Jeff Bezos on Amazon, Blue Origin, and Decision-Making
In his first long-form conversation of this kind, Jeff Bezos moves from the physics and manufacturing of Blue Origin’s New Glenn rocket and his vision for large-scale space settlement to the decision-making architecture and writing discipline — two-way doors, disagree and commit, the six-page memo — he built at Amazon and is now installing at Blue Origin.
Key ideas
- Rockets love to be big, until manufacturing makes them hate it. Avionics and turbo-pump tolerances scale favourably with size, so a large rocket carries proportionally less parasitic mass and looser manufacturing tolerances than a small one — but building large structures (fairings, tanks, launch pads driven 50–150 feet into the ground) is disproportionately hard. Bezos treats rate manufacturing, not the first vehicle, as the true engineering challenge behind New Glenn.
- Two-way doors versus one-way doors govern decision speed. A reversible decision should be delegated and made fast; an irreversible one — Amazon Prime’s launch, a rocket’s propellant choice — should be elevated to senior people and deliberately slowed down. See Two-Way Door Decisions.
- Disagree and commit replaces compromise and attrition. Compromise (splitting the difference on a factual question) and war of attrition (whoever is more stubborn wins) both fail to find truth. The alternative: voice a genuine objection fully, then, once truly heard, commit to the decision made. See Disagree and Commit.
- A skeptical view of proxies defends against organisational decay. Metrics adopted for good reasons quietly stop tracking the truth they were meant to represent as conditions shift — Bezos’s own customer-service call, which found a ten-minute wait against a metric showing under sixty seconds, is his central illustration. Fighting this decay is one of the defences against what he calls ‘Day 2’: stasis, irrelevance, and death.
- A crisp document produces a productively messy meeting. Amazon and Blue Origin meetings open with a silent 30-minute reading of a six-page narrative memo rather than a PowerPoint deck, because full sentences cannot hide sloppy thinking the way bullet points can, and because a persuasion tool is the wrong instrument for a truth-seeking discussion.
Content
New Glenn and the physics of scale
New Glenn is a heavy-lift launch vehicle carrying roughly 45 metric tons to low Earth orbit on about 3.9 million pounds of thrust, its reusable booster powered by seven BE-4 engines burning liquefied natural gas and liquid oxygen in an oxygen-rich staged-combustion cycle ‘pioneered by the Russians’; the expendable upper stage uses two BE-3U hydrogen engines, trading hydrogen’s low density (and correspondingly large tanks) for its high specific impulse where it matters most. Bezos’s recurring line, ‘rockets love to be big’, reflects two physical facts: avionics and guidance systems are roughly constant mass regardless of vehicle size, so they are a heavy tax on a small rocket and a rounding error on a large one, and turbo-pump manufacturing tolerances — the unavoidable gap between a rotating impeller and its housing — shrink in relative terms as pumps scale up. The counterweight is manufacturing: large carbon-composite fairings, aluminium-lithium tanks joined by friction-stir welding, and launch infrastructure requiring dozens of pilings driven 50 to 150 feet into the ground are disproportionately hard to build at any size. Bezos distinguishes producing a first article — hard, but not the real challenge — from rate manufacturing: sustaining a cadence of 24 launches a year requires an upper stage every two weeks and an engine every week, which he judges at least as difficult as the original vehicle design.
Decisiveness and the two-way/one-way door test
Bezos wants Blue Origin to become, in his words, ‘the world’s most decisive company across any industry’, the direct analogue of Amazon’s ambition to be the most customer-obsessed. The mechanism he uses to explain how large organisations stay fast is the reversibility test: most decisions are two-way doors — cheap to reverse if wrong — and should be delegated to individuals or small teams and made quickly. A minority are one-way doors — consequential and difficult to reverse — and these should be elevated to the most senior people, who should deliberately slow the process down. As Amazon’s CEO, Bezos describes his own role on these as ‘chief slow-down officer’. He offers Amazon Prime’s launch and the choice of New Glenn’s propellants (liquefied natural gas for the booster, hydrogen for the upper stage — ‘quick-drying cement’, in his phrase) as decisions that function as one-way doors even where reversibility is, strictly, possible. Full treatment: Two-Way Door Decisions.
Truth-telling, escalation, and disagree and commit
Bezos rejects two common ways organisations resolve disagreement. Compromise — averaging two guesses at a room’s ceiling height instead of measuring it — is low-effort but abandons the question of what is actually true. War of attrition — whichever party is more stubborn or has more energy prevails — is worse, because it is demoralising and no closer to truth. His prescribed alternative is escalation: disagreeing parties who respect each other but cannot agree should ask a shared superior to decide, rather than exhausting each other. Where a decision has to be made by someone more senior, the disagree and commit principle applies: the objecting party is obligated to voice their genuine concern, all the way up the chain if necessary, and once they are truly heard — not merely tolerated — they commit fully to the decision made, rather than complying grudgingly. Bezos also structures meetings to protect dissent before it happens: he speaks last, on the reasoning that even confident, high-judgement people will second-guess their own view once a senior figure has stated theirs first. Full treatment: Disagree and Commit.
Space settlement and O’Neill colonies
Bezos’s long-run vision is a solar system supporting a trillion humans living in large rotating O’Neill colonies — built from lunar regolith and asteroid-belt material, spun for Earth-normal gravity, and sited in Earth-vicinity orbits rather than on planetary surfaces, which he judges too small and fragile to expand into indefinitely. The case for urgency is thermodynamic rather than purely exploratory: continued growth in energy use per capita, which Bezos regards as essential to rising living standards, is incompatible with a finite planet, so heavy industry has to move off-world to preserve Earth itself — which he calls ‘the good planet’ after robotic probes found nothing comparable elsewhere in the solar system. He frames Blue Origin’s mission as building ‘heavy infrastructure’ — the space-age analogue of the credit-card payment network and postal service Amazon inherited, and the dial-up telephone network the early internet piggybacked on — so that a future generation of space entrepreneurs can found ventures ‘in a dorm room’ the way two people could once start an internet company without having to build the underlying network themselves. Nearer term, Blue Ring (a servicing spacecraft offering power, compute, thermal management, and communications to payloads — ‘a little bit like Amazon Web Services… for space’) and the Mark 1 and Mark 2 lunar landers are the concrete steps toward that infrastructure, alongside in-situ work extracting oxygen and manufacturing solar cells from lunar regolith.
Day 1, customer obsession, and the skeptical view of proxies
Bezos describes ‘Day 1 thinking’ — a phrase he first used in his 1997 shareholder letter — as a discipline of renewal: starting fresh each day, revisiting even long-settled principles (‘unless you know a better way’ is appended to every list of team tenets), and never being trapped by self-consistency or history. The alternative, ‘Day 2’, is ‘stasis, followed by irrelevance, followed by excruciating painful decline, followed by death’. Of the defences against Day 2 he lists in his final shareholder letter as CEO — customer obsession, a skeptical view of proxies, eager adoption of external trends, high-velocity decision-making — he elaborates most on proxies: a metric adopted for good reason (such as customer contacts per unit sold, as a stand-in for customer happiness) can quietly stop tracking the truth it was invented to represent as conditions shift, and an organisation must periodically ask whether the proxy still serves the goal it was built for. His illustration: dialling Amazon’s own customer-service line during a weekly business review and waiting over ten minutes, despite a metric showing under sixty seconds — proof the company was measuring the wrong thing. His related maxim: ‘when the data and the anecdotes disagree, the anecdotes are usually right’ — not because anecdote trumps data, but because the disagreement itself signals flawed measurement. Distinct from big, decade-stable customer wants (low prices, fast delivery, wide selection), Amazon also tracks ‘paper cuts’ — small, individually low-priority friction points requiring dedicated teams, since teams working on major initiatives never reach them. Bezos offers 1-Click ordering as an invention that eliminated a cluster of such paper cuts.
The six-page memo: crisp document, messy meeting
Amazon and Blue Origin meetings typically open with a silent 30-minute ‘study hall’, in which participants read a six-page narratively structured memo rather than sit through a slideshow. Bezos gives three reasons for preferring memos to PowerPoint: a slide deck is built to persuade, which is the wrong instrument for internal truth-seeking; bullet points can conceal sloppy thinking in a way full sentences with topic sentences cannot; and reading the whole document in advance avoids senior people interrupting with questions the next slide would have answered. A good six-page memo can take two weeks to write, rewrite, and stress-test with colleagues before a meeting. Bezos’s own preferred meeting has ‘a crisp document and a messy meeting’ — the document should be unambiguous, but the discussion itself should wander toward a solution nobody yet knows, since efficiency and genuine invention are frequently at odds.
AI, the 10,000-Year Clock, and mortality
Bezos regards large language models as ‘discoveries’ rather than ‘inventions’ — their capabilities are found through experiment rather than designed in advance, unlike an engineered object such as an aircraft — and states he is net optimistic that powerful AI is more likely to help humanity than harm it, including as a possible hedge against other existential risks, while flagging narrow, non-general AI weapons as a serious near-term danger. The 10,000-Year Clock — a roughly 500-foot mechanical clock built inside a Texas mountain, conceived by Danny Hillis — is presented as a deliberately slow symbol intended to stretch human thinking horizons past the five-year window most institutions default to; Bezos treats long-term thinking as a lever capable of solving problems that are unsolvable on a short horizon. He closes by saying he no longer fears death as he once did, preferring to optimise for health span — full health followed by a quick end — over raw longevity.
Related
- Two-Way Door Decisions — created from this source
- Disagree and Commit — created from this source
- Jeff Bezos — speaker
- Lex Fridman — host