Agentic Era
AI moves from tool to colleague. Autonomous agents, the future of work, and creativity within everyone's reach.
30 of 30 articles
2026. 03. 28
An Agent Team’s Trust Budget: Operational Design to Reduce Escalation Latency
As automation expands, recovery of trust matters more than performance alone. Managing escalation latency prevents small incidents from becoming widespread distrust.
2026. 01. 04
The Democratization of Creativity, or the End of Art?
An age in which everyone can become an artist may paradoxically be one in which nobody can.
2026. 01. 07
From Tools to Colleagues: The Rise of AI Agents
AI is no longer content simply to answer our questions. Now agents set goals, execute them and create on their own.
2026. 03. 02
Operating an Agent Team: From One Agent to Multi-Agent Orchestration
A team with separated roles delivers results more reliably than a capable agent working alone.
2026. 03. 05
The Real Standard for Multi-Agent Operations: Failure Budgets and Approval Gates
Agent-team performance is determined more by operating rules than model performance. The key is allowing failures while controlling losses.
2026. 03. 05
Agentic Era CH6. You Cannot Protect What You Cannot See: Runtime Observability and Safe Operations
Agent systems need visibility before performance. Bringing logs, traces and policy events onto one screen reduces incidents and builds trust.
2026. 03. 06
Agentic Era CH7. Hand Off Without Stopping: Human Handoff and Escalation Design
Good agents do not handle everything alone. Operational quality comes from a loop that detects failure quickly, hands it to people precisely, and returns to automation.
2026. 03. 09
Memory Is Harder to Retrieve Than Store: Designing Agent Memory Layers
Agent performance depends more on retrieving the right context at the right time than on model size. Well-designed memory layers improve cost, quality, and operational stability together.
2026. 03. 12
Good Agents Filter First: Context Firewalls and Policy Routing
Most performance degradation and dangerous automation arise from disordered context intake rather than inadequate models. Context firewalls and policy routing can improve quality, speed and safety together.
2026. 03. 15
Automation Is a Contract, Not a Permission: Operator Agreements by Autonomy Layer
As agent automation expands, operator agreements matter more than model performance. Separate autonomy layers and specify authority, verification, and recovery responsibilities in each to gain speed and safety together.
2026. 03. 18
A Turning Point in Agent Operations: Change Control Windows for Runtime Changes
When models, prompts and tool configurations change daily, stability requires designing the timing and validation windows for changes rather than preventing change itself.
2026-10-23
Five Checks Before Connecting an MCP Server
Before you connect an MCP server to an AI app, check who runs it, which tools it opens, what data leaves, how far permissions reach and whether approval and records remain. This guide orders those checks from the official MCP documents and OpenAI's guide.
2026-10-16
Five Steps to Start AI Automation
This general introduction to AI automation, current as of October 14, 2026, explains how to start with small tasks and check results, while noting that each tool’s features and pricing must be rechecked on its official website before use.
2026-10-16
Check Rules in Your AI Brand Folder
A fictional-brand exercise shows how to connect a brand folder and use a check question to assess whether AI has read the rules, without treating this as evidence of business results.
2026-10-16
AI Roles Without Invented Metrics
A fictional-brand exercise defines two AI employees in four sections, establishes a rule for marking missing metrics as no data, and distinguishes role creation from work actually performed.
2026-10-16
Seven Aside Steps From Tables to Phone Instructions
Seven steps show how to start with a verified read-only task in the Aside AI browser, then expand to skills, routines, and remote instructions while keeping their limits in view.
2026-10-16
Six Steps for a Claude Code Website
A practice project featuring a fictional house shows how to separate planning, rules, scene creation, and visual checks when building a 3D website.
2026-10-16
Three Checks After Delegating to dots
A fictional-brand exercise shows how to check dated files and prohibited actions, then read the content yourself after delegating work in a single request.
2026-10-16
Google Flow Prompts That Preserve Bottle Shapes
A comparison of three prompts, each generated three times, explores why explicitly naming the shape and checking every result matter.
2026-10-16
Check Before Spending Free Google Flow Credits
This guide separates the free-credit figures recorded in the October 12, 2026 video from the generation settings you need to check in your own account today.
2026-10-16
LightCraft and PhotoCraft With Ten RAW Photos
A video editing and exporting ten Canon RAW photos with LightCraft and PhotoCraft highlights issues to check, including boundary bands and preset settings that did not transfer.
2026-10-16
REA Note Alarms Without Cloning an App
A personal learning exercise reimplements selected Windows Sticky Notes behaviors, distinguishing the demonstrated alarm from its operating conditions and missing features.
2026-10-16
Give AI Skills Recent Post Records
A fictional-brand exercise compares content plans before and after recent post records are supplied to an AI skill, while noting the limits of what was verified.
2026-10-16
Four Checks Before Deleting Temporary Files
Before automatically deleting temporary files, check when work was completed, whether review is pending, and whether each copy matches an original that will be kept.
2026-10-16
Three AI Tools and Four Tasks
Observations from a video comparing ChatGPT, Claude, and Aside on four identical tasks highlight the need to check account settings and the evidence behind results.
2026-10-16
Six AI Uses From One Week’s Experiments
This retrospective review of eight full-length videos and seven Shorts published from the morning of October 9 through the morning of October 16 summarizes six AI experiments, their limitations, and the process of checking results.
2026-10-14
Check the Evidence in AI Meeting Notes
Keep preferences from becoming commitments in AI meeting notes. An original, unexecuted prompt and review method for checking sources, owners, dates and conditions.
2024. 05. 20
Getting the Most Out of N8N Workflows
An in-depth guide to N8N architecture and practical applications that go beyond no-code limitations. The ultimate guide to business automation.
2026-09-01
Should an Agent Have the Authority to Refuse?
An agent that does as it is told is easy to manage. But it also faithfully executes flawed instructions. The real boundary of autonomy lies in what it can choose not to do, rather than only what it can do.
2026-10-05
An Agent Did It Once. Can You Trust That?
A demo where an agent finishes a task once says little about tomorrow. This chapter looks at repeating the same task, grading the final state instead of the agent's words, and what published research and engineering write-ups say about both.