The minimally viable productivity system: three components everyone needs

Executive overview

Most people either suffer from too little organisation — dropping balls, missing deadlines, no progress on meaningful work — or tip into over-optimisation, where life becomes about execution for its own sake. The question worth asking is: what is the minimum set of rules and tools that escapes the first problem without creating the second?

The MVPS (minimally viable productivity system) has exactly three components: task management, workload management, and time control — and even bare-bones versions of each are enough.

The three goals any system must satisfy

• Reduce stress from disorganisation and forgotten commitments
• Build a reputation as someone who can be counted on
• Preserve space to make progress on important, non-urgent work

Task management: a trusted system outside your brain

• Core requirement: obligations are stored somewhere external that you review regularly
• Bare bones: a calendar for time-sensitive items + a legal pad or text file for everything else
• Crossing off items and copying uncrossed ones to a fresh page (bullet journal style) is sufficient
• More advanced: status boards (e.g. Trello) with columns per role and rows per status — waiting on, to discuss, in progress
• The key property: your mind releases anxiety once it trusts that nothing is lost

Workload management: controlling the volume of commitments

• Task management tracks what you've agreed to do; workload management controls how much you agree to
• You need: (1) a way to estimate current workload, (2) a sense of your personal maximum, (3) rules to keep the two in balance
• Pre-scheduling big commitments at the point of agreement forces a reality check — if you can't find the hours, you can't take the work
• Quotas keep recurring obligation types (peer review, advisory calls, committee work) from strangling your schedule
• Project counts set a hard limit: pick a number through experience and stop taking on new work when you hit it
• Advanced: Kanban-style WIP limits — separate what is actively being worked on from what is waiting; nothing in the waiting column gets administrative overhead until it moves

Time control: intention over reaction

• Default mode for most people is reactive: responding to email, Slack, and ambient stress, with distraction filling the gaps
• Bare minimum: a five-minute morning review — check calendar and task list, decide what matters today, identify when you'll do it
• More advanced: multi-scale planning across semester, week, and day
  • Semester plan sets big-picture goals
  • Weekly plan plays chess with the calendar — move and protect blocks for the big rocks
  • Daily time-block plan assigns every hour a job; adjustments are expected but intention is set first

Choosing your own implementation

• These are components, not prescriptions — analogue or digital, simple or complex, any implementation that satisfies the property is valid
• The three components together prevent the worst failures of under-organisation without requiring an optimisation mindset
• Start minimal; add complexity only if a specific problem demands it

Q&A highlights

Passion trap vs purpose trap

• The passion trap: assuming job satisfaction comes from matching job content to pre-existing interests — it doesn't
• The purpose trap: letting a sense of mission blind you to other factors (autonomy, mastery, financial sustainability, lifestyle fit) that make a job worth keeping
• Purpose can be one input into lifestyle-centric planning, not the only one
• Build career capital first; leverage it to shape work toward what resonates

Capturing ideas from podcasts on the go

• Jot timestamps in a notes app and transcribe later
• Voice-dictate emails to yourself mid-workout or mid-commute
• A small field notebook in your pocket for quick timestamp notes

Managing PhD students without email overhead

• Research supports daily 10-minute stand-ups over weekly hour-long check-ins
• Students always know what they're working on; they can't get stuck for more than a day
• Longer one-on-one sessions are reserved for specific identified blockers

Federal workers navigating a values gap

• Distinguish: actively doing something against your values (disqualifying) vs resources for work you care about being reduced (frustrating but different)
• Don't personalise — in large organisations, dramatic exits go unnoticed
• Think in terms of lifestyle-centric trade-offs: the job may be supporting many things beyond its mission
• If change is needed, take the time to make it correctly rather than making a showy exit

Writing a nonfiction book on the side

• Nonfiction is sold before it is written — motivation is the contract, not willpower
• The path: query agents with a one-page letter → agent helps craft a proposal → publisher buys it → then write
• To sell, you need: a compelling idea, a large enough audience that feels they must read it, and credentials to write it
• All three are hard to satisfy simultaneously; use failed queries as diagnostic feedback
• Fiction works differently — you write the book first; self-publishing nonfiction typically reaches almost no one

Case study: career transition and digital reset (Kelly)

• Lifestyle-centric planning led to substitute teaching — not the obvious dream job, but right for the actual constraints
• Paired with an analogue reset: book everywhere, less appealing phone, no algorithmically curated content
• During a family medical crisis: standards should drop temporarily; crisis lifestyle planning has different goals (be the person others count on, prioritise self-care over productivity)
• Return to the bigger vision when the emergency passes

AGI, superintelligence, and the Frankenstein factors

AGI is a quality threshold, not a new capability

• Current language models already write memos, code, and jokes — AGI just means they do these things as well as a competent human
• It has real economic and security consequences, but it is not the Skynet scenario

The four Frankenstein factors required for autonomous risk

1. Understanding — the ability to reason about complex concepts (what language models provide)
2. World modelling — a maintained, updated state of the system's situation
3. Incentives — a value function that drives action
4. Actuation — the ability to affect the external world

• Current AI development is almost entirely focused on understanding; world modelling, incentives, and actuation are left to the humans using the models
• Factors 2–4 are engineered, not trained — they can be explicitly controlled and constrained (as demonstrated by the Cicero diplomacy system, which was programmed not to lie)
• This is intentional AI (IAI): autonomous systems where the dangerous factors are hand-designed and therefore controllable

Why superintelligence may not be computationally feasible

• Most problems are mathematically unsolvable (Turing, 1930s); of those that are solvable, most are computationally intractable
• There is no theoretical basis for assuming intelligence scales indefinitely with compute
• The halting problem is the canonical example: no program can decide whether an arbitrary program halts — this is provably unsolvable

The realistic near-term risk

• Not a sentient system that resists being turned off
• More likely: a supercharged autonomous agent (like a Morris worm) that spirals out of control through actuation before anyone intervenes
• This is a tractable safety problem — limit actuation, keep humans in the loop — and is the right focus for AI safety work in the next few years