, and feedback loops that improve content work.
How AI Agents Rebuilt My 120-Post Blog From Scratch
I rebuilt a 120-post Hugo blog with AI agents, human review, MCP, GitHub Actions, a cover pipeline, and a measured GEO experiment. Here is what held up.
Where AI Agents Still Have a Blue Ocean
A practical map of defensible AI agent businesses: vertical workflows, accountable delivery, agent infrastructure, and moats that survive model progress.
How to Build Real Trust in Unattended AI Agents That Act
A practical trust stack for unattended AI agents: hard tool controls, regression evals, checkpoints, rollback, and focused human review for risky actions.
Build a Personal Intelligence System That Leads to Action
Build a personal intelligence system that turns monitored sources into decisions, with honest boundaries for RSS, RSSHub, agents, MCP, safety, and evaluation.
Agent Fleet Economics in 2026: Testing Low-Cost APIs and Open-Weight Options
A dated, reproducible agent-fleet cost model that tests low-cost APIs, weighs open-weight options, and routes work by measured success and operational risk.
When the AI Agent Starts Prompting You, What Has Actually Changed
A practical framework for proactive AI agents: memory, runtimes, triggers, interruption budgets, human approval, and the discipline to stay quiet by default.
AI News Pipelines: Automation Limits and Human Judgment
AI news pipelines can collect, translate, summarize, and deduplicate at scale. This field-tested guide shows where automation ends and human judgment begins.
What Remains When You Take It All Apart: On No-Self, Suffering, and "You Could Say It Either Way"
Starting from a single image — a shaft of sunlight passing through an eye clouded by cataracts — this piece dismantles the self step by step: no-self, the five aggregates, dependent origination, until the sheer relativity of concepts makes you dizzy: 'You can say it either way, so what's left to reach for?' It refuses to hand you an answer. Instead it argues that relativity is a property of…
From Chatbot to Agent to Skill: Turning Judgment into a System
A practical Chatbot-to-Agent-to-Skill framework for turning domain judgment into reusable workflows, with Amazon Ads boundaries, metrics, and human approval.
Give AI Tasks, Not Just Direction: Define Done First
Explore direction with AI, then define the finish line yourself. A six-part task card turns open-ended conversation into work you can test, review, and trust.
Friction Is Growth: When AI Removes All Resistance for You, Deliberately Keep Some
AI's core selling point is removing friction — faster, less effort, painless. But almost all human growth comes from friction: forcing the vague into the clear, working through what you don't yet know, being jolted by an objection. This essay argues that in an era where everything can be painless, learning to tell which friction to remove and which to deliberately keep is becoming a new form of…
AI Workflow Quality Gates: A Practical Engineering Guide
A practical framework for reliable AI workflows: separate claims from logs, preserve evidence, run evals, set failure thresholds, and escalate risk to humans.
Build an AI Second Brain with Claude and Obsidian
Build a safe AI second brain with Obsidian and Claude: structure your vault, isolate sensitive notes, preview changes, back up data, test, and roll back.