Eighty essays, hundreds of patterns, but only a handful of iron laws that compound across domains. The useful stuff survives [[Inversion]]. Flip the problem, change the narrative, strip the hype. What's left standing is what matters. ## What Separates Winners from the Forgettable **Most people fail because they believe in imaginary constraints.** [[High Agency]] is the recognition that if something is physically possible, the only real barrier is stupidity. Shannon didn't beat roulette with luck. The Wright brothers didn't wait for permission. They asked "what prevents flight?" and then systematically removed each barrier. High agency boils down to: clear thinking (mental models that cut through fog), a bias to action (movement beats meditation), and productive disagreeability (questioning authority without being disagreeable). Everyone assumes there are adults in charge who have things figured out. There aren't. That realization is freedom disguised as chaos. [[Effectual Reasoning]] flips the traditional startup playbook. Start with what you have, not where you want to end up. Think in terms of affordable loss, not expected return. Co-create with people building the future instead of competing for scraps of the present. Prediction is astrology. Creation is [[Wright's Law]] playing out in real time. The lottery test is a good filter. If you won a billion dollars tomorrow, what changes? If the answer is "I'd do something else," you're building a bridge to somewhere else. Forrest Heath at Somos Internet would deploy that capital into exactly what he's already building. Elon risked everything on Tesla and SpaceX because Mars requires scale, and scale requires companies. Big problems need institutional force. You either build the institution or the problem stays unsolved. ## Mental Models for Outliers [[Pattern breaking ideas - the key to outlier startups|Pattern-breaking ideas]] come from three forces. Inflections: external shifts that change behavior. Insights: non-obvious truths hiding in plain sight. And the ability to live in the future, building what's missing. Airbnb saw the cultural shift toward trusting strangers plus surplus assets creating value. Tesla saw the energy transition plus the insight that luxury could fund the scale needed for mass production. [[Charlie Munger]]'s line applies: show me the incentives, I'll show you the outcome. Edge cases become the center. Outliers don't follow recipes. They harness forces larger than themselves. [[The Deep Tech Growth Cycle is different]] because atoms resist compression in ways bits don't. Software scales cheaply. Hardware doesn't. Two paths have to merge for commercial success: tech-led innovation (platform play, generic capability) and problem-led innovation (solution play, specific value). The merge points: improving worthy existing applications, solving the previously unsolvable, or making the impossible routine. You get a lollapalooza effect when both paths converge. Bet on what's stable across time. Ask: what sucks, what's strategically critical, and what's complex enough to defend? That's where the [[Bottleneck Business]] opportunities live. [[AI era Defensibility|Defensibility is sequential, not static.]] Think of the castle metaphor. The bailey is your fast outer defense: distribution, velocity, brand. The motte is your slow inner defense: network effects, switching costs, workflow lock-in. Early stage, you fight in the bailey. As you scale, you retreat to the motte. Google built both. Groupon built neither. Speed gets you in the door; staying power keeps you there. Network effects are monopoly in disguise. Build both or get commoditized. The [[Technical Moat Assessment Framework]] is useful here. Selling requires models, not emotions. [[When to Sell - Lessons from Investment Legends]] covers this in depth. O'Neil watches price versus the 200-day moving average. Lynch sells when PEG exceeds 2x (you're paying for fantasy). Buffett looks for eroding moats and better opportunities elsewhere. Soros uses hard stops at 7-8% loss and locks profits at 20-25%. Graham sticks to intrinsic value divergence. Whatever your model, have one. Discipline beats intuition because psychology kills portfolios. ## Atoms, Bits, and Progress AI is predictive, not intelligent. Anthropomorphizing it is the 21st century version of thinking the sun revolves around Earth. These models compress data into statistical distributions and predict next tokens. They don't think. The danger is misunderstanding the tool and then misusing it. The opportunity: [[AI Verification]] at inhuman scale. Stop asking "can AI think?" and start asking "what patterns can AI find that humans can't?" [[Control the browser & win the customer|Browser control is the next monopoly.]] Whoever controls the default wins the habit. Netscape lost to IE not on quality but on defaults. Chrome won on speed plus defaults. The next browser war is about agency, who acts on your behalf. The winner reduces friction to zero. [[The directional arrows of inevitable progress|Progress has directional arrows]], not a random walk. Energy density increases. Computation costs fall. Human-machine interfaces tighten. Space launches cheapen. Autonomous systems proliferate. These aren't predictions. They're [[Wright's Law]], thermodynamics, and information theory compounding over decades. Stand where the arrows converge. Avoid fighting entropy. ## Atoms Reclaim Value from Bits [[The Watt-Bit Spread and the Future of AI Power Markets|The Watt-Bit Spread]] exposes a core market tension. AI companies maximize throughput, so power delivered now is worth 10x power delivered later. Utilities minimize overhead, so an electron in 2027 equals an electron in 2030 to them. That mismatch in time preferences creates massive pricing inefficiency. AI needs atoms (electricity) to create bits (computation), but the atom suppliers don't price for urgency. Whoever solves grid-scale energy storage captures the spread. Long-duration storage, superconductors, grid enhancement. Picks and shovels for the gold rush. [[The Data Center is the Computer - How the Network Shapes Performance|The data center is the computer]] because the network defines throughput, not the processors. Cloud workloads use north-south traffic and basic Ethernet. Generative AI needs east-west traffic and [[Spectrum-X]] fabric. AI factories with millions of GPUs demand [[NVLink]] and [[InfiniBand]]. Different workloads, different infrastructure. Know your [[Bottleneck Business|bottleneck]]. CPU-bound? Add compute. Network-bound? Add bandwidth. Memory-bound? Add VRAM. Optimize the wrong thing and you waste everything. [[Timing is everything]] because digital events require temporal ordering. As systems distribute across continents, clock synchronization goes from a nice-to-have to absolutely critical. Regulators now demand nanosecond precision. Modern applications depend on machines agreeing on "when" as tightly as they agree on "what." Relativity matters at scale. Light speed creates lag. Cheap atomic clocks, network time protocol, GPS disciplined oscillators. Atoms keeping bits honest. Data centers can make or break electricity affordability. PJM capacity prices hit $329 per megawatt-day, seven times higher than two years ago. Peak demand drives infrastructure cost, and the grid gets built for roughly 35 critical hours per year. If data centers could shift load during those peaks, infrastructure costs would drop significantly. Require flexibility at connection time. Build for typical case plus flexible load, not worst case. Demand response creates real value. ## Markets, Crises, and Uncertainty [[Navigating the Crisis Cycle - Tariffs, Uncertainty, and the Price of Risk|Crisis cycles rhyme because human psychology doesn't evolve.]] Trigger, reaction, aftershocks, resolution. In the 2025 Tariff Crisis, ERP spiked from 4.54% to 5.08% (fear premium rising), Treasury yields fell from 4.27% to 4.01% (flight to safety). Same three camps show up every time: "I told you so" (confirmation bias), "buy the dip" (recency bias), and "wait and see" (analysis paralysis). Understand the process, don't predict the outcome. ERP spikes are emotional signals, not rational forecasts. Monitor the variables, adjust for aftershocks, and remember that uncertainty is information, not paralysis. [[Making sense of a shifting world|Globalization is cresting.]] China drove 38% of global growth from 2010-2023, but 2008 exposed how fragile the system really was. Brexit, Trump, trade wars. All symptoms of one cause: gains got concentrated, pain got distributed, and democracy didn't adjust fast enough. The disruption playbook moved from business (Amazon, Tesla) to politics (Bukele, Milei). "Move fast and break things" works when the downside is bankruptcy. It fails in government where the downside is societal collapse. Politics is now a market driver. Fundamentals matter, but policy matters more. Hedge accordingly. [[Unravelling Stablecoins - a short journey through modern digital dollars|Stablecoins are speedrunning banking history.]] Fiat-backed (centralized, 1:1 redemption), asset-backed (decentralized, on-chain collateral), and strategy-backed (actively managed). Over $160B on-chain, fiat-backed dominating at 94%+ market share. Same evolution we saw in [[Lessons from early american banking for stablecoins - when a dollar was not always a dollar|early American banking]]: fractional reserves, bank runs, eventual central clearing. Trust requires either transparency or enforcement. Crypto provides the first through blockchain, TradFi provides the second through FDIC. Always clarify the backing, weigh the transparency, and remember that higher yields signal higher risk. ## Principles That Compound [[Operating Principles for Growth - Mark Leonard & Constellation Software|Constellation Software]] is [[Charlie Munger]]'s Berkshire model applied to software. $25M into $31B through decentralization (keep units small, human-scale, autonomous), experimentation (scientific method, MECE analysis, one-year post-mortems on every acquisition), and alignment (employees invest bonuses in company stock). Over 100 employee-millionaires by 2015. Small units move fast, experiments teach cheaply, and alignment kills principal-agent problems. [[Jamie Dimon]] runs JPMorgan on similar principles. [[Lessons from the Titans - Key Principles for Progress]] maps this thinking across industries. [[Level 5 Leadership]] combines humility and will. Out of 1,435 Fortune 500 companies, only 11 made the leap from good to great. All had Level 5 leaders, people ambitious for the mission, not their own ego. Not charismatic icons. Quietly resolved. Personal humility plus professional will creates sustainable success. The inverse holds too: ego plus ambition creates a flash followed by collapse. Mission alignment outlasts personality. [[Why Giving Away Business Models is Genius|Giving away business models]] drives winner-take-most outcomes. Commoditize the core, monetize the edge. Google gave away free search, charged for ads. Spotify gave away free music, charged for subscriptions. AWS commoditized infrastructure, monetized scale. [[Sell your sawdust]] is the same idea at a smaller scale. Network effects make free more valuable (more users, more value), data becomes currency (attention is the actual product), and bits have zero marginal cost. The value created exceeds the disruption caused. At least until regulation rebalances things. ## Where It All Connects Everything links because the underlying forces are consistent. [[High Agency]] enables [[Pattern breaking ideas - the key to outlier startups|pattern-breaking ideas]]. [[Effectual Reasoning]] aligns with [[The Deep Tech Growth Cycle is different|deep tech growth cycles]]. [[AI era Defensibility|Defensibility sequencing]] mirrors [[Operating Principles for Growth - Mark Leonard & Constellation Software|Constellation's operating principles]]. [[The Watt-Bit Spread and the Future of AI Power Markets|Energy economics]] shapes [[Control the browser & win the customer|browser wars]]. [[The directional arrows of inevitable progress|Directional arrows]] converge where [[Making sense of a shifting world|markets meet geopolitics]]. The winners over the next decade won't predict best. They'll position at the intersections. Physical meets digital as atoms reclaim value from bits. Energy meets computation as thermodynamics constrains what's possible. Agency meets effectuation as builders shape reality instead of observing it. Fast meets defensible as companies combine velocity with staying power. Find the merge points. Bet on what's stable. Work with entropy instead of fighting it. Build the machine that lets you do your life's work at scale. [[Inversion|Invert, always invert]]. What survives inversion is worth building. --- Related: [[First Principles and Mental Models MoC]] | [[Investing/Investing System MoC]] | [[Foundational Models MOC]] | [[Defensibility Principles MOC]] #firstprinciple #synthesis #deeptech #mentalmodels #kp