DAILY RECORD / 2025
365 published entries in 2025.
NOTES
The end of the third year
Retrospective AI note: The end of the third year.
A practical hope
Retrospective AI note: A practical hope.
The next question
Retrospective AI note: The next question.
A year of systems
Retrospective AI note: A year of systems.
The right kind of ambition
Retrospective AI note: The right kind of ambition.
The archive as evidence
Retrospective AI note: The archive as evidence.
Intelligence and meaning
Retrospective AI note: Intelligence and meaning.
A quiet test
Retrospective AI note: A quiet test.
The work I want to do
Retrospective AI note: The work I want to do.
The opportunity in redesign
Retrospective AI note: The opportunity in redesign.
What agents make visible
Retrospective AI note: What agents make visible.
The challenge of speed
Retrospective AI note: The challenge of speed.
The institutional imperative
Retrospective AI note: The institutional imperative.
A more mature optimism
Retrospective AI note: A more mature optimism.
The discipline of authorship
Retrospective AI note: The discipline of authorship.
The next educational mission
Retrospective AI note: The next educational mission.
A public responsibility
Retrospective AI note: A public responsibility.
The limits of convenience
Retrospective AI note: The limits of convenience.
Better objectives
Retrospective AI note: Better objectives.
The risk of optimisation without values
Retrospective AI note: The risk of optimisation without values.
The case for human diversity
Retrospective AI note: The case for human diversity.
Intelligence needs pluralism
Retrospective AI note: Intelligence needs pluralism.
The value of local knowledge
Retrospective AI note: The value of local knowledge.
A different kind of privacy
Retrospective AI note: A different kind of privacy.
The new trust boundary
Retrospective AI note: The new trust boundary.
The meaning of control
Retrospective AI note: The meaning of control.
More than the next release
Retrospective AI note: More than the next release.
The value of deliberate work
Retrospective AI note: The value of deliberate work.
A more demanding future
Retrospective AI note: A more demanding future.
The task of integration
Retrospective AI note: The task of integration.
The year in retrospect
Retrospective AI note: The year in retrospect.
November’s lesson
Retrospective AI note: November’s lesson.
A personal standard
Retrospective AI note: A personal standard.
The work of a generation
Retrospective AI note: The work of a generation.
Gratitude and responsibility
Retrospective AI note: Gratitude and responsibility.
The value of informed restraint
Retrospective AI note: The value of informed restraint.
The changing meaning of expertise
Retrospective AI note: The changing meaning of expertise.
Claude Opus 4.5 and the professional frontier
Retrospective AI note: Claude Opus 4.5 and the professional frontier.
Institutional trust compounds slowly
Retrospective AI note: Institutional trust compounds slowly.
The future of organisations
Retrospective AI note: The future of organisations.
The importance of visible agency
Retrospective AI note: The importance of visible agency.
The danger of seamlessness
Retrospective AI note: The danger of seamlessness.
The system-level race
Retrospective AI note: The system-level race.
Gemini 3 raises the integrated frontier
Retrospective AI note: Gemini 3 raises the integrated frontier.
A choice about concentration
Retrospective AI note: A choice about concentration.
The limits of inevitability
Retrospective AI note: The limits of inevitability.
A larger definition of building
Retrospective AI note: A larger definition of building.
The value of translation
Retrospective AI note: The value of translation.
The new public intellectual task
Retrospective AI note: The new public intellectual task.
From model to civilisation
Retrospective AI note: From model to civilisation.
The role of imagination
Retrospective AI note: The role of imagination.
The value of self-authorship
Retrospective AI note: The value of self-authorship.
More than assistance
Retrospective AI note: More than assistance.
The right to opt out
Retrospective AI note: The right to opt out.
The right to challenge
Retrospective AI note: The right to challenge.
The right to understand
Retrospective AI note: The right to understand.
A new social contract
Retrospective AI note: A new social contract.
The importance of constraints
Retrospective AI note: The importance of constraints.
The evidence of usefulness
Retrospective AI note: The evidence of usefulness.
What progress should mean
Retrospective AI note: What progress should mean.
The maturity test
Retrospective AI note: The maturity test.
October’s lesson
Retrospective AI note: October’s lesson.
Better questions for policy
Retrospective AI note: Better questions for policy.
The cost of abstraction
Retrospective AI note: The cost of abstraction.
A public language for risk
Retrospective AI note: A public language for risk.
The need for standards
Retrospective AI note: The need for standards.
The difficult middle
Retrospective AI note: The difficult middle.
The risk of overfitting
Retrospective AI note: The risk of overfitting.
The end of generic software
Retrospective AI note: The end of generic software.
The right kind of friction
Retrospective AI note: The right kind of friction.
A case for interoperability
Retrospective AI note: A case for interoperability.
The danger of lock-in
Retrospective AI note: The danger of lock-in.
Systems should preserve exits
Retrospective AI note: Systems should preserve exits.
A deeper definition of agency
Retrospective AI note: A deeper definition of agency.
What remains human
Retrospective AI note: What remains human.
The myth of effortless intelligence
Retrospective AI note: The myth of effortless intelligence.
The infrastructure of trust
Retrospective AI note: The infrastructure of trust.
The future of public services
Retrospective AI note: The future of public services.
The value of contestability
Retrospective AI note: The value of contestability.
Institutions need constitutions
Retrospective AI note: Institutions need constitutions.
The agent as an institution
Retrospective AI note: The agent as an institution.
The accountability gradient
Retrospective AI note: The accountability gradient.
The next professional skill
Retrospective AI note: The next professional skill.
The value of review
Retrospective AI note: The value of review.
The modern apprenticeship
Retrospective AI note: The modern apprenticeship.
Reliability over charisma
Retrospective AI note: Reliability over charisma.
The importance of edge cases
Retrospective AI note: The importance of edge cases.
The new evaluation challenge
Retrospective AI note: The new evaluation challenge.
The burden of continuity
Retrospective AI note: The burden of continuity.
Memory is governance
Retrospective AI note: Memory is governance.
The meaning of persistence
Retrospective AI note: The meaning of persistence.
Long-horizon work
Retrospective AI note: Long-horizon work.
September’s lesson
Retrospective AI note: September’s lesson.
Claude Sonnet 4.5 raises the agentic bar
Retrospective AI note: Claude Sonnet 4.5 raises the agentic bar.
The value of human pace
Retrospective AI note: The value of human pace.
The quality of a system’s silence
Retrospective AI note: The quality of a system’s silence.
A choice about attention
Retrospective AI note: A choice about attention.
Better than personalisation
Retrospective AI note: Better than personalisation.
The knowledge graph of a life
Retrospective AI note: The knowledge graph of a life.
Context is not a database
Retrospective AI note: Context is not a database.
The new middle layer
Retrospective AI note: The new middle layer.
The agent economy
Retrospective AI note: The agent economy.
A different form of ambition
Retrospective AI note: A different form of ambition.
The future of small teams
Retrospective AI note: The future of small teams.
The hard problem of scale
Retrospective AI note: The hard problem of scale.
The responsibility of builders
Retrospective AI note: The responsibility of builders.
More than compliance
Retrospective AI note: More than compliance.
The capacity to disagree
Retrospective AI note: The capacity to disagree.
A public standard for personal AI
Retrospective AI note: A public standard for personal AI.
The danger of invisible ranking
Retrospective AI note: The danger of invisible ranking.
The civic interface
Retrospective AI note: The civic interface.
The need for interpreters
Retrospective AI note: The need for interpreters.
The institutions we need
Retrospective AI note: The institutions we need.
Collective intelligence is fragile
Retrospective AI note: Collective intelligence is fragile.
Beyond individual productivity
Retrospective AI note: Beyond individual productivity.
The problem selection advantage
Retrospective AI note: The problem selection advantage.
A new hierarchy of skills
Retrospective AI note: A new hierarchy of skills.
The value of refusal
Retrospective AI note: The value of refusal.
The return of the editor
Retrospective AI note: The return of the editor.
A more rigorous optimism
Retrospective AI note: A more rigorous optimism.
The standard of evidence
Retrospective AI note: The standard of evidence.
The new normal
Retrospective AI note: The new normal.
August’s lesson
Retrospective AI note: August’s lesson.
The choice ahead
Retrospective AI note: The choice ahead.
A more demanding optimism
Retrospective AI note: A more demanding optimism.
The discipline of revision
Retrospective AI note: The discipline of revision.
A living body of work
Retrospective AI note: A living body of work.
The power of a clear thesis
Retrospective AI note: The power of a clear thesis.
More than acceleration
Retrospective AI note: More than acceleration.
What is worth preserving
Retrospective AI note: What is worth preserving.
The next institutional challenge
Retrospective AI note: The next institutional challenge.
The ethics of distribution
Retrospective AI note: The ethics of distribution.
The need for public tools
Retrospective AI note: The need for public tools.
More capable publics
Retrospective AI note: More capable publics.
The social value question
Retrospective AI note: The social value question.
A better measure of progress
Retrospective AI note: A better measure of progress.
The role of craft
Retrospective AI note: The role of craft.
The importance of practice
Retrospective AI note: The importance of practice.
The danger of skipped formation
Retrospective AI note: The danger of skipped formation.
A different apprenticeship model
Retrospective AI note: A different apprenticeship model.
The future of expertise
Retrospective AI note: The future of expertise.
The problem of comparative advantage
Retrospective AI note: The problem of comparative advantage.
A higher baseline
Retrospective AI note: A higher baseline.
Simplicity can hide complexity
Retrospective AI note: Simplicity can hide complexity.
The end of model selection?
Retrospective AI note: The end of model selection.
Intelligence as routing
Retrospective AI note: Intelligence as routing.
GPT‑5 changes the default
Retrospective AI note: GPT‑5 changes the default.
The new bottleneck
Retrospective AI note: The new bottleneck.
A broader imagination
Retrospective AI note: A broader imagination.
The test of convenience
Retrospective AI note: The test of convenience.
The disappearing interface
Retrospective AI note: The disappearing interface.
Integration over novelty
Retrospective AI note: Integration over novelty.
The next capability threshold
Retrospective AI note: The next capability threshold.
July’s lesson
Retrospective AI note: July’s lesson.
The right use of power
Retrospective AI note: The right use of power.
More than an assistant
Retrospective AI note: More than an assistant.
The value of accountability
Retrospective AI note: The value of accountability.
A question of identity
Retrospective AI note: A question of identity.
The architecture of choice
Retrospective AI note: The architecture of choice.
The danger of passive convenience
Retrospective AI note: The danger of passive convenience.
The limits of autonomy
Retrospective AI note: The limits of autonomy.
A new entry point
Retrospective AI note: A new entry point.
The work of specification
Retrospective AI note: The work of specification.
The economy of delegation
Retrospective AI note: The economy of delegation.
The need for checkpoints
Retrospective AI note: The need for checkpoints.
Autonomy is not abandonment
Retrospective AI note: Autonomy is not abandonment.
The permission model matters
Retrospective AI note: The permission model matters.
ChatGPT agent bridges research and action
Retrospective AI note: ChatGPT agent bridges research and action.
The end of a false separation
Retrospective AI note: The end of a false separation.
The browser was only the beginning
Retrospective AI note: The browser was only the beginning.
Agency requires recourse
Retrospective AI note: Agency requires recourse.
The source of confidence
Retrospective AI note: The source of confidence.
A long-term project
Retrospective AI note: A long-term project.
Building institutional capacity
Retrospective AI note: Building institutional capacity.
The role of public education
Retrospective AI note: The role of public education.
The unequal transition
Retrospective AI note: The unequal transition.
A more serious conversation about work
Retrospective AI note: A more serious conversation about work.
The task is not the job
Retrospective AI note: The task is not the job.
Human judgment as governance
Retrospective AI note: Human judgment as governance.
The need for legibility
Retrospective AI note: The need for legibility.
The cost of invisible action
Retrospective AI note: The cost of invisible action.
The boundary of agency
Retrospective AI note: The boundary of agency.
The collaboration question
Retrospective AI note: The collaboration question.
From tool to colleague
Retrospective AI note: From tool to colleague.
June’s lesson
Retrospective AI note: June’s lesson.
The future is relational
Retrospective AI note: The future is relational.
What I want to build
Retrospective AI note: What I want to build.
The power of a small group
Retrospective AI note: The power of a small group.
More capacity, same meaning
Retrospective AI note: More capacity, same meaning.
The social layer
Retrospective AI note: The social layer.
The importance of dissent
Retrospective AI note: The importance of dissent.
A different research discipline
Retrospective AI note: A different research discipline.
The risk of intellectual outsourcing
Retrospective AI note: The risk of intellectual outsourcing.
Harder questions
Retrospective AI note: Harder questions.
The education of judgment
Retrospective AI note: The education of judgment.
Better than a knowledge base
Retrospective AI note: Better than a knowledge base.
The value of tacit knowledge
Retrospective AI note: The value of tacit knowledge.
A new organisational memory
Retrospective AI note: A new organisational memory.
The role of internal critics
Retrospective AI note: The role of internal critics.
The organisation learns slowly
Retrospective AI note: The organisation learns slowly.
The implementation gap
Retrospective AI note: The implementation gap.
A richer standard of value
Retrospective AI note: A richer standard of value.
The economics of patience
Retrospective AI note: The economics of patience.
Time as a resource
Retrospective AI note: Time as a resource.
Reliability becomes worth waiting for
Retrospective AI note: Reliability becomes worth waiting for.
The limit of optimisation
Retrospective AI note: The limit of optimisation.
The value of stopping
Retrospective AI note: The value of stopping.
The quality of oversight
Retrospective AI note: The quality of oversight.
The human authority problem
Retrospective AI note: The human authority problem.
The difference between access and autonomy
Retrospective AI note: The difference between access and autonomy.
A question of concentration
Retrospective AI note: A question of concentration.
The hidden infrastructure
Retrospective AI note: The hidden infrastructure.
Price changes imagination
Retrospective AI note: Price changes imagination.
The cost of intelligence
Retrospective AI note: The cost of intelligence.
May’s lesson
Retrospective AI note: May’s lesson.
The story behind the system
Retrospective AI note: The story behind the system.
A wider field of creation
Retrospective AI note: A wider field of creation.
The role of mentorship
Retrospective AI note: The role of mentorship.
The new apprenticeship
Retrospective AI note: The new apprenticeship.
Context as organisational memory
Retrospective AI note: Context as organisational memory.
The future of software teams
Retrospective AI note: The future of software teams.
The problem of drift
Retrospective AI note: The problem of drift.
Long-horizon intelligence
Retrospective AI note: Long-horizon intelligence.
Claude 4 and sustained work
Retrospective AI note: Claude 4 and sustained work.
Video gains sound
Retrospective AI note: Video gains sound.
AI becomes multimedia infrastructure
Retrospective AI note: AI becomes multimedia infrastructure.
A new production function
Retrospective AI note: A new production function.
The question of substitution
Retrospective AI note: The question of substitution.
The meaning of intelligence
Retrospective AI note: The meaning of intelligence.
The limits of policy language
Retrospective AI note: The limits of policy language.
Safety as operational practice
Retrospective AI note: Safety as operational practice.
The safety threshold rises
Retrospective AI note: The safety threshold rises.
The political economy of capability
Retrospective AI note: The political economy of capability.
Learning to direct systems
Retrospective AI note: Learning to direct systems.
The next educational divide
Retrospective AI note: The next educational divide.
Depth over speed
Retrospective AI note: Depth over speed.
The research agent question
Retrospective AI note: The research agent question.
Freshness is not truth
Retrospective AI note: Freshness is not truth.
Web search becomes an API primitive
Retrospective AI note: Web search becomes an API primitive.
Discovery still needs institutions
Retrospective AI note: Discovery still needs institutions.
AI for science
Retrospective AI note: AI for science.
The value of interruption
Retrospective AI note: The value of interruption.
Better than seamless
Retrospective AI note: Better than seamless.
The architecture of trust
Retrospective AI note: The architecture of trust.
Connected intelligence
Retrospective AI note: Connected intelligence.
April’s lesson
Retrospective AI note: April’s lesson.
Personal AI arrives as a product category
Retrospective AI note: Personal AI arrives as a product category.
The hard work of standards
Retrospective AI note: The hard work of standards.
A personal principle
Retrospective AI note: A personal principle.
The source of originality
Retrospective AI note: The source of originality.
Authorship after generation
Retrospective AI note: Authorship after generation.
The media environment shifts again
Retrospective AI note: The media environment shifts again.
Image generation becomes infrastructure
Retrospective AI note: Image generation becomes infrastructure.
The public-interest standard
Retrospective AI note: The public-interest standard.
The politics of access
Retrospective AI note: The politics of access.
A new professional standard
Retrospective AI note: A new professional standard.
The return of craftsmanship
Retrospective AI note: The return of craftsmanship.
Delegation needs boundaries
Retrospective AI note: Delegation needs boundaries.
The tool-use threshold
Retrospective AI note: The tool-use threshold.
Reasoning models become tool users
Retrospective AI note: Reasoning models become tool users.
Context is not comprehension
Retrospective AI note: Context is not comprehension.
GPT‑4.1 makes context practical
Retrospective AI note: GPT‑4.1 makes context practical.
The role of refusal
Retrospective AI note: The role of refusal.
The personal relationship
Retrospective AI note: The personal relationship.
The danger of generic intelligence
Retrospective AI note: The danger of generic intelligence.
A different startup thesis
Retrospective AI note: A different startup thesis.
The capability stack
Retrospective AI note: The capability stack.
The infrastructure beneath freedom
Retrospective AI note: The infrastructure beneath freedom.
Openness has conditions
Retrospective AI note: Openness has conditions.
The strategic value of openness
Retrospective AI note: The strategic value of openness.
Llama 4 and open capability
Retrospective AI note: Llama 4 and open capability.
Learning as agency
Retrospective AI note: Learning as agency.
The university question
Retrospective AI note: The university question.
Education enters the next phase
Retrospective AI note: Education enters the next phase.
A system of systems
Retrospective AI note: A system of systems.
March’s lesson
Retrospective AI note: March’s lesson.
The difference between capability and wisdom
Retrospective AI note: The difference between capability and wisdom.
A new creative discipline
Retrospective AI note: A new creative discipline.
The visual epistemology problem
Retrospective AI note: The visual epistemology problem.
The end of single-modal assumptions
Retrospective AI note: The end of single-modal assumptions.
The convergence of mediums
Retrospective AI note: The convergence of mediums.
Thinking becomes a product feature
Retrospective AI note: Thinking becomes a product feature.
Intellectual independence
Retrospective AI note: Intellectual independence.
Research as a relationship
Retrospective AI note: Research as a relationship.
The problem with polished synthesis
Retrospective AI note: The problem with polished synthesis.
The source trail
Retrospective AI note: The source trail.
AI gains web access
Retrospective AI note: AI gains web access.
A new information environment
Retrospective AI note: A new information environment.
The importance of provenance
Retrospective AI note: The importance of provenance.
Coordination as a moat
Retrospective AI note: Coordination as a moat.
New forms of management
Retrospective AI note: New forms of management.
The future of work is organisational
Retrospective AI note: The future of work is organisational.
A system should be interruptible
Retrospective AI note: A system should be interruptible.
Observability matters
Retrospective AI note: Observability matters.
From chat to workflow
Retrospective AI note: From chat to workflow.
The agent stack arrives
Retrospective AI note: The agent stack arrives.
The moral weight of defaults
Retrospective AI note: The moral weight of defaults.
Personalisation and power
Retrospective AI note: Personalisation and power.
What systems remember
Retrospective AI note: What systems remember.
The advantage of proximity
Retrospective AI note: The advantage of proximity.
More output, harder selection
Retrospective AI note: More output, harder selection.
The new operating model
Retrospective AI note: The new operating model.
Better than benchmark theatre
Retrospective AI note: Better than benchmark theatre.
The evaluation problem
Retrospective AI note: The evaluation problem.
The test of usefulness
Retrospective AI note: The test of usefulness.
Beyond the model
Retrospective AI note: Beyond the model.
February’s lesson
Retrospective AI note: February’s lesson.
GPT‑4.5 and two paths to intelligence
Retrospective AI note: GPT‑4.5 and two paths to intelligence.
Architecture still matters
Retrospective AI note: Architecture still matters.
Software as delegation
Retrospective AI note: Software as delegation.
Claude Code changes the workflow
Retrospective AI note: Claude Code changes the workflow.
The next wave of builders
Retrospective AI note: The next wave of builders.
The role of friction
Retrospective AI note: The role of friction.
A different kind of education
Retrospective AI note: A different kind of education.
The craft of judgment
Retrospective AI note: The craft of judgment.
The scarcity of attention
Retrospective AI note: The scarcity of attention.
More intelligence, more noise
Retrospective AI note: More intelligence, more noise.
Proof of work returns
Retrospective AI note: Proof of work returns.
The future of expertise
Retrospective AI note: The future of expertise.
A public-interest test
Retrospective AI note: A public-interest test.
Capability is becoming geopolitical
Retrospective AI note: Capability is becoming geopolitical.
The politics of infrastructure
Retrospective AI note: The politics of infrastructure.
Context without surveillance
Retrospective AI note: Context without surveillance.
The right kind of help
Retrospective AI note: The right kind of help.
Learning versus completion
Retrospective AI note: Learning versus completion.
Better educational systems
Retrospective AI note: Better educational systems.
The hidden curriculum
Retrospective AI note: The hidden curriculum.
Technical power, social consequence
Retrospective AI note: Technical power, social consequence.
The question of dependence
Retrospective AI note: The question of dependence.
A new managerial skill
Retrospective AI note: A new managerial skill.
The coordination problem
Retrospective AI note: The coordination problem.
What agents reveal
Retrospective AI note: What agents reveal.
A deeper kind of product work
Retrospective AI note: A deeper kind of product work.
The architecture of accountability
Retrospective AI note: The architecture of accountability.
January’s lesson
Retrospective AI note: January’s lesson.
More than automation
Retrospective AI note: More than automation.
A new interface for trust
Retrospective AI note: A new interface for trust.
Better oversight
Retrospective AI note: Better oversight.
The necessity of reversibility
Retrospective AI note: The necessity of reversibility.
Agentic risk
Retrospective AI note: Agentic risk.
The unit of work changes
Retrospective AI note: The unit of work changes.
Delegation is a design problem
Retrospective AI note: Delegation is a design problem.
Operator makes delegation concrete
Retrospective AI note: Operator makes delegation concrete.
The cost curve matters
Retrospective AI note: The cost curve matters.
Openness is not neutrality
Retrospective AI note: Openness is not neutrality.
DeepSeek changes the economics
Retrospective AI note: DeepSeek changes the economics.
The open question
Retrospective AI note: The open question.
What remains scarce
Retrospective AI note: What remains scarce.
Intelligence without judgment
Retrospective AI note: Intelligence without judgment.
Better than speed
Retrospective AI note: Better than speed.
The cost of convenience
Retrospective AI note: The cost of convenience.
A new kind of literacy
Retrospective AI note: A new kind of literacy.
The human signature
Retrospective AI note: The human signature.
The institutional lag
Retrospective AI note: The institutional lag.
More than productivity
Retrospective AI note: More than productivity.
Better questions, higher stakes
Retrospective AI note: Better questions, higher stakes.
The bottleneck moves
Retrospective AI note: The bottleneck moves.
A system needs a theory
Retrospective AI note: A system needs a theory.
The context economy
Retrospective AI note: The context economy.
The value of interpretation
Retrospective AI note: The value of interpretation.
A better definition of leverage
Retrospective AI note: A better definition of leverage.
Intelligence is becoming infrastructure
Retrospective AI note: Intelligence is becoming infrastructure.
The new asymmetry
Retrospective AI note: The new asymmetry.
Capability is not adoption
Retrospective AI note: Capability is not adoption.
The question changes
Retrospective AI note: The question changes.