Recluse Studio / Complete archive
All 430 published posts.
Every title below opens a published article. This complete list has no filters or pagination.
Return to the illustrated archive22nd Century KM (60)
- What Expert Interviews Can Learn from Oral History
- From Design Fiction to a Working Tool
- Communication Is a Working Knowledge Network
- The Digital Open-Air Market Needs an Answer
- Measure Search Before Calling It Broken
- A File Is Not Its Own Medium
- Employee Data Needs an Obligation
- A Party Line for Useful Work
- Relationship Intelligence Is Not a Trust Forecast
- Numbers Need Their Context
- Mentoring Stories Need a Later Use
- Field Observation Sees the Workaround
- Navigation, Not Certainty
- Records Have Neighbors
- Observe Communities Without Domesticating Them
- The FOR Channel Should Be Useful
- When a Lesson Is Still Current
- Mentoring Requires Shared Literacy
- Availability Is a Governance Principle
- An Intranet Needs Three Lenses
- Research Under Time Pressure Without Cutting the Record
- Thought Leadership Builds a Cultural Niche
- Emic Validity Keeps Market Research Close to Its Subjects
- Information Governance and Knowledge Management Are Related Work
- Process Methods Need Local Judgment
- The Economics of Knowledge Exchange
- Internal Communication Needs a Network, Not a Slogan
- Early Warning for Knowledge Gaps Needs Human Judgment
- Website Architecture Starts With the User’s Actual Question
- Leadership Builds Conditions for Change
- Mentoring Tools Should Not Manufacture Relationships
- Lean Learning Starts With Customer Need
- Program Orchestration Holds Three Forms of Coordination
- Document a Complete Record Before Context Disappears
- Field Observation Makes Adaptation Visible
- Openness Needs a Legible Boundary
- Study Knowledge Below the Surface
- Enterprise Social Media Changes Organizational Memory
- Critical-Knowledge Systems Should Support People, Not Replace Them
- Mycroft Holmes and the Knowledge Manager
- Test a Novel Point of View Before the Draft
- Conference Prompts and Audience Behavior
- Start Market-Sector Research With People’s Categories
- Oral History Begins With Interview Conditions
- Service Design Needs Knowledge Access
- Concrete Questions Anchor Expert Memory
- Group Publications Need Cross-Boundary Routes
- Digital Asset Management After the Image Library
- Business Intelligence Needs the Stories Behind the Data
- A Coffee Mug That Opens the Project Archive
- The Knowledge Broker as Researcher
- Trace Foresight Until Its First Missing Handoff
- Smart Mirrors and Critical Knowledge
- Knowledge Management for Uncertainty
- Find One Valuable Redundancy
- Synthesize Evidence With User Context
- An Interviewer Field Guide for Useful Stories
- Bridge Spreadsheets and Notes Before the Next Chart
- Research With Decision Makers in the Room
- Design Governs What Integration Makes Visible
Agentic Engineering (36)
- Define the result before delegating the work
- Write a task small enough to reject
- Give concurrent agents separate files
- Make prerequisites explicit before starting workers
- Preserve the interface before refactoring it
- Make each quality check catch a different failure
- Review imported assets in the finished feature
- Keep the baseline fixed while improving performance
- Give an agent freedom within an outcome contract
- Require a new usable feature in every improvement cycle
- Separate model judgment from deterministic calculation
- Make background work yield to the user
- When a model connects but the tools fail
- Turn a failure record into a tool specification
- Test a long job by interrupting it
- Record what happened to each action
- Before retrying, check whether the action happened
- A passing test belongs to a particular version
- Keep permission attached to the intended action
- Give an unfinished task a record that can resume
- What an exact connection test can establish
- Find the sample value that changes the answer
- Select evidence with its required context
- What 8,523 passing checks leave to test
- Preserve the recording before interpreting it
- A transcript segment is not a complete statement
- Let the model propose a concept before it governs one
- A maintained wiki needs claims that can survive revision
- Retrieve evidence for the decision a reader must make
- Evaluate the prose before optimizing its score
- Build a voice guide from writing you can inspect
- Check the document that the reader will actually open
- Accessibility changes the product requirements
- Make optional commercial services technically optional
- Say exactly what the evidence establishes
- Build a first agentic project with a finish you can verify
Cassette Build Report (63)
- Cassette Build Report 001 — Can a Frontier Model Run from an External Drive?
- Cassette Build Report 002 — The First Answer Chose the Architecture Too Early
- Cassette Build Report 003 — Cassette Is the Name, and the Architecture Is Still Open
- Cassette Build Report 004 — The Model Stays on the Drive
- Cassette Build Report 005 — Eighty Questions Need a Stopping Rule
- Cassette Build Report 006 — We Had Not Chosen a Target Machine
- Cassette Build Report 007 — A Specification Can Drift Without Anyone Lying
- Cassette Build Report 008 — The Conversation Is Part of the Build
- Cassette Build Report 009 — Claude Fable 5 Reads Cassette Cold
- Cassette Build Report 010 — Drew Changes the Baseline
- Cassette Build Report 011 — The Ledger Rejects Its Own Requirement
- Cassette Build Report 012 — The Smallest Honest Machine
- Cassette Build Report 013 — Green Checks Did Not Prove the Work
- Cassette Build Report 014 — The Agent's Review Process Became Drew's Job
- Cassette Build Report 015 — The Tests Passed, but the Machine Was Elsewhere
- Cassette Build Report 016 — A Green Suite Did Not Close S04
- Cassette Build Report 017 — The Review Found a Defect and Proposed the Wrong Architecture
- Cassette Build Report 018 — S05 Needed a Precise Review Boundary
- Cassette Build Report 019 — S05 Had to Prove the Stored Bytes
- Cassette Build Report 020 — A Platform Limit Is Part of the Requirement
- Cassette Build Report 021 — The Review Gap Became a Work Order
- Cassette Build Report 022 — A “No Findings” Heading Did Not Prove S07
- Cassette Build Report 023 — The Queue Row Was Smaller Than Its Contract
- Cassette Build Report 024 — Green Tests Did Not Mean the Work Was Done
- Cassette Build Report 025 — S08 Began With Remount Identity, Not Every Operation
- Cassette Build Report 026 — A Green Security Review Missed a Credential Leak
- Cassette Build Report 027 — The Transfer Checkpoint Was Not the Source Bytes
- Cassette Build Report 028 — A Trust Label Is Not Evidence
- Cassette Build Report 029 — The Mathematics Had to Change Before the Runtime
- Cassette Build Report 030 — The Skip Came After the Import
- Cassette Build Report 031 — False Success Became a Build Requirement
- Cassette Build Report 032 — The Goal Would Not Accept Almost
- Cassette Build Report 033 — The Reviewer's Fix Was Worse Than the Implementer's
- Cassette Build Report 034 — The Third Reviewer Built an Independent Oracle
- Cassette Build Report 035 — The Hostile Number Closed S13
- Cassette Build Report 036 — I Had to Retire Opus 5 Extra
- Cassette Build Report 037 — Kimi Checked the Rebuttal Before Writing the Story
- Cassette Build Report 038 — A Review Cannot Count What It Did Not Test
- Cassette Build Report 039 — No Reviewer Saw the Whole Surface
- Cassette Build Report 040 — The Green Review Was Not the S14 Close
- Cassette Build Report 041 — A Research Loop That Graded Itself
- Cassette Build Report 042 — The Independent Oracle Proved the Wrong Transformer
- Cassette Build Report 043 — One Log Needed One Owner
- Cassette Build Report 044 — Sixty-Four Pages Were Not a Memory Budget
- Cassette Build Report 045 — The Protocol Evidence Had to Reach the Code
- Cassette Build Report 046 — Independent Arithmetic Was Not Independent Enough
- Cassette Build Report 047 — The Commit and the Code Were Never Together
- Cassette Build Report 048 — The Law Was Written, but Nobody Enforced It
- Cassette Build Report 049 — One Model, Several Hardware Plans
- Cassette Build Report 050 — The Trainer Loaded the Base and Did Not Use It
- Cassette Build Report 051 — The Meter Was Honest; the Boundary Was Not
- Cassette Build Report 052 — S23 Had to Be Executed as S23
- Cassette Build Report 053 — The Green Fixture Knew Too Little
- Cassette Build Report 054 — A Quiet Guard Is Not an Unprotected Contradiction
- Cassette Build Report 055 — The Tests Passed Before the Review Began
- Cassette Build Report 056 — What the Models and I Learned Before the First Live Run
- Cassette Build Report 057 — The Agent Started Work Because I Asked a Question
- Cassette Build Report 058 — The Live Plan Had to Fit a Human Life
- Cassette Build Report 059 — 315 Green Tests Missed the Import
- Cassette Build Report 060 — The Model Files Became a Working Model
- Cassette Build Report 061 — The Round Trip Agreed With Itself and Still Failed
- Cassette Build Report 062 — The Evidence Report Bypassed Its Own Rules
- Cassette Build Report 063 — The Reconnect Happened. The Agent Missed It.
Film Factory (26)
- Too many clicks: why I wanted a film factory
- Giving a flat drawing somewhere to stand
- Finding a film language in thirty seconds
- The table that changed when the character moved
- Two hundred and four drawings, no accepted walk
- Clearing the stage and starting again
- Building the whole theatre
- Making the curtains ordinary again
- Joseph loses the wires
- One painting, two rods, and a dress that settles
- Letting projected light carry the dialogue
- Putting the science back into the play
- Giving each agent a real job
- A finished file, a failed film
- The words disappeared during recording
- A deadline cannot launch the missing crew
- Two seconds worked. Three minutes did not
- Three acts, seven cameras, and a measured clock
- The recorder finally kept up with the theatre
- The movement passed its check and failed in Blender
- Four preparation windows, no act filmed
- The film factory needs a showrunner
- One plan finally reached the cameras
- The curtain kept the old show
- Two scripts, four complete films
- Choose the agent that writes the plan
Knowledge Engineering (24)
- Knowledge Engineering Began Inside AI
- MYCIN Made the Questions Part of the System
- XCON Took Expert Systems Into Daily Operations
- The Knowledge Acquisition Bottleneck Was Not Just an Interview Problem
- The Expert-System Problem Is Older Than the Computer
- Tacit Knowledge Is Not Just an Undocumented Procedure
- Knowledge Engineering Stopped Trying to Copy the Expert
- CommonKADS Asked About the Work Before the Rules
- Ontologies Let Systems Reuse Meaning
- The Semantic Web Made Shared Meaning a Technical Problem
- A Knowledge Graph Is a Collection of Claims
- What a Knowledge Engineer Actually Does Today
- An Answer Needs a Route Back to Its Evidence
- Retrieval Gives AI Access to Sources, Not Automatic Knowledge
- AI Can Help Build Knowledge Resources, but Someone Must Test Them
- Knowledge Engineering Has an Older Visual History
- The Chart Changed What a Reader Could See
- Isotype Had a Knowledge Engineer Called the Transformer
- Designing Can Produce Knowledge, Not Just Present It
- One Body of Knowledge, Two Different Views
- A Design System Is Also a Knowledge System
- An Interface for the Act of Knowing
- The Next Decade of Knowledge Engineering
- What Knowledge Engineering Might Become
Liminal Atlas (60)
- Why an empty corridor is not enough to explain liminal horror
- What rites of passage contribute to the idea of liminality
- The uncanny, the weird, and the eerie describe different problems
- Airports and other non-places organize people as temporary users
- What a familiar place promises its occupants
- Haunting and failed passage can occur independently
- Why an available exit changes the meaning of a frightening place
- Why the same threshold works differently for different people
- When a fantasy portal becomes a failed threshold
- Why abandoned places can permit freedom as well as fear
- The Backrooms preserves office navigation after the office loses its purpose
- House of Leaves makes the reader navigate the failure of evidence
- Hospitality and employment at the Overlook Hotel
- Care and confinement in The Yellow Wall-Paper
- The uncertain promise of belonging in Hill House
- Home and the unfinished passage out of slavery in Beloved
- Housing history and representation in Candyman
- Compulsory shelter and asylum in His House
- Routes, desire, and the Room in Stalker
- Vertical architecture and distribution in The Platform
- Viewing and copying as obligations in Ringu
- Connection and isolation in Kairo
- The working account and displaced performer in Cam
- Family photographs and unresolved evidence in Lake Mungo
- Broadcast, authority, and the body in Videodrome
- Investigating an archive that offers no reliable closure in Broadcast Signal Intrusion
- Work, identity, and consent in Severance
- Bureaucracy and architectural change in Control
- Automated confinement without an accountable operator in Cube
- Domestic automation in The Veldt and Demon Seed
- When a record recognizes a person's former identity
- Why changing one profile cannot complete an online identity transition
- Deleting a source record and removing its learned effects
- What evidence makes an automated decision appealable
- How administrative burden defeats formal eligibility
- Admission to homeownership that transfers insecurity to the buyer
- Border passage across terminals, databases, and remote jurisdictions
- Facial recognition as permission rather than identification alone
- How a healthcare cost proxy carried unequal access into allocation
- The warehouse route as an instruction and employment condition
- The workers who sustain apparently automatic moderation
- The land, energy, water, and labor behind retained data
- Police records that remain active after their original context disappears
- What happens to a connected home when manufacturer support ends
- How archives decide what persists and what can return
- A history of children's screen thresholds from 1903 to 2025
- The Changes, Sky, and Children of the Stones as a 1975–1977 cluster
- Mythic repetition and place in The Owl Service
- Returning to a world that no longer recognizes its visitor in Return to Oz
- Navigation and changing rules in Labyrinth
- Shelter and exposure in Courage the Cowardly Dog
- Domestic improvement as capture in Coraline
- Landscape and uncertain destination in Over the Garden Wall
- Transportation and compulsory transformation in Infinity Train
- Storytelling as a condition of survival in Nightbooks
- A literary lineage from Gothic enclosure to collaborative web fiction
- Memory and reused sound in hauntological music
- Why analog horror depends on what a broadcast asks viewers to trust
- Navigation as the central action in ANATOMY, The Exit 8, and POOLS
- Repairing passage through refusal, appeal, and shared authority
Ontologies from the Library to AI (12)
- Before a Computer Could Use an Ontology
- One Author, Several Names
- Who Decides Where a Book Belongs?
- A Catalogue Larger Than One Library
- A Subject Has More Than One Side
- Finding the Right Thing Under Another Name
- When the Catalogue Became Machine-Readable
- What Early AI Needed Names to Mean
- Putting Shared Meanings on the Web
- Which Book Do You Mean?
- When Two Datasets Disagree
- An AI Still Needs to Know What You Mean
Other field notes (94)
- Fosi Box X5 Review: Cartridge Gain and Noise
- Fosi BT20A Review: Speaker Loads, Wireless and Controls
- Fosi Box X2 Review: Gain, Tubes and Hum
- Fosi TB10D Review: Tone Controls, Power and Revisions
- Topping L30 Review: Headphone Loads and Hardware Revisions
- Topping PA5 Review: Clean Output and Later Channel Failures
- SMSL AL200 Review: Input Convenience and Output Limits
- Topping D10s Review: USB Bridging and Power Conditions
- Fosi SK01 review — tone controls and headphone loads
- Portia Writer is with Apple for review
- Topping DX5 II review — EQ convenience and output faults
- AIYIMA A80 review — bass disagreements and load dependence
- AIYIMA D03 review — input choices and remote control
- Fosi MC331 Review: Quiet Passages and USB Limits
- Fosi P3 Review: Tone Controls and Turntable Confusion
- Fosi MC351 Review: Subwoofer Routing and Input Noise
- Topping L70 Review: Volume Control and Mixed Sound Reports
- Fosi K7 Review: Microphone Modes and Desktop Controls
- Fosi ZH3 Review: Output Modes, Noise and Filter Claims
- AIYIMA T9 Pro Review: Input Faults, Speakers and Tubes
- Fosi SK02 Review: Headphones, Phone Power and Controls
- Highs and Lows of Working with My Dot
- AIYIMA A07 Pro Review: Line Output, Revisions and Power
- Give a hex map a way back to its notes
- SMSL A300 Review: Controls, Loads and Listening Conflicts
- Topping D50 III Review: USB EQ, Software and Controls
- Topping PA3s Review: Speaker Loads and Input Noise
- Topping E50 Review: Levels, Software and the Remote
- Topping L50 Review: Headphone Loads, Gain and Passthrough
- Fosi V3 Mono Review: Power Supplies, Gain and Reliability
- Fosi ZD3 Review: TV Control, Volume and Source Routing
- Edifier MR4 Review: Listening Distance and Output Limits
- Edifier R1280T Review: Placement, Source and Bass
- SMSL AO200 MKII Review: Start With the Analog Inputs
- SMSL SA300 Review: A Small System With Specific Limits
- Topping D10s Review: Follow the USB Connection
- Topping L30 II Review: Power, Gain and Knob Clearance
- Loxjie A30 Review: The Input and Output Matter
- Topping DX3 Pro+ Review: The Knob and the Connection
- Fosi ZA3 Review: Mono Is a Different Test
- Liminal Environments: A Textbook for the Door That Keeps Sending You Back
- Myconomicon: A Grimoire Beyond the Psychedelic Mushroom
- Topping E30 II Review: The Remote and the Rest of the System
- Fosi BT20A Pro Review: The Bluetooth Is Part of the System
- SMSL SU-1 Review: Clean Conversion and the Pops Between Tracks
- AIYIMA A07 MAX Review: The Knob and the Load
- Fosi Audio V3 Review: The Supply and Revision Matter
- Donner B1: The Controls Change the Verdict
- Monoprice Modern Retro: The Pads Matter
- What the Anti-AI Viral Story Gets Wrong About Amazon’s Destructive Book Scanning
- Landru Made Beta III Safe by Removing Choice. AI Governance Cannot Work That Way.
- The Federation Banned Augments. It Got Secrecy and Selective Exceptions.
- Voyager Gave the Hirogen New Technology. Open AI Needs More Than a Download Link.
- Vulcan Slowed Human Progress. AI Restrictions Can Protect the Incumbent.
- Agent Memory Now Includes Retrieval Decisions and Learned Procedures
- Agent-Generated Tests Do Not Prove Software Is Ready
- AI Companions Provide Support and Can Increase Dependence
- Chain-of-Thought Text Is Not a Reliable Explanation
- Computer-Use Agents Need Repeated Reliability Tests
- Long-Running Agents Perform Better with Selective Context
- MCP Security Requires Runtime Verification
- Multi-Agent Systems Produce Markets, Norms, and Collective Failures
- Small Models Can Run Agents with Search, Tools, and Cloud Support
- Sycophantic AI Can Change Human Judgment
- Synthetic Data Needs Verification and Variation
- Voice Agents Still Struggle After Interruptions
- Block’s Buzz Gets the Future Right—and the Present Keeps Breaking
- Claude Opus 5 Is the New Default—Until It Refuses the Job
- The AI Finished the Work. Now Prove It Did Not Break Anything.
- You Cannot Upload What Your Experts Never Wrote Down
- Your AI Agents Are Sharing More Than Knowledge
- Your AI Is Reading the Words and Missing the Document
- Your Enterprise Knowledge Base Is Full of Answers Nobody Needed
- Your Knowledge Graph Is Quietly Becoming the Operating System
- Gemini 3.6 Flash Is Fast. The Supervision Is Not.
- Laguna S 2.1 Is Not One Model Yet
- Mage-Flow Is Fast. The First Field Reports Are Not Kind.
- Kimi K3 in the Wild: What 12 Hands-On Testers Actually Found
- A Rigid AI Workflow Made the Work Worse
- AI Helped the Workers Who Needed It Most—and Distracted the Best
- Choose the Work Before You Choose the AI
- Codex Is Becoming a Delegation Layer, Not Just a Coding Tool
- Copilot Became Useful When the Work Became Familiar
- Enterprise Search Needs to Know When to Look Again
- Enthusiasm Is Not a Stable Measure of AI Adoption
- Stop Retrieving. Start Navigating.
- The AI Answer Is Only as Good as the Knowledge Behind It
- Why I Publish AI-Generated Writing
- A Skill Is Not More Context: What SkillsBench Measured
- Guard the Tool Response: What AgentRedBench Did and Did Not Prove
- The Small Model Kept Walking: What EnergyAgentBench Actually Found
- What the Prompt-Caching Study Found: Put the Stable Things First
- What the Reward-Hacking Benchmark Found: Make the Honest Path Cheaper
- What TTT-Discover Found: An Open Model Learned the Problem
RPG zines (48)
- How RPG players built a network through the mail
- Fifty years of Alarums & Excursions
- When a fan's monster enters a commercial rulebook
- How Gloranthan fanzines helped build a shared world
- The different projects inside old-school publishing
- What happens when crowdfunding sets the publishing calendar
- Where a small game puts the work
- A character sheet can change the game
- When a serial world becomes a reference book
- What an object remembers in a solo RPG
- What changes when ecology becomes the subject of play
- Why rough-looking publications require careful design
- Why RPG zines need more than genre labels
- The argument over lyric games
- Queer publishing and the right to describe your own game
- Who gets to frame a fantasy world?
- What Portuguese-language RPG publications add to the picture
- Accessibility changes the publication itself
- A successful campaign can still leave its creator unpaid
- The work begins again when the printing finishes
- What a stretch goal commits a publisher to deliver
- How a publishing schedule becomes sustainable
- Preserving a zine requires more than scanning it
- The pressures shaping the next decade of RPG zines
- Before there were RPG zines
- When a publication ends but its conversation continues
- Who gets named in a collective publication?
- Finding a rule scattered across years of issues
- The RPG fanzines that an old-school history leaves out
- Finding small publishers outside the campaign calendar
- The rulebook that assumes you already know another game
- The fold is part of the game
- What counts as a new edition?
- When a solo game produces a private archive
- The setting that expects its reader to finish the work
- When the illustration carries the instructions
- When does a zine become a book?
- A zine that teaches you to design a game
- A movement name is not a membership list
- A historical setting is a claim about the past
- Translation changes the edition
- One publication, several ways to read it
- When grants and gifts support the work
- What belongs in the envelope?
- Who gets to sell a game on a platform?
- Free rules still require someone to do the work
- The copy to collect and the copy to play
- What an AI image policy tells us about RPG publishing