Here I try to keep track of original ideas I've had over the years, which I do for a number of reasons:
These ideas feel especially relevant given what's happening in AI and technology right now—agentic systems going production, the AI bubble debate, workforce disruption hitting specific industries, China's open-source dominance, and the growing need to figure out where AI should and shouldn't be used.
| Idea | Why It's Relevant | Reference |
|---|---|---|
| Design AI systems to be replaceable by smarter models—never encode clever human heuristics | Claude Opus 4.6 and GPT-5.3-Codex just dropped; today's workarounds are tomorrow's dead code | Bitter Lesson Engineering |
| ASI-like outcomes come from universal outer loops comparing current→desired state, not new models | Agentic AI is going production—the outer loop matters more than the model | Pursuing the Algorithm |
| Generalized hill-climbing at the runtime/scaffolding layer is the path to continuous AI improvement | MCP and tool-use are mainstream; scaffolding is where the real gains are | Generalized Hill-Climbing |
| The real bubble is human labor, not AI spending—companies never wanted to hire people | Meta just committed $60B to AI chips; the spending looks less like a bubble when you see what it replaces | The Bubble Is Labor |
| A major AI investment crash, AGI arriving, and mass job loss can all happen simultaneously | The AI bubble debate assumes these are mutually exclusive—they're not | Revisiting the AI Bubble |
| AI collapses tool + operator + outcome into a single unit, eliminating the human operator | Seedance 2.0 spooked Hollywood—the tool-operator-outcome collapse is hitting creative industries now | AI Gives You Outcomes |
| People are building their own bespoke software instead of buying SaaS, collapsing product categories | Agentic coding tools are making self-software the default for technical users | AI Self-Software |
| China's global strategy resembles private equity—patiently waiting for economies to decay, then acquiring | 80% of AI startups now build on Chinese open-source models; this is the strategy working | China Private Equity |
| Prompt injection is a vulnerability, not merely a design flaw—it meets the standard definition | More tools + more MCP connections = larger attack surface; prompt injection matters more, not less | Is PI a Vulnerability? |
| Security becomes an AI-vs-AI speed competition over asset and vulnerability management | Deepfake verification requirements and AI-powered attacks are making security an arms race | Cybersec Changes 2026 |
| Quality signals are inverting: excellence will be assumed AI-generated while imperfection signals authenticity | As practical AI beats hype, the question of what's "real" becomes the cultural fault line | AI Quality Paradox |
| Use the "Job vs Gym" distinction: outsource where the goal is the outcome, never where the goal is the effort | The most important framework for deciding where AI belongs in your life | Keep Robots Out of the Gym |
Ideas from 2024-2026, organized by theme.
| Idea | Date | Reference |
|---|---|---|
| Design AI systems to be replaceable by smarter models—never encode clever human heuristics | 2026-02 | Bitter Lesson Engineering |
| AI lets you do way more stuff: Previous tech is tooling that humans use; AI is tech that can do work itself | 2025-02 | Essay |
| AI will replace entire companies and products, not just automate jobs within existing companies | 2025-07 | AI Creative Destruction |
| Tiny teams using AI will outcompete traditional companies by shipping features almost instantly | 2025-07 | AI Creative Destruction |
| Two to three person teams with AI can undercut competitors while maintaining better margins | 2025-07 | AI Creative Destruction |
| Current companies are inefficient versions of what they could be with AI transformation applied | 2025-07 | AI Creative Destruction |
| One to ten percent of current developers can become full-stack builders | 2025-05 | AI Solution Factories |
| Companies will shift from high-leverage single products to identifying and solving any problems | 2025-05 | AI Solution Factories |
| Knowledge workers making under $80K will be replaced by AI systems soon | 2024-11 | UL 457 |
| Companies will use AI agents to multiply their effective headcount by 10-50x | 2024-11 | UL 457 |
| Being able to create ideas, implement them, and market them defines future workforce survival | 2024-11 | UL 457 |
| AI businesses focused on tooling will crash while companies creating with AI will boom | 2024-10 | AI Enablement |
| Small teams of 1-20 people will function like 200-20,000 | 2024-10 | AI Enablement |
| AI is pure accelerant for creation and business enablement, not a space or technology | 2024-10 | AI Enablement |
| B2B services are especially vulnerable to AI because they function largely as middleware | 2023-09 | B2B Services |
| AI collapses tool + operator + outcome into a single unit, eliminating the human operator | 2025-10 | AI Gives You Outcomes |
| AGI will come from AI systems (scaffolding + models + tools), not standalone models alone | 2025-10 | Karpathy Wrong on AGI |
| The real data problem is nobody looking at existing business data, not training data scarcity | 2025-11 | AI Data Problem |
| If a system produces outputs requiring understanding to create, it understands—mechanism is irrelevant | 2025-11 | AI Understanding Outputs |
| Quality signals are inverting: excellence will be assumed AI-generated while imperfection signals authenticity | 2025-11 | AI Quality Paradox |
| The real bubble is human labor, not AI spending—companies never wanted to hire people | 2025-12 | The Bubble Is Labor |
| People are building their own bespoke software instead of buying SaaS, collapsing product categories | 2025-12 | AI Self-Software |
| Personal AI maturity follows a 9-tier progression from basic chatbot usage to full AI companion | 2025-12 | PAIMM |
| The defining shift of 2026 is AI becoming verifiable via TDD/eval, enabling automated improvement loops | 2026-01 | AI Changes 2026 |
| ASI-like outcomes come from universal outer loops comparing current→desired state, not new models | 2026-01 | Pursuing the Algorithm |
| Generalized hill-climbing at the runtime/scaffolding layer is the path to continuous AI improvement | 2026-02 | Generalized Hill-Climbing |
| A major AI investment crash, AGI arriving, and mass job loss can all happen simultaneously | 2025-10 | Revisiting the AI Bubble |
| Idea | Date | Reference |
|---|---|---|
| AI will dramatically widen existing educational gaps—making great students extraordinary, bad students worse | 2025-06 | AI Effect on Education |
| Developer excitement about AI power is directly proportional to business owners' excitement about replacing workers | 2025-07 | AI Morose Mania |
| First week of July 2025 might be remembered as the start of the proto-AGI era | 2025-07 | AI Morose Mania |
| Platforms should show all legal speech but let users filter their own experience completely | 2025-01 | Reality Through Filters |
| Server-side reality with client-side filtering will solve platform censorship debates | 2025-01 | Reality Through Filters |
| Be careful not to outsource work that is only useful when done yourself personally | 2025-01 | Fast vs Slow AI |
| Meaning and suffering are closely tied together—like muscle teardown during weightlifting | 2025-01 | Fast vs Slow AI |
| U-shaped future: high chances for Western blossoming OR truly Orwellian AI dystopia | 2024-11 | Frontview Mirror |
| The game is getting really good at adjusting plans, not having the perfect plan | 2024-11 | Frontview Mirror |
| Life satisfaction directly correlates with ability to complete: "I solve [problem] by doing [action]" | 2024-10 | Most Important Sentence |
| Monday feelings serve as the best metric for measuring overall life and career satisfaction | 2024-10 | Most Important Sentence |
| Understanding emerges from ability to perform tasks, not from subjective experience | 2024-10 | Deutsch and AI |
| Consciousness is evolutionary illusion giving perception of authorship/agency for social advantage | 2024-10 | Deutsch and AI |
| Human creativity is mechanistic process not requiring free will—making AI creativity theoretically possible | 2024-10 | Deutsch and AI |
| People use ideology and emotion as lens to interpret reality rather than actual facts | 2024-11 | N/A |
| The smartest people about X are the most interested in it. IQ helps, but interest has to come first | 2024 | N/A |
| Humans "overfit" during their lives like AI models do—need deliberate entropy injection to counteract | 2025-10 | Humans Need Entropy |
| Maximum life enjoyment comes from consciously oscillating between exploration (novelty) and exploitation (pattern) | 2025-10 | Explore vs Exploit |
| Attention and novelty literally slow subjective time—pursue them to extend your experienced lifespan | 2025-10 | Magnifying Time |
| UBI alone won't manage AI displacement—governments will also need immersive games and entertainment | 2025-10 | UBI + Games |
| Humans may not have "real" world models either—ours may be as emergent and flimsy as AI's | 2025-09 | Do Humans Have World Models? |
| Use the "Job vs Gym" distinction: outsource where the goal is the outcome, never where the goal is the effort | 2025-11 | Keep Robots Out of the Gym |
| China's global strategy resembles private equity—patiently waiting for economies to decay, then acquiring | 2025-12 | China Private Equity |
| Anti-fragility requires understanding the world, opinions on how it should differ, and skills to change it | 2026-02 | Antifragile Framework |
| What people call Claude Code addiction is actually addiction to creation—categorically better than consumption | 2026-01 | CC Addiction = Creativity |
| Idea | Date | Reference |
|---|---|---|
| The future of hacking is dueling banjos of attacker's AI vs. defender's AI | 2025 | Future of Hacking |
| Security should disappear into engineering SOPs rather than exist as separate bolt-on products | 2025-07 | AI Creative Destruction |
| AI agents will automatically build, test, and analyze code—making security part of engineering | 2025-07 | AI Creative Destruction |
| Security baseline for technology is set by market of people's behavior, not individual efforts | 2025-05 | AI Security |
| We will get precisely the right amount of AI security—no more and no less | 2025-05 | AI Security |
| Security awareness training fails because attacks trigger System 1 thinking, not System 2 reasoning | 2024-10 | UL 456 |
| Only defense against AI spearphishing is mindfulness techniques that teach pausing when stressed | 2024-10 | UL 456 |
| Prompt injection requires comparing requestor's implied expectation against what actually happens | 2024-10 | UL 456 |
| AI avatar breaches will expose users' deepest secrets given to AI companions | 2024-10 | N/A |
| Most AI hacking will be normal hacking of regular infrastructure used by AI companies | 2024-10 | N/A |
| The result of agents is your adversary having 2x, 5x, 50x larger teams in 1-3 years | 2024 | N/A |
| Prompt injection is a vulnerability, not merely a design flaw—it meets the standard definition | 2025-11 | Is PI a Vulnerability? |
| Prompt injection strings are like Metasploit modules—should be shared openly, not hoarded as secrets | 2025-11 | PI OPSEC |
| Security products fail by communicating activities to practitioners instead of communicating value to the business | 2025-12 | New Cybersec Product |
| Security becomes an AI-vs-AI speed competition over asset and vulnerability management | 2025-12 | Cybersec Changes 2026 |
| Anthropic's advantage is an integrated ecosystem like Apple's iPhone era, not the model itself | 2025-12 | Anthropic's Vision |
| # | Idea | Date | Reference |
|---|---|---|---|
| ⭐ | Universal Daemonization & Digital Assistants: Everything gets an API; AI assistants query those APIs on our behalf; we stop interacting with the world directly | 2014 | The Real Internet of Things, HouSecCon Talk (2014), UD is the Future of IoT, UD + Augmented Reality, Launching Daemon |
| ⭐ | SPQA Architecture: Most software will shift from circuit-based to understanding-based systems with four components: State, Policy, Questions, Actions | 2023 | SPQA Essay |
| ⭐ | Human 3.0: The next evolution of humanity is AI-augmented humans who can create, analyze, and build at unprecedented scale | 2020s | Human 3.0 |
| ⭐ | Layer-Dependent Reality: Reality depends on who's asking and for what purpose; free will is meaningless at the physics layer but essential at the human layer | Ongoing | N/A |
| ⭐ | The Meaning Crisis: We got meaning from religion, family, and hard work—we're losing all three, and AI accelerates the loss of work | 2024 | N/A |
| ⭐ | Personal AI Infrastructure (PAI): A unified system for human augmentation built on nested loops of current→desired state and the scientific method | 2025 | PAI |
| 1 | AI lets you do way more stuff: Previous tech is tooling that humans use; AI is tech that can do work itself - that's worth trillions | 2025-02-21 | AI Lets You Do Way More Stuff |
| 2 | Intelligence Tasks require human judgment pattern recognition and decision making beyond simple automation programming | 2025-07-21 | AI Workforce Volume Difficulty |
| 3 | Visualizing work capacity as area under curve combining both volume and difficulty dimensions together | 2025-07-21 | AI Workforce Volume Difficulty |
| 4 | Most Intelligence Tasks are not being done at all due to insufficient human capacity available | 2025-07-21 | AI Workforce Volume Difficulty |
| 5 | AI expands capacity on both axes handling more volume and higher difficulty tasks simultaneously | 2025-07-21 | AI Workforce Volume Difficulty |
| 6 | KISAC framework measures Intelligence Task performance through Knowledge Intelligence Speed Accuracy Cost metrics | 2025-07-21 | AI Workforce Volume Difficulty |
| 7 | AI role is not replacement but expansion illuminating dark corners of never completed work | 2025-07-21 | AI Workforce Volume Difficulty |
| 8 | Current companies are inefficient versions of what they could be with AI transformation applied | 2025-07-08 | AI Creative Destruction |
| 9 | AI will replace entire companies and products not just automate jobs within existing companies | 2025-07-08 | AI Creative Destruction |
| 10 | Tiny teams using AI will outcompete traditional companies by shipping features almost instantly | 2025-07-08 | AI Creative Destruction |
| 11 | Two to three person teams with AI can undercut competitors while maintaining better margins | 2025-07-08 | AI Creative Destruction |
| 12 | Security should disappear into engineering SOPs rather than exist as separate bolt on products | 2025-07-08 | AI Creative Destruction |
| 13 | AI agents will automatically build test and analyze code making security part of engineering | 2025-07-08 | AI Creative Destruction |
| 14 | Death by thousand startups as tiny teams launch products making entire software categories obsolete | 2025-07-08 | AI Creative Destruction |
| 15 | AI changes fundamental economics of software creation enabling one person to replace fifty people | 2025-07-08 | AI Creative Destruction |
| 16 | First week of July 2025 might be remembered as the start of proto AGI era | 2025-07-04 | AI Morose Mania |
| 17 | Developer excitement about AI power directly proportional to business owners excitement about replacing workers | 2025-07-04 | AI Morose Mania |
| 18 | AI productivity gains won't offset economic damage from losing hundreds of thousands of jobs | 2025-07-04 | AI Morose Mania |
| 19 | AI will dramatically widen existing educational gaps making great students extraordinary bad students worse | 2025-06-01 | AI Effect on Education |
| 20 | AI like reading is powerful self improvement tool mostly used by self improvement oriented people | 2025-06-01 | AI Effect on Education |
| 21 | Smart cybersecurity professionals split into two camps with radically different views on AI technology | 2025-05-31 | AI Friend Groups |
| 22 | AI skeptics damage careers by convincing impressionable people to dismiss transformative technology prematurely | 2025-05-31 | AI Friend Groups |
| 23 | People's fundamental outlook on change and life determines their stance on AI advancement | 2025-05-31 | AI Friend Groups |
| 24 | AI embracers transform anxiety about rapid change into excitement and productive energy instead | 2025-05-31 | AI Friend Groups |
| 25 | Evaluate AI opinions by examining speaker's underlying worldview about change and technological progress | 2025-05-31 | AI Friend Groups |
| 26 | Security baseline for technology is set by market of people's behavior not individual efforts | 2025-05-29 | AI Security |
| 27 | AI security baseline moves like mass of dozen suns based on user emotion inducing events | 2025-05-29 | AI Security |
| 28 | We will get precisely the right amount of AI security no more and no less | 2025-05-29 | AI Security |
| 29 | Global security baseline for technology is largely unmovable by individuals because anchored to human behavior | 2025-05-29 | AI Security |
| 30 | Security increases only when users demand it due to emotion inducing risk events not before | 2025-05-29 | AI Security |
| 31 | Large companies will transform into conglomerates of hundreds of thousands of micro-companies internally | 2025-05-06 | AI Solution Factories |
| 32 | Only one to ten percent of current developers can become full-stack builders needed | 2025-05-06 | AI Solution Factories |
| 33 | Companies will shift from high-leverage single products to identifying and solving any problems | 2025-05-06 | AI Solution Factories |
| 34 | Universal business engine: identify problems, analyze solutions, build better ones, distribute through AI | 2025-05-06 | AI Solution Factories |
| 35 | Future builders need problem understanding, sentiment analysis, engineering, and go-to-market skills combined | 2025-05-06 | AI Solution Factories |
| 36 | Business becomes directory of products where question is which shop builds superior solution | 2025-05-06 | AI Solution Factories |
| 37 | The total TAM for AI is: 1) total human workforce cost, and 2) what companies will pay to start/10x/100x their business | 2025-01 | N/A |
| 38 | AI's total addressable market combines human workforce costs and business multiplication investment opportunities globally | 2025-01-28 | AI TAM |
| 39 | Companies will pay to ten times or thousand times multiply their business using AI | 2025-01-28 | AI TAM |
| 40 | The AI market opportunity represents hundreds of trillions of dollars in potential value creation | 2025-01-28 | AI TAM |
| 41 | Be careful not to outsource work that is only useful when done yourself personally | 2025-01-22 | Fast vs Slow AI |
| 42 | Meaning and suffering are closely tied together like muscle teardown during weightlifting exercise sessions | 2025-01-22 | Fast vs Slow AI |
| 43 | Use AI for discovery of content but switch to natural mode for learning consumption | 2025-01-22 | Fast vs Slow AI |
| 44 | Learning and experiencing are more similar to muscle teardown than most people have realized | 2025-01-22 | Fast vs Slow AI |
| 45 | We have added more atmospheric CO2 in last thirty years than previous thirty years | 2025-01-19 | CO2 Levels |
| 46 | Earth breathes through seasonal CO2 fluctuations but each year peaks and valleys rise higher | 2025-01-19 | CO2 Levels |
| 47 | Platforms should show all legal speech but let users filter their own experience completely | 2025-01-09 | Reality Through Filters |
| 48 | AI powered digital assistants will become personal content filters understanding users better than platforms | 2025-01-09 | Reality Through Filters |
| 49 | Server side reality with client side filtering will solve platform censorship debates within years | 2025-01-09 | Reality Through Filters |
| 50 | Platform confusion stems from not declaring whether they are public squares or private communities | 2025-01-09 | Reality Through Filters |
| 51 | Modern tools enable scripts with embedded dependencies eliminating traditional setup friction between writing and running | 2025-01-08 | Executable TypeScript |
| 52 | Bun compiles TypeScript into standalone binaries including runtime dependencies and code in one executable | 2025-01-08 | Executable TypeScript |
| 53 | Self-contained scripts represent fundamental shift in code distribution eliminating install these dependencies first paradigm | 2025-01-08 | Executable TypeScript |
| 54 | The ideal number of human employees in any company is zero not one person | 2025-01-06 | AI Founder Augmentation |
| 55 | Jobs exist because founders want to make things but cannot do it alone themselves | 2025-01-06 | AI Founder Augmentation |
| 56 | AI isn't replacing human workers it's replacing the need for human workers entirely everywhere | 2025-01-06 | AI Founder Augmentation |
| 57 | AI will make founders into superhumans capable of doing work of ten thousand employees | 2025-01-06 | AI Founder Augmentation |
| 58 | Future survival requires being full spectrum individual eclectic polymathic and centered on human connection | 2025-01-06 | AI Founder Augmentation |
| 59 | Left center's inability to address immigration gives all power to far right extremists | 2024-11-17 | Frontview Mirror |
| 60 | AI Crash Takeoff where bubble companies die while solid AI fundamentals companies hockey stick | 2024-11-17 | Frontview Mirror |
| 61 | Agentic systems not standalone models will achieve AGI by replicating competent human knowledge workers | 2024-11-17 | Frontview Mirror |
| 62 | AI Employee purchases will actually be complex networks of agents operating as one system | 2024-11-17 | Frontview Mirror |
| 63 | Russia plays middle position between US China benefiting from both while influencing mutual hatred | 2024-11-17 | Frontview Mirror |
| 64 | Framing and positive momentum belief more powerful than traditional economic fundamentals for growth success | 2024-11-17 | Frontview Mirror |
| 65 | Countries do business with China because they must not because they want to anymore | 2024-11-17 | Frontview Mirror |
| 66 | The game is getting really good at adjusting plans not having the perfect plan | 2024-11-17 | Frontview Mirror |
| 67 | U shaped future with high chances for Western blossoming or truly Orwellian AI dystopia | 2024-11-17 | Frontview Mirror |
| 68 | AI agents' maturation in 2025 will be the biggest advancement in artificial intelligence technology | 2024-11-14 | UL 459 |
| 69 | AI benefits will come from tooling and ecosystems around models not the models themselves | 2024-11-14 | UL 459 |
| 70 | RebootAI offline AI oracle for emergencies providing survival knowledge without internet connectivity available | 2024-11-14 | UL 459 |
| 71 | Treating bullying as collective issue rather than individual one significantly reduces occurrence in schools | 2024-11-14 | UL 459 |
| 72 | Next AI rigs should be Mac Mini clusters or standard big beefy NVIDIA-based boxes | 2024-11-14 | UL 459 |
| 73 | AI system that uses smartest available AI to judge other AI systems' task performance | 2024-11-11 | AI Rating AI |
| 74 | High-quality prompting architecture can replace complex AI evaluation frameworks for assessing AI sophistication | 2024-11-11 | AI Rating AI |
| 75 | Rating AI outputs across sixteen thousand dimensions using seed analysis types like Attention | 2024-11-11 | AI Rating AI |
| 76 | Framework where newest best model becomes judge and new models plug in for testing | 2024-11-11 | AI Rating AI |
| 77 | AI models predictably score on human scale from high school to PhD across tasks | 2024-11-11 | AI Rating AI |
| 78 | Using AI to rate quality of other AI outputs with structured assessment scale | 2024-11-08 | UL 458 |
| 79 | Companies will use AI agents to multiply their effective headcount by ten times | 2024-11-08 | UL 458 |
| 80 | Crypto is back as gambling and money stores not as transformative technology idea | 2024-11-08 | UL 458 |
| 81 | Silicon Valley tech leaders will create liberal policies lifting everyone not just elite | 2024-11-08 | UL 458 |
| 82 | Waking up tomorrow is worth more than million dollars so treat it accordingly | 2024-11-08 | UL 458 |
| 83 | Knowledge workers making under eighty thousand dollars will be replaced by AI systems soon | 2024-11-03 | UL 457 |
| 84 | Companies will use AI agents to multiply their effective headcount by ten to fifty times | 2024-11-03 | UL 457 |
| 85 | Being able to create ideas implement them and market them defines future workforce survival | 2024-11-03 | UL 457 |
| 86 | The ease of instant AI use matters more than the intelligence of the models themselves | 2024-11-03 | UL 457 |
| 87 | Entrepreneurs report lower stress levels and more meaningful careers compared to traditional employees overall | 2024-11-03 | UL 457 |
| 88 | People use ideology and emotion as lens to interpret reality rather than actual facts | 2024-11-03 | N/A |
| 89 | Battle is largely about narratives and beliefs not policy or outcomes or actual results | 2024-11-03 | N/A |
| 90 | Three inception levels of being wrong about Trump exist simultaneously in public discourse today | 2024-11-03 | N/A |
| 91 | Success of country might depend on perceptions and resulting actions not actual policy outcomes | 2024-11-03 | N/A |
| 92 | Security awareness training fails because attacks trigger System 1 thinking not System 2 reasoning | 2024-10-26 | UL 456 |
| 93 | Only defense against AI spearphishing is mindfulness techniques that teach pausing when stressed | 2024-10-26 | UL 456 |
| 94 | World model development combined with post-training will unlock both AGI and ASI capabilities | 2024-10-26 | UL 456 |
| 95 | If AI understands world mechanics and scientific method that might be all it needs | 2024-10-26 | UL 456 |
| 96 | Google's browser-based Jarvis implementation is short-sighted compared to Apple's deep OS integration | 2024-10-26 | UL 456 |
| 97 | Prompt injection requires comparing requestor's implied expectation against what actually happens not just | 2024-10-26 | UL 456 |
| 98 | Life satisfaction directly correlates with ability to complete specific problem solution sentence template | 2024-10-24 | Most Important Sentence |
| 99 | Monday feelings serve as best metric for measuring overall life and career satisfaction levels | 2024-10-24 | Most Important Sentence |
| 100 | Finding meaningful problems and solution paths creates molten core for thriving career and life | 2024-10-24 | Most Important Sentence |
| 101 | AI businesses focused on tooling will crash while companies creating with AI will boom | 2024-10-23 | AI Enablement |
| 102 | AI is pure accelerant for creation and business enablement not a space or technology | 2024-10-23 | AI Enablement |
| 103 | Small teams of one to twenty people will function like two hundred to twenty thousand | 2024-10-23 | AI Enablement |
| 104 | Human creativity points will multiply from nineteen to twenty million through AI magnification globally | 2024-10-23 | AI Enablement |
| 105 | AI enablement tools will plateau becoming invisible hard to differentiate and monetize killing many companies | 2024-10-23 | AI Enablement |
| 106 | AI is not a space or tech but pure intelligence and business accelerant | 2024-10-18 | UL 455 |
| 107 | AI crash in tools will coincide with boom in AI-enabled creation companies | 2024-10-18 | UL 455 |
| 108 | Small teams using AI will function like companies with ten to thousand times more people | 2024-10-18 | UL 455 |
| 109 | AI is a magnifier and multiplier of human creativity not just a technology | 2024-10-18 | UL 455 |
| 110 | Human creativity points will multiply from nineteen to twenty million through AI enablement | 2024-10-18 | UL 455 |
| 111 | Eight billion humans will have their creativity unlocked by AI creating unprecedented innovation | 2024-10-18 | UL 455 |
| 112 | AI avatar breaches will expose users' deepest secrets given to AI companions | 2024-10-13 | N/A |
| 113 | Gullibility is the vulnerability while conspiracy theories are merely the threat vector | 2024-10-13 | N/A |
| 114 | AI gets exponentially better the more honest and forthcoming you are with it | 2024-10-13 | N/A |
| 115 | Most AI hacking will be normal hacking of regular infrastructure used by companies | 2024-10-13 | N/A |
| 116 | People are allergic to subtlety and want crisp clear answers of good or evil | 2024-10-13 | N/A |
| 117 | Genius creators unlikely to hit timelines perfectly and act normally comes with territory | 2024-10-13 | N/A |
| 118 | Innovation critics suffer from not reading enough biographies of great people like Elon | 2024-10-13 | N/A |
| 119 | US has gullibility problem not misinformation problem we're eating crap not being fed | 2024-10-13 | N/A |
| 120 | Need to reduce vulnerability not remove threats since threats will always exist there | 2024-10-13 | N/A |
| 121 | Digital assistants with premium API access will have massive advantage over regular consumer AI | 2024-10-06 | UL 453 |
| 122 | Thinking about yourself is the source of all unhappiness according to Naval's philosophical framing | 2024-10-06 | UL 453 |
| 123 | Too much therapy becomes rumination rather than solving problems and moving forward with life | 2024-10-06 | UL 453 |
| 124 | People will need certifications to access less nerfed AI models with fewer content filters | 2024-10-06 | UL 453 |
| 125 | Premium AI subscriptions at thirty eight thousand monthly will provide vastly superior real world insights | 2024-10-06 | UL 453 |
| 126 | AI containing security rather than security being separate from AI technology and development systems | 2024-10-04 | How Projects Fit Together |
| 127 | Humans understand desired shape of outputs better than AIs despite AIs superior mental capabilities | 2024-10-04 | How Projects Fit Together |
| 128 | Human three point zero represents AI augmented humans as next stage of human evolution | 2024-10-04 | How Projects Fit Together |
| 129 | Substrate as alternative operating system for humanity designed for upgraded human versions H2 H3 | 2024-10-04 | How Projects Fit Together |
| 130 | Humans trained for economic utility rather than becoming full spectrum people causes societal problems | 2024-10-04 | How Projects Fit Together |
| 131 | Complete system combining philosophy life mission AI augmentation and behavior change for human flourishing | 2024-10-04 | How Projects Fit Together |
| 132 | Understanding emerges from ability to perform tasks not from subjective experience according to constructor theory | 2024-10-02 | Deutsch and AI |
| 133 | AI understanding can be tested through analogical reasoning counterfactual scenarios and conceptual combination experiments | 2024-10-02 | Deutsch and AI |
| 134 | Understanding has two tiers functional understanding for applying knowledge and creative understanding for innovation | 2024-10-02 | Deutsch and AI |
| 135 | Consciousness is evolutionary illusion giving perception of authorship agency for social blame praise advantage | 2024-10-02 | Deutsch and AI |
| 136 | Villages with blame praise concepts outcompete those without by reinforcing social norms through authorship | 2024-10-02 | Deutsch and AI |
| 137 | Human creativity is mechanistic process not requiring free will making AI creativity theoretically possible | 2024-10-02 | Deutsch and AI |
| 138 | Many Nobel Prize discoveries are incremental advances that current AI could potentially have discovered | 2024-10-02 | Deutsch and AI |
| 139 | B2B services are especially vulnerable to AI because they function largely as middleware between parties | 2023-09-21 | B2B Services |
| 140 | Current AI could eliminate fifty to seventy percent revenue and sixty to eighty percent jobs | 2023-09-21 | B2B Services |
| 141 | AI impact on B2B services could represent twenty two percent reduction of total US GDP | 2023-09-21 | B2B Services |
| 142 | AGI with IQ between 120 and 160 would cause seventy percent revenue loss eighty percent jobs | 2023-09-21 | B2B Services |
| 143 | B2C services less vulnerable than B2B because product goes directly from creator to buyer customer | 2023-09-21 | B2B Services |
| 144 | More advanced AI cuts deeper into high skill knowledge work and creative work than basic | 2023-09-21 | B2B Services |
| 145 | Digital Assistants: Everyone has a DA; they know our full context; they advocate for us 24/7 | 2016 | Real Internet of Things |
| 146 | Universal Daemonization: Everything gets an API; the APIs will be used by our DAs | 2014 | Real Internet of Things, HouSecCon Talk (2014), UD is the Future of IoT, UD + AR, Launching Daemon |
| 147 | DA's Mediate Everything for Us: Most services are used by DAs, not humans | 2016 | Real Internet of Things |
| 148 | Everyone, including people, will have an API | 2016 | N/A |
| 149 | The problem with governments, economies, and societies is best thought of as problems with the population that make them up | 2017 | N/A |
| 150 | Personal daemons will broadcast preferences | 2014 | HouSecCon Talk |
| 151 | Venues personalize based on customer preferences | 2014 | N/A |
| 152 | People will lack meaning and look for it in games | 2006 | N/A |
| 153 | Humans may not actually have real world models either ours may be as emergent as AIs | 2025-09-29 | Do Humans Have World Models? |
| 154 | AI collapses tool operator and outcome into single unit eliminating the human operator entirely | 2025-10-06 | AI Gives You Outcomes |
| 155 | A major AI investment crash AGI arriving and mass job loss can happen simultaneously | 2025-10-07 | Revisiting the AI Bubble |
| 156 | Maximum life enjoyment from consciously oscillating between exploration novelty and exploitation pattern modes | 2025-10-12 | Explore vs Exploit |
| 157 | UBI alone will not manage AI displacement governments will also need immersive games entertainment | 2025-10-13 | UBI + Games |
| 158 | Attention and novelty literally slow subjective time pursue them to extend experienced lifespan | 2025-10-14 | Magnifying Time |
| 159 | AGI will come from AI systems scaffolding models tools not standalone models alone | 2025-10-19 | Karpathy Wrong on AGI |
| 160 | Humans overfit during their lives like AI models need deliberate entropy injection to counteract | 2025-10-28 | Humans Need Entropy |
| 161 | The real data problem is nobody looking at existing business data not training data scarcity | 2025-11-12 | AI Data Problem |
| 162 | If a system produces outputs requiring understanding to create it understands mechanism is irrelevant | 2025-11-21 | AI Understanding Outputs |
| 163 | Use Job vs Gym distinction outsource where goal is outcome never where goal is effort | 2025-11-24 | Keep Robots Out of the Gym |
| 164 | Quality signals inverting excellence assumed AI generated while imperfection signals human authenticity | 2025-11-24 | AI Quality Paradox |
| 165 | Prompt injection is a vulnerability not merely a design flaw meets the standard definition | 2025-11-25 | Is PI a Vulnerability? |
| 166 | Prompt injection strings are like Metasploit modules should be shared openly not hoarded | 2025-11-24 | PI OPSEC |
| 167 | The real bubble is human labor not AI spending companies never wanted to hire people | 2025-12-08 | The Bubble Is Labor |
| 168 | Anthropic advantage is integrated ecosystem like Apples iPhone era not the model itself | 2025-12-11 | Anthropic's Vision |
| 169 | Security products fail communicating activities to practitioners instead of value to the business | 2025-12-12 | New Cybersec Product |
| 170 | Personal AI maturity follows nine tier progression from basic chatbot to full AI companion | 2025-12-14 | PAIMM |
| 171 | China global strategy resembles private equity patiently waiting for economies to decay then acquiring | 2025-12-15 | China Private Equity |
| 172 | People building their own bespoke software instead of buying SaaS collapsing product categories | 2025-12-22 | AI Self-Software |
| 173 | Security becomes an AI vs AI speed competition over asset and vulnerability management | 2025-12-31 | Cybersec Changes 2026 |
| 174 | The defining shift of 2026 is AI becoming verifiable via TDD eval enabling automated improvement | 2026-01-05 | AI Changes 2026 |
| 175 | What people call Claude Code addiction is actually addiction to creation better than consumption | 2026-01-08 | CC Addiction = Creativity |
| 176 | ASI like outcomes come from universal outer loops comparing current to desired state not models | 2026-01-17 | Pursuing the Algorithm |
| 177 | Anti fragility requires understanding the world opinions on how it should differ skills to change | 2026-02-19 | Antifragile Framework |
| 178 | Generalized hill climbing at runtime scaffolding layer is the path to continuous AI improvement | 2026-02-14 | Generalized Hill-Climbing |
| 179 | Design AI systems to be replaceable by smarter models never encode clever human heuristics | 2026-02-22 | Bitter Lesson Engineering |
This section has more context on some of the bigger concepts.
If you think about most products and services, they're largely just executing on these components in different ways.
I've come to believe that complex human issues like free will and existentialism come down to understanding that reality has multiple layers. I feel like The Fabric of Reality, by David Deusch, is a good starting point for this idea, although I've been grappling with it for years before reading it in 2024.
The idea is practical rather than technical. It's the notion that each layer of reality is valid on its own, and there isn't an actual "true" reality. Or to put in another way, the way you frame your question and your desired outcomes determines which layers of reality—and thus which truths—get engaged for the answer.
E.g.:
Technically, your neighbor is a mechanistic creature that lacks free will, so he didn't have a choice about whether or not to take your figs. But right now your kid is asking why he can't have a fig, and why your neighbor is selling them across the street. So the physics conversation is rather moot.
Same with enjoying ice cream, or friendship, or trying to do good in the world. At some layer they're likely all meaningless, but that doesn't matter for humans because we spend our entire lives in layers where they do.
Humans vibrate at a particular set of layers based on our physical makeup combined with the rules of the universe, and the reactions between those ultimately determine what is real to us. And it's the same for anything else. Meaning, reality, and truth are all dependent on who's asking the questions, and for what reasons.
These are two terms I came up with in 2008 that we would eventually use for talking about people who were augmented with tech—or not. My goal was to have them be easy to say and single syllable.
First, augmented.
And then non-augmented.
Again, a major reason for doing this is to encourage others—namely you—to build your own version of this page. Please make one and send it to me.