the deck
Mindset · Dan Koe · future/proof · 23 Dec 2025

How to fix your entire life in one day

Resolutions fail 80–90% of the time because people change their actions and never touch the thing underneath them — who they believe they are. Seven ideas, then a one-day pen-and-paper protocol.

Reformatted for internal use · animations are CSS & SVG, no external assets
Idea I

You aren't there because you aren't the person who'd be there

A resolution has two requirements. People only ever do the first.

  • 2Change your actions toward the goal. Second-order, least important — the part people white-knuckle for three weeks.
  • 1Change who you are so behaviour just follows. The bodybuilder doesn't grind to eat well — he'd have to grind to eat badly.

Build a great life on a rotting foundation and it collapses back to baseline by February. You must own the lifestyle that creates the outcome long before the outcome arrives — and keep it afterward.

ROTTING BASE SOLID BASE
Same house. The base decides whether it stands.

What this section means

Stop asking "what habits do I need?" Ask "what kind of person has these by default?" — then become that first. Habits that don't serve the new identity should start to feel disgusting, because you can now see what they compound into.

Idea II

You aren't there because you don't actually want to be

Trust only movement. Life happens at the level of events, not of words.— Alfred Adler

All behaviour is goal-oriented — but most goals are unconscious, and some quietly harm you while sounding respectable.

  • A"I procrastinate because I lack discipline" → real goal: avoid the judgment of finishing and shipping.
  • B"I stay in the dead-end job because I'm not a risk-taker" → real goal: safety, predictability, not looking like a failure to people who also stayed.

A goal isn't a to-do line. A goal is a lens — it decides which information, ideas and resources you even notice. Real change means swapping the lens, not the checklist.

STATED GOAL — the lens you admit to hidden goal — the one your behaviour serves
Watch the behaviour, not the words. It always lands somewhere.

What this section means

Before setting any resolution, name the unconscious payoff of your current behaviour. If you can't say what staying stuck is protecting, your hidden goal will quietly override the shiny one.

Idea III

You aren't there because you're afraid to be there

Identity is a loop that has run on you since childhood. Each pass hardens it.

1 · want a goal 2 · perceive via it 3 · notice "relevant" only 4 · act, get feedback 5 · repeat → automatic 6 · "I am the type who…" 7 · defend the identity 8 · it shapes new goals CUT HERE (6→7)
Threaten someone's identity and they go into fight-or-flight — same as a threat to the body.

Once physical survival is handled (easy today), you start surviving on the ideological level — protecting and reproducing your mind. Challenge a belief someone is fused with and they feel physically slapped. Same mechanism whether it's a politics, a religion, or seeing yourself as "a gamer" who wouldn't do the harder thing.

What this section means

The block isn't laziness — it's a defended self. Change requires letting a version of you die, and paying the social cost of no longer being that person. Say the identity statement out loud: "I am the type of person who…"

Idea IV

The life you want lives at a specific level of mind

The mind evolves through predictable stages (Maslow, Loevinger / Cook-Greuter, Spiral Dynamics, Koe's own Human 3.0). The 80/20 of nine ego stages:

MOST READERS HOVER 4 – 8 (a huge gap) 1 Impulsive2 Self-Protect3 Conformist 4 Self-Aware5 Conscient.6 Individualist 7 Strategist8 Construct-Aware9 Unitive
Readers near 8 are here to learn or pass time. Readers near 4 are here for a change.
  • 3Conformist — you are your group; its rules feel like reality.
  • 4Self-Aware — you notice an inner life that doesn't match the outside, but don't know what to do with it.
  • 5Conscientious — you build your own principles and hold yourself to them.
  • 6Individualist — you see your principles were shaped by context, and hold them loosely.
  • 7Strategist — you work systems while watching your own blind spots.

What this section means

Your stage isn't the point — the pattern of moving between stages is, and it's the same every time: tension → disorientation → reorganisation at a higher level. The protocol runs that pattern on purpose.

Idea V

Intelligence is the ability to get what you want out of life

The only real test of intelligence is if you get what you want out of life.— Naval Ravikant

Three ingredients: agency (likelihood you act), opportunity (worth of the goal you can reach — digital abundance is right there), and intelligence (ability to convert the other two). Read through cybernetics — Greek kybernetikos, "good at steering" — intelligence is a feedback loop:

Goal Act Sense Compare iterate · persist a ship blown off course keeps correcting toward the destination
Low intelligence gets stuck on a problem. High intelligence assumes any problem yields on a long enough timescale.

Most people's goals were assigned: go to school, get the job, get offended, play victim, retire at 65 — a known path that doesn't work. To get smarter: reject the known path, dive into the unknown, set higher goals, embrace the chaos, study nature's general principles, become a deep generalist.

What this section means

Treat your life like a control system. Pick a real target, make a move, measure honestly, adjust, repeat. The only true failure mode is refusing to learn from the last loop.

Idea VI · the protocol

Launch into a new life — in one day

Pen, paper, an open mind. No AI for the contemplation. The arc: reach dissonance → move through uncertainty → hit discovery, where clarity is loud enough that distractions lose their pull.

Dissonancefed up Uncertaintyexperiment / wander Discovery6 yrs of progress in 6 months
Morning15–30 min

Psychological excavation → Vision & Anti-Vision

Build a new lens. It won't fit yet — good.
PainWhat dull, persistent dissatisfaction have you learned to tolerate? (If you don't hate it, you'll tolerate it.)
PainWhat do you complain about repeatedly but never change? Write your 3 most-voiced complaints of the past year.
PainFor each: what would someone watching your behaviour conclude you actually want?
PainWhat truth about your current life would be unbearable to admit to someone you deeply respect?
Anti-visionIf nothing changes for 5 years, describe an average Tuesday — bed, body, first thought, who's around, 9am–6pm, how you feel at 10pm.
Anti-visionNow 10 years. What did you miss? Which doors closed? Who gave up on you? What's said about you when you're not in the room?
Anti-visionEnd of life, the safe version, pattern never broken. What was the cost? What did you never let yourself feel, try, or become?
Anti-visionWho is already living that future — 5, 10, 20 years down the same track? What do you feel imagining you become them?
Anti-visionWhat identity would you have to give up to change? What does it cost you socially to stop being that person?
Anti-visionMost embarrassing reason you haven't changed — the one that sounds weak, scared or lazy rather than reasonable.
Anti-visionIf your behaviour is self-protection: what exactly are you protecting, and what is it costing you?
Vision MVPForget practicality. Snap your fingers — a different life in 3 years. What's an average Tuesday, same detail?
Vision MVPWhat would you have to believe about yourself for that life to feel natural? "I am the type of person who…"
Vision MVPOne thing you'd do this week if you were already that person.
Daytimeinterrupts

Interrupt autopilot → break unconscious patterns

Set these as phone reminders now, question included. Random, non-conflicting, ideally during commutes / walks.
What am I avoiding right now by doing what I'm doing?
If someone filmed the last two hours, what would they conclude I want from my life?
Am I moving toward the life I hate, or the life I want?
What's the most important thing I'm pretending isn't important?
What did I do today out of identity protection rather than genuine desire?
When did I feel most alive today? When did I feel most dead?
WalkingWhat changes if I stop needing people to see me as [that identity]?
WalkingWhere am I trading aliveness for safety?
WalkingWhat's the smallest version of the person I want to become that I could be tomorrow?
Eveningsynthesise

Synthesise insight → enter a season of progress

TrueAfter today, what feels most true about why you've been stuck?
EnemyName the actual enemy — not circumstances, not people. The internal pattern or belief running the show.
Anti-visionOne sentence: what you refuse to let your life become. It should make you feel something.
VisionOne sentence: what you're building toward, knowing it will evolve.
1-year lensWhat would have to be true in one year to know you broke the old pattern? One concrete thing.
1-month lensWhat would have to be true in one month for the one-year lens to stay possible?
Daily lens2–3 actions you can timeblock tomorrow that the person you're becoming would simply do.

Why goals here aren't normal goals

A goal is a point of view you swap into to enter the right state for the next action — not a finish line. There is no finish line; enjoyment is in the progress.

Idea VII

Turn your life into a video game

The optimal state of inner experience is one in which there is order in consciousness… skills match the opportunities for action.— Mihaly Csikszentmihalyi

On one page, write six components. Together they build a small world you have no choice but to get obsessed with.

VISION constraints — the rules daily levers — the quests 1-month project — the boss fight 1-year goal — the mission anti-vision — what's at stake if you lose
Shiny objects fly in and bounce off. The more you play, the stronger the field.
how you win

Vision

The ideal life you think you want — improved as you go. Valid until the game evolves.

what's at stake

Anti-vision

The bane of your existence. What happens if you lose or quit.

the mission

1-year goal

Your sole priority. Is it visibly closer to the life you want?

the boss fight

1-month project

What to learn / build / acquire. Where you gain XP and loot.

the quests

Daily levers

Priority, needle-moving tasks that unlock new opportunities.

the rules

Constraints

What you won't sacrifice. Limits that force creativity.

What this section means

Games are the poster child for flow: clear goal, clear stakes, matched difficulty, tight feedback. Reverse-engineer those four into your life and obsession stops requiring discipline.

The concentric circles become who you are

Dissonance made you fed up. Uncertainty made you experiment. Discovery made distractions weightless.

The more you play the game, the stronger the field — and soon you wouldn't have it any other way.

— Dan Koe · future/proof · 23 Dec 2025
Apply it · 16 years in tech

16 years in. AI engineering. UK. Target: a frontier lab.

The essay's claim: you don't get to Anthropic or OpenAI by applying harder. You become the person a frontier lab already wants, then the offer follows. Framework applied — not insider hiring advice.

Idea I → applied

Be the hire before you're hired

The labs don't hire "a senior engineer with 16 YOE." They hire someone visibly already doing frontier-shaped work — evals, inference & performance, interpretability tooling, agent harnesses, red-teaming — in public.

The tell from the essay: if reading a paper, writing, or building in public feels like grind you force yourself into, you don't have the lifestyle that produces the outcome. Someone in that job does those by default — they'd have to grind to stop.

  • +16 YOE helps: you own ambiguous, underspecified problems and lead without hand-holding.
  • 16 YOE traps: the identity "I'm past proving myself." You'll share rooms with people who've published since their PhD. Proof-of-work resets.

The move

Aim to be someone who gets a warm referral from inside — not someone optimising a cold application.

Idea II → applied

Check the hidden goal

"I want to work at a frontier lab" — now watch the behaviour. Are your evenings already going to evals, model-behaviour writeups, PRs to repos the labs read (vLLM, transformers, the labs' SDKs, MCP servers)? If not, the stated goal may be serving a hidden one: prestige, a cleaner escape from the current job, external validation.

  • AAnthropic — safety framing, research-adjacent product, an unusually heavy writing culture. If you can't write clearly, that's the gap.
  • OOpenAI — scale, shipping velocity, huge product surface. Bias toward people who move fast and ship.

The move

If the real goal is "leave my current job," a frontier lab is a slow, low-odds route. Name it — your answer to "why this lab" is itself interview signal.

Idea III → applied

The identity that has to die

  • 1"I'm an implementer, not a researcher." Labs hire strong engineers into research-facing roles. The wall is self-imposed.
  • 2"16 YOE = senior = I skip the proof-of-work." You don't.
  • 3"Posting in public is cringe / hype-chasing." Shipping evals and being wrong in front of peers is the social cost the essay says you must pay.
  • 4Going from big-fish to relative beginner in a room of specialists.

Write the sentence

"I am the type of person who ships public work on frontier-AI problems every week, and talks to the people building them." Then notice what it costs you socially to become that.

Ideas IV + V → applied

Swap the lens, then run the loop

Move off the assigned ladder (hit milestones, effort → reward, wait to be promoted) onto the Strategist lens: play the actual hiring system — referrals + visible artefacts + being known to people inside, not LeetCode plus cold apply.

settarget ship sensesignal adjust × 15–20
Most people run this loop once, get auto-rejected, and quit. That is the whole filter.
  • GGoal: final-round or offer at a frontier lab within 12 months (London — both have offices, no move needed).
  • AAct: ship one artefact.
  • SSense: recruiter replies, referral offers, callback rate, repo stars, whether insiders reply to your DMs.
  • CCompare & adjust: one rejection is not a stop. Change the artefact, the pitch, the target team, the referrer — and loop.
Idea VI → applied

Your 1-day protocol, pre-filled

Run the day as the essay lays it out. Below is a starting fill to react against and rewrite — not to copy.

Anti-visionIt's 2031. Still a senior AI engineer at a consultancy / product company. You follow frontier work on X, forward launch posts into Slack, and tell yourself you "could have applied." The gap to the field widened every year you didn't ship.
Vision MVPYou work on model capability or safety at a frontier lab. Your name is on evals / systems / research that shipped. Reading papers and writing clearly are just what your days contain.
Identity"I'm the type of person who publishes work on frontier-AI problems every week and is in conversation with people building them."
1-year lensA final-round at Anthropic or OpenAI — or a portfolio (an eval suite + 2 technical writeups + a merged PR in a lab-relevant repo) strong enough that a referral is trivial to ask for.
1-month projectPick one boss fight: (a) a public eval / benchmark on a capability you know deeply, (b) a clean reproduction of a recent paper + writeup, (c) a real contribution to a repo the labs use, (d) an agent harness that does something non-trivial and is honest about its failure modes.
Daily levers2h deep work on the project · 1 paper skimmed to notes · 1 public artefact per week · 1 outreach to someone inside per week.
ConstraintsKeep the paycheck — don't quit yet · protect health and family time · ship quality, never spam.
Idea VII → applied

The board

how you win

Vision

Engineer at a frontier lab, on work that ships.

what's at stake

Anti-vision

Permanent spectator — watching the field from Twitter.

the mission

1-year goal

Final round / offer, or a referral-ready portfolio.

the boss fight

1-month project

The one flagship artefact — eval, repro, OSS PR, or harness.

the quests

Daily levers

2h build · 1 paper · weekly artefact · weekly outreach.

the rules

Constraints

Keep the job. Keep your health. Ship quality, not noise.

the two paths

Known path vs. real path

Doesn't work

  • Grind LeetCode for months as the main plan
  • Cold-apply through the careers page and wait
  • "Get more senior" at the current job and hope it transfers
  • Read about the field privately, share nothing
  • One application, one rejection, quit

Works

  • Referral from someone inside — network deliberately: London AI meetups, OSS maintainers, ex-colleagues who moved
  • Public artefacts matched to a specific team's work
  • Merged PRs in repos the labs actually use
  • A blog with real depth on model behaviour (Anthropic runs on writing — practise it)
  • 15–20 loops: iterate the artefact, pitch, target, referrer

The whole thing in one line

Stop trying to be picked. Become the person for whom the pick is obvious — then it's just loops.

Dissonance → uncertainty → discovery

Get fed up enough with being a spectator. Experiment through the messy middle. Then ship loudly enough that the frontier stops feeling far away.

framework: Dan Koe · application: yours to run
Apply it · starting fresh

New to tech, aiming at an AI engineering job

The essay says no credential is the gate and "applying harder" isn't the lever. Become someone who already builds and ships AI things in public — the junior offer follows. Framework applied, not a syllabus.

Idea I → applied

Be a builder before you're hired

Entry AI engineering today ≈ a software engineer who's fluent with LLM APIs, retrieval, evals, and prompt / agent design — and can ship. SWE fundamentals still decide most interviews; the LLM layer sits on top.

Junior offers go to people with a visible trail: 3–4 deployed projects, real commit history, short writeups. Not certificates. The essay's tell: if building only happens when you force it — and mostly you watch tutorials — you don't have the identity yet. Someone in that job builds because they can't not.

The move

Become the candidate a hiring manager can literally see working: deployed links, a README, a build log of what broke.

Idea II → applied

Check the hidden goal

"I want an AI engineering job." Now watch the behaviour. Are you shipping with the APIs every week, or collecting courses? Tutorial-hell is often an unconscious goal — feel productive, avoid the judgment of shipping something real and flawed.

  • ?Do you want AI specifically, or a stable, well-paid tech job? Both are fine — but they lead to different plans and different applications. Say which.
  • ?Is the next course genuinely the missing piece, or the place you go to feel safe instead of shipping?

The move

Finish one real, deployed project before buying the next course. The course is usually the hiding place.

Idea III → applied

The identity that has to die

  • 1"I'm not technical / not a real engineer yet." You become one by shipping, publicly, before you feel ready.
  • 2"One more course and then I'll start." There is no ready.
  • 3"I'll build in public once I'm good enough." Being visibly mid is exactly the cost the essay says to pay.
  • 4"I'm too late / too old to switch." That's the protocol's "most embarrassing reason" question — answer it honestly.

Write the sentence

"I'm someone who ships a small AI project every two weeks and writes up what broke." Then notice what it costs you socially to say that out loud.

Ideas IV + V → applied

Run your own system, then run the loop

Move off Conformist (follow the assigned curriculum, wait for a credential to grant permission) to running your own system: learn what your project needs, judge your own output, set your own milestones. The credential is not the gate — a portfolio is.

Your opportunity capital is low right now, so loop volume matters most.

settarget shipsmall sensesignal adjust × 30–50
Most people send ~10 applications, hear nothing, decide "the market is bad", and stop. That's the filter.
  • GGoal: first paid AI-adjacent role in ~9–12 months — junior AI engineer, LLM-focused SWE, AI product engineer, applied-AI at a non-tech company, forward-deployed, AI dev-rel.
  • AAct: ship a small thing.
  • SSense: replies, interview invites, GitHub stars, feedback on posts, referral offers.
  • CAdjust: widen the targets, rewrite the résumé to the job description, swap the project.
Idea VI → applied

Your 1-day protocol, pre-filled

Run the day as the essay lays it out. Starting fill to rewrite in your own words:

Anti-visionThree years on. Still "preparing." A dozen half-finished courses, no deployed project, applications sent in bursts when motivation spikes. You've decided the market is the problem.
Vision MVPJunior AI engineer at a real company. You ship features that use LLMs, you're learning on the job, you're paid to get better.
Identity"I ship a small AI project every two weeks and write about what I learned."
1-year lensA paid role — or a portfolio of 3–4 deployed projects + 5 writeups + 1 merged open-source PR that makes recruiters reply.
1-month projectBuild one genuinely useful thing end-to-end and deploy it: a RAG app over a corpus you actually care about; an agent that does one real task reliably; an eval harness for a small model; an LLM-powered tool you'd use yourself. Demo link, README, honest writeup of what broke.
Daily levers1–2h building · learn one concept by using it (not just watching) · ship or post weekly · 3 tailored applications + 1 human outreach per week.
ConstraintsKeep income coming (job / part-time) · no new course until this project ships · protect sleep.
Idea VII → applied

The board

how you win

Vision

Junior AI engineer, paid to learn on real systems.

what's at stake

Anti-vision

Perpetual "preparing" — courses, no shipped work.

the mission

1-year goal

A paid role, or a recruiter-magnet portfolio.

the boss fight

1-month project

One deployed, useful project + honest writeup.

the quests

Daily levers

1–2h build · 1 concept applied · weekly ship · 3 apps + 1 outreach.

the rules

Constraints

Keep income. No new course till this ships. Protect sleep.

the two paths

Known path vs. real path

Doesn't work

  • Collect certificates and wait to feel ready
  • Tutorial-hell: watch, don't build
  • Mass-spray 200 identical applications
  • Learn entirely in private
  • Assume a bootcamp will place you
  • Only apply to labs and FAANG

Works

  • 3–4 deployed projects that solve a real problem
  • A short build log for each — what broke, what you'd change
  • One PR to a library you use (an SDK, an eval tool, a framework)
  • 3 applications a week, tailored to the job description
  • One human referral via communities — local meetups, OSS, Discords
  • Wide funnel: applied-AI roles at ordinary companies, not just labs
  • Rewrite the résumé from each round of feedback

The whole thing in one line

Stop preparing to be picked. Ship small, ship visibly, run the loop 40 times — someone replies.

Dissonance → uncertainty → discovery

Get sick enough of "preparing." Build messy things through the middle. Then ship loud enough that someone on the other side replies.

framework: Dan Koe · application: yours to run