2026-07-255 min read
The Same Question, A Different Assistant
A multilingual assistant does not only translate its answers. It may change how readily it challenges, reassures, explains, and admits doubt.
AIDesignTrust
Read the memoField Notes
Selected observations from the overlap of systems, selfhood, and synthetic confidence.
An archive of arguments, reports, and studio notes. Some are practical memos. Some are atmospheric. The useful ones leave a handle on the work.
2026-07-247 min read
The Best Model for the Job Was at Home
During a live security incident, the useful AI was not the most capable model on the market. It was the one responders could run on their own infrastructure, keeping the work inside their environment.
AISystemsWork
Read the memo2026-07-216 min read
When Work Stops Looking Like A Conversation
Agentic work is beginning to have rhythms instead of turns. Draft marks now need to travel upstream from the finished artifact into the run that produced it.
AIWorkInterfaces
Read the memo2026-07-206 min read
The Web Has Two Front Doors
As sites begin exposing structured tools for browser agents, the old human-facing page is getting a second entrance. The design question is what the new door is allowed to do.
AIWebInterfaces
Read the memo2026-07-196 min read
The Useful Agent Has Blinders
As workplace agents gain more access, the better design question is not how much context they can swallow. It is what they are mercifully kept from seeing.
AIWorkInterfaces
Read the memo2026-07-186 min read
When The Repository Starts Keeping Score
Repository-level AI metrics can make agent work easier to see. They can also teach organizations to mistake visible motion for software health.
AIWorkQuality
Read the memo2026-07-156 min read
The Cheapest Model Is A Management Decision
As AI products begin routing work across models, providers, budgets, and fallback paths, the humane question is whether people can see what kind of intelligence their work was assigned.
AIWorkDesign
Read the memo2026-07-146 min read
The Assistant Keeps A Draft Of You
Memory-enabled AI is starting to build working theories of its users. The humane design question is whether people can inspect, revise, and expire those theories without becoming administrators of their own personalities.
AIDesignLife
Read the memo2026-07-136 min read
Polished Work Needs Draft Marks
As AI agents get better at producing finished-looking decks, documents, spreadsheets, and interfaces, the humane design question is whether review still has somewhere to attach.
AIWorkDesign
Read the memo2026-07-126 min read
Agent Traffic Needs House Rules
As AI agents start visiting the web on our behalf, the humane question is no longer only whether bots may enter. It is what kind of visitor each bot is pretending to be.
AIInterfacesWeb
Read the memo2026-07-116 min read
Capability Gates Need Receipts
As frontier models become more powerful and more restricted, the humane design question is whether access decisions leave enough evidence for people to understand the bargain.
AITrustWork
Read the memo2026-07-116 min read
AI Slop Is Scalable Authenticity
AI slop feels uncanny because it scales the polished, repeatable authenticity that platform culture had already taught people to perform.
AICultureWork
Read the memo2026-07-106 min read
Saved Time Needs Somewhere To Go
As AI tools begin returning hours to workers, the humane workplace question is what organizations do with that time before it quietly becomes more work.
AIWorkDesign
Read the memo2026-07-106 min read
AI Rollouts Need Rehearsal Rooms
As agentic tools move into everyday work, the humane adoption problem is whether people get enough visible practice to learn what good judgment looks like together.
AIWorkLearning
Read the memo2026-07-076 min read
Customer Service AI Needs An Appeal Path
As customer-service agents start resolving, routing, and acting across support systems, the humane interface question is whether people can contest the answer without disappearing into the workflow.
AIWorkDesign
Read the memo2026-07-066 min read
AI Capacity Needs A Civic Interface
As AI companies turn intelligence into chips, leases, power contracts, and neighborhood infrastructure, the humane design question is whether people can see the capacity they are being asked to live with.
AIDesignInfrastructure
Read the memo2026-07-036 min read
Agent Controls Need Weight
As AI work moves from chat into background agents, shortcuts, and physical controls, the humane interface question is whether control feels like responsibility or just another way to go faster.
AIDesignWork
Read the memo2026-07-025 min read
Science Agents Need Strange Questions
As AI research workbenches start running analyses, querying databases, and producing reproducible artifacts, the humane scientific question is whether they preserve exploration or quietly narrow it.
AIScienceWork
Read the memo2026-07-018 min read
Agentic Development Needs Load-Bearing Pauses
As coding agents keep working after people stop, the surprise infrastructure bill is not only a cost problem. It is a sign that our software delivery systems were designed around human pauses.
AIWorkInfrastructure
Read the memo2026-06-306 min read
Institutional AI Needs A Training Record
As companies turn their own work history into AI capability, the humane design question is whether people can see what the system learned from them.
AIWorkDesign
Read the memo2026-06-306 min read
AI Meetings Need A Guest List
As AI notetakers move from video calls into everyday rooms, the humane design question is whether people know who is listening, what is being kept, and when the room can go back to being a room.
AIWorkDesign
Read the memo2026-06-276 min read
AI Work Needs A Shared Floor
As agentic tools become ordinary in some corners of work and exotic in others, the humane design question is how teams keep a shared sense of the work itself.
AIWorkDesign
Read the memo2026-06-266 min read
AI Patches Need Maintainers
As AI security tools learn to find and draft fixes faster, the humane infrastructure question is whether maintainers gain capacity or inherit a louder queue.
AISecurityWork
Read the memo2026-06-256 min read
Agent Work Needs Role Lines
As coding agents make technical execution easier to delegate across an organization, the humane design problem is deciding where capability ends and responsibility begins.
AIWorkDesign
Read the memo2026-06-246 min read
AI Speed Needs A Second Clock
As AI tools make work feel faster at the keyboard, the humane measurement problem is whether the whole system is actually getting calmer, clearer, and better.
AIWorkQuality
Read the memo2026-06-236 min read
Creative Agents Need Rough Edges
As AI moves deeper into design and editing tools, the humane question is whether the work still leaves enough texture for taste to attach.
AIDesignWork
Read the memo2026-06-226 min read
Health AI Needs A Handoff
As medical models become more capable, the humane design problem is not only whether they can find the right answer. It is how uncertainty, context, and responsibility move back to human care.
AIHealthDesign
Read the memo2026-06-216 min read
AI Classrooms Need A Syllabus
As study bots and teacher copilots move into schoolwork, the humane question is not only whether students used AI. It is whether anyone knows what kind of learning the tool was supposed to leave behind.
AIEducationDesign
Read the memo2026-06-206 min read
Workspaces Need Housekeeping
As agents move into real files, tickets, calendars, and repos, the hidden work is not only giving them better tools. It is making the work itself legible enough to share.
AIWorkDesign
Read the memo2026-06-196 min read
Local AI Needs A Front Door
As AI moves into browsers, operating systems, and everyday apps, running locally is useful. It is not the same as knowing who is inside the room.
AIInterfacesPrivacy
Read the memo2026-06-186 min read
Scheduled AI Needs A Shelf Life
As assistants learn to watch, remind, and run in the background, the humane interface question is not how often AI can return. It is when delegated attention should expire.
AIAttentionDesign
Read the memo2026-06-176 min read
Agent Actions Need Repair Paths
As AI agents move from suggestions into work systems, the humane interface problem is no longer only approval. It is how people recover when a plausible action has already changed the world.
AIDesignWork
Read the memo2026-06-166 min read
Junior Work Needs A Floor
AI may remove some of the routine work that used to define early-career jobs. The humane question is what replaces it as a place to practice judgment before judgment is demanded.
AIWorkLearning
Read the memo2026-06-167 min read
Agent Work Needs Stop Conditions
Coding agents do not only answer prompts. They inherit a workspace, a tool belt, a loop, and sometimes a very loose idea of when the work is over.
AIEngineeringWork
Read the memo2026-06-155 min read
AI Needs A Degraded Mode
When a model is rerouted, restricted, or suddenly unavailable, the humane fallback is not to keep the interface looking normal. It is to help people understand what changed and keep the work intact.
AISystemsWork
Read the memo2026-06-146 min read
Reproduction Needs Skepticism
AI agents are getting better at rerunning published research. The harder question is whether they can help institutions challenge a result instead of merely producing it again.
AIScienceQuality
Read the memo2026-06-136 min read
Public Input Needs Somewhere To Go
AI companies are getting better at asking people what they fear and hope for. Listening only becomes accountability when the public can see what changed and who still has the power to say no.
AISocietyTrust
Read the memo2026-06-125 min read
AI Workflows Have Scar Tissue
A prompt stack is often a record of old failures, anxious fixes, and institutional memory. Smarter models do not erase that history; they make us decide what still deserves to survive.
AIWorkSystems
Read the memo2026-06-127 min read
AI Needs a Nutrition Label
As AI-generated work moves from answers into actions, people need a small, visible way to inspect what the work depends on before they trust it.
AIInterfacesTrust
Read the memo2026-06-116 min read
Shopping Agents Need A Receipt
As AI agents move from finding things to buying them, checkout has to preserve consent, regret, and responsibility after the button disappears.
AIInterfacesWork
Read the memo2026-06-106 min read
Job Search Needs A Human Shape
As AI starts helping people find jobs, rewrite resumes, and reorganize work itself, the humane question is not whether the search gets faster. It is whether anyone can still be recognized inside it.
AIWorkInterfaces
Read the memo2026-06-096 min read
AI Needs A Good Logout
As assistants become persistent agents across phones, browsers, workspaces, and local machines, the humane design problem is not only what they can do. It is how cleanly they can stop.
AIInterfacesWork
Read the memo2026-06-086 min read
Second Opinions Need Standards
As AI tools start asking other AI tools to critique their plans, tests, and output, the humane question is whether the second opinion is defending a real standard or just making uncertainty look managed.
AIQualityWork
Read the memo2026-06-086 min read
Agent Work Needs Handles
As AI work moves out of chat and into canvases, sites, agents windows, and annotated artifacts, the humane design question is whether people can still grab the work where judgment actually happens.
AIInterfacesWork
Read the memo2026-06-056 min read
Agent Plans Need Witnesses
As AI agents learn to plan before they act, the humane design problem is not making the plan look authoritative. It is making sure somebody can still notice what the plan leaves out.
AIWorkDesign
Read the memo2026-06-046 min read
Advice Bots Need Doubt
As people ask AI systems what to do with their jobs, bodies, money, children, and relationships, the humane design problem is not better confidence. It is better doubt.
AIHealthInterfaces
Read the memo2026-06-036 min read
The Office Graph Needs Manners
As AI tools learn from the relationships, files, meetings, messages, and permissions around work, the humane design question is no longer only what context they can see. It is how politely they use it.
AIWorkDesign
Read the memo2026-06-028 min read
Disposable Software Needs Aftercare
As coding agents move into knowledge work, more teams will make small tools for small moments. The hard part is deciding what deserves to keep existing.
AIWorkSoftware
Read the memo2026-06-026 min read
Agents Need Badges
As AI agents get identities, permissions, sponsors, and audit trails, the important design question is not only what they can do. It is how they become legible members of the workplace.
AIWorkTrust
Read the memo2026-06-018 min read
People Are Not Exception Queues
As agents take on more execution, the humane workplace question is not whether humans stay in the loop. It is what kind of loop we are asking them to live inside.
AIWorkDesign
Read the memo2026-05-297 min read
Work Files Became Agentic Rooms
As AI moves from answering beside our files to acting inside them, documents, spreadsheets, and decks are becoming rooms where judgment, authority, and evidence have to coexist.
AIWorkInterfaces
Read the memo2026-05-296 min read
AI Memory Needs Edges
As assistants remember across chats, projects, repositories, and connected tools, the humane design question is no longer whether memory helps. It is where memory should stop.
AIInterfacesTrust
Read the memo2026-05-276 min read
Search Needs Exits
As AI search becomes a place where answers, interfaces, planning, and transactions happen, the humane design question is whether people can still leave the answer cleanly.
AIInterfacesKnowledge
Read the memo2026-05-265 min read
Sandboxes Are The Agent Interface
As agents move from chat into managed work environments, the most important interface may be the room we let them work inside.
AIInterfacesTrust
Read the memo2026-05-266 min read
Provenance Became Part Of The Interface
As AI agents write code, sign commits, and leave traces through our tools, authorship stops being a footer. It becomes part of whether work can be trusted.
AITrustWork
Read the memo2026-05-266 min read
Instructions Became Infrastructure
As agents begin to run from Markdown files, skills, goals, and workflow rules, the quality of ordinary instructions starts to matter like software architecture.
AIWorkInfrastructure
Read the memo2026-05-266 min read
Good Tech Ideas Can Come From Anybody
AI tools may not make everyone a software engineer. Their more interesting promise is that they let more people bring a serious idea close enough to software to be heard.
AISoftwareCreativity
Read the memo2026-05-255 min read
AI Code Review Needs Disagreement
As AI review moves from comments into suggested fixes and agentic repair, the important question is whether teams still have enough friction to notice what quality actually requires.
AICodingQuality
Read the memo2026-05-226 min read
Background Agents Need Endings
As AI work moves from active sessions into phones, schedules, locked computers, and remote agent threads, the humane design problem is no longer only how work begins. It is how it stops.
AIWorkAttention
Read the memo2026-05-196 min read
Agent-Readable Software Needs Caretakers
As software gets redesigned for agents as well as humans, the hidden work is not making everything autonomous. It is keeping systems legible enough to trust.
AIInterfacesMaintenance
Read the memo2026-05-186 min read
Ambient AI Needs Friction
As AI moves into default models, keyboards, cursors, widgets, and operating systems, the humane design problem is no longer access. It is knowing when help should slow itself down.
AIInterfacesAgency
Read the memo2026-05-186 min read
AI Is Thinning The Apprenticeship Layer
The first labor problem may not be that AI eliminates entire professions. It may be that it quietly removes the messy early work where people learn how judgment is made.
AIWorkLearning
Read the memo2026-05-176 min read
Work Chat Became An Execution Surface
As agents move into Slack channels, schedules, and shared directories, the workplace conversation layer is becoming a place where work gets triggered, reviewed, and quietly delegated.
AIWorkInterfaces
Read the memo2026-05-155 min read
The Design-Code Loop Needs Taste
As AI collapses the distance between canvas and code, the scarce work is no longer translation. It is knowing what should survive the loop.
AIDesignQuality
Read the memo2026-05-156 min read
AI Metrics Are Management Design
As AI work becomes metered by team, model, feature, and cost, the question is no longer whether people use the tools. It is what the organization teaches them to value.
AIWorkQuality
Read the memo2026-05-143 min read
Deep Research Changed The Shape Of Knowledge Work
The important shift is not that AI can summarize faster. It is that research is starting to look like delegated labor instead of assisted search.
AIResearchWork
Read the memo2026-05-133 min read
MCP Made AI Tools Less Lonely
The breakthrough is not one more model. It is a shared way for agents to reach tools, data, and systems without every product pretending to be a closed universe.
AIToolsInfrastructure
Read the memo2026-05-123 min read
Context Engineering Is Replacing Prompt Engineering
The frontier is moving away from clever wording and toward the design of what the model can see, retrieve, compress, and ignore.
AIDesignSystems
Read the memo2026-05-113 min read
Plugins Are Becoming The Real AI Product
The model may still be the engine, but the product advantage is moving toward packaged capability: tools, rules, workflows, and embedded know-how.
AIProductWorkflows
Read the memo2026-05-103 min read
The Browser Is Turning Into An Agent Workbench
The browser used to be where we did the work. Increasingly it is becoming the place where we supervise work being done on our behalf.
AIWebAttention
Read the memo2026-05-093 min read
Multimodal Memory Will Change What Search Feels Like
When text, images, audio, and video can live in the same retrieval space, search stops being a library query and starts becoming a memory system.
AISearchMedia
Read the memo2026-05-083 min read
Coding Agents Made Management Visible
The real bottleneck in software was never just typing speed. It was the hidden work of scope, delegation, review, context, and trust.
AICodingManagement
Read the memo2026-05-073 min read
Model Choice Became An Interface Decision
When multiple frontier agents live inside the same product, model selection stops feeling like loyalty and starts feeling like workflow design.
AIProductStrategy
Read the memo2026-05-062 min read
AI Capability Is Outpacing Workplace Theory
The tools are getting better at acting, researching, coding, and remembering, while many organizations still think work is a matter of attendance, output, and managerial visibility.
AIWorkHealth
Read the memo2026-05-052 min read
Quality Is The Next AI Feature
After the fascination with speed, the durable differentiator may be whether a tool helps people produce work that is calmer, cleaner, and more worth trusting.
AIQualitySociety
Read the memo2026-04-095 min read
The Problem With OKRs
OKRs can name a destination, but they often miss the conditions that make health, judgment, and quality possible in the first place.
EcologyWorkEmergence
Read the memo2026-03-192 min read
The AI Morning Routine Worked Until A Human Woke Up
A useful routine generated by a model can still fail on contact with energy levels, mood, weather, and the older operating system known as a body.
AIWorkReflection
Read the memo2026-03-122 min read
Weight Loss Became Easier When I Stopped Negotiating With Myself
The shift was not toward harsher discipline, but toward better systems, fewer identity dramas, and more attention to what my days were already teaching me.
HealthMindfulnessSystems
Read the memo2026-03-052 min read
Catan, Theme Parks, and Travel Are All Lessons in Experience Design
Scarcity, anticipation, pacing, and surprise are not side effects. They are the architecture of how systems become memorable.
GamesTravelDesign
Read the memo