Wiki

Agent Proficiency — The skill of effectively managing and directing AI agents

AI, Society & Epistemic Collapse — Four complementary frames on AI's societal impact beyond the labor/safety axes: Kingsbury on epistemic collapse (information ecology pollution, end of evidence, hyperscale propaganda), Diamandis on speciation (five forks driven by exponential tech), Klein/Sun on the political-material backlash (data center resistance as the site where AI's concentrated costs collide with democratic voice), and Acemoglu on the investment-inequality trap (AI's required revenue scale forces either a massive capital-share surge in national income or an economically costly crash

Business Moats in AI — As AI commoditizes everything "hard to do," the only durable moats are things "hard to get"

Context Engineering — The discipline of filling a model's context window with exactly the right information for the next step. In Karpathy's Software 3.0 framing, the context window is the new programming surface; the same weights produce a good or bad agent depending entirely on what's in the window. Everything reduces to four operations

LLM Knowledge Bases — Using LLMs to compile and maintain personal knowledge bases from raw sources

Model Quantization — Quantization reduces LLM memory footprint by storing weights as small integers instead of full-precision floats

Physical AI Systems — Frontier AI systems extending beyond language into the physical world

Product Management Dynamics — Four lenses on product management. Berne's Transactional Analysis names ten recurring "games" in PM meetings and prescribes an antithesis for each: refuse the assigned role and respond from Adult ego state. Doshi's Antithesis Principle adds a meta-cognitive layer: every observation about human nature yields an obvious outward tactic and a non-obvious inward warning. Elman reframes the PM's core artifact as story over spec and argues AI's collapse of build cost inverts the development loop, making judgment and curation the whole job. Winters validates this empirically: PM teams are dramatically more productive but not more impactful, because AI removed the engineering bottleneck without replacing the prioritization mechanism

Psychology & Trauma — Behaviors that look like personality

Psychology of Death — Carl Jung's psychological framework for death

Synthetic Personas — LLMs prompted to take on personas produce average, not diverse, responses

The Future of Knowledge Work — AI is commoditizing knowledge

World Models — An alternative AI paradigm to LLMs

Agent Harness — The harness is everything that wraps a model to make it an agent

Agentic Engineering — The emerging discipline of designing, optimizing, and orchestrating AI agents

AI Evals — Effective AI evaluation starts with manually reviewing production traces

AI-Native Product Development — The shift from traditional PRDs and sprint planning to rapid prototyping with AI agents

Claude Code Skill Frameworks — Claude Code is a full agent platform, not just a coding tool

Graph Engineering — Graph engineering organizes AI work as nodes (bounded tasks) and edges (real data dependencies), enabling parallel execution wherever tasks are independent. The core skill is the "fake-edge test"

Loop Engineering — Loop engineering is the practice of replacing yourself as the person who prompts the agent

Managed Agents — Hosted agent infrastructure and products spanning enterprise and consumer

Voice AI Infrastructure — Voice AI spans two layers: transport infrastructure that makes latency invisible, and model architecture that makes conversation natural. OpenAI's relay-plus-transceiver split solves WebRTC-in-Kubernetes at scale, while GPT-Live's full-duplex architecture and delegation pattern decouple continuous interaction from deep reasoning. In production, the cascaded pipeline (STT → LLM → TTS) remains dominant because it lets teams enforce guardrails, swap components per-market, and shave latency at each stage independently

AI Safety & Interpretability — The effort to understand and align AI models

Claude Mythos Preview — Anthropic's most capable model as of April 2026

Frontier Models — The frontier model landscape (2026) is defined by rapid competitive rotation among Anthropic, OpenAI, and Google, inference-time multi-agent scaling as an alternative to ever-larger single models, a narrowing gap between frontier and open-weight local models, and a paradigm shift where the binding constraint on progress is high-quality training data

AGI Definitions — "AGI" is contested and often misused. On the most prominent definitions

AI Careers — The AI job market is bifurcating into "big AI" and "small AI," but the highest-leverage roles are emerging at the intersection of building and strategy

AI Organization Design — Block (Jack Dorsey) proposes replacing corporate hierarchy with an "intelligence layer"

AI Regulation — The US is developing a fragmented patchwork of 40+ state AI laws while the Trump administration pushes for federal preemption

AI Startup Distribution — Code is commoditized

AI User Perspectives — Anthropic's 81K-person global survey (Dec 2025) found users primarily want AI for professional excellence and life management

Consumer AI Apps — Consumer AI usage is broad (nearly half of U.S. consumers) but shallow

Geopolitics & World Order — The post-1945 rules-based international order led by the US is fragmenting into a might-is-right multipolar system. Multiple active conflicts

Services-as-Software — The next trillion-dollar company will sell the work, not the tool

Solo Business — A solo business (or "one-person empire") stacks three of MJ DeMarco's five "money trees"

Tech Crash Cycles — Every major technology surge follows the same Carlota Perez pattern

Venture Capital Access — Private markets are democratizing. USVC (AngelList) is the first indexed, low-fee vehicle giving retail investors access to top-tier venture capital

Vertical AI — Domain-specific AI applications are finding product-market fit faster than horizontal AI, in both enterprise (legal, healthcare, customer service) and consumer (personal assistants). Harvey ($0→$200M ARR in 36 months) and Sierra ($0→$165M ARR in 26 months) prove the enterprise playbook

Y Combinator AI Thesis (2026) — YC's 2026 bet areas span AI-native agencies, "Cursor for PM," AI hedge funds, government AI tools, spatial reasoning models, and reindustrialized manufacturing

Andrej Karpathy — Former Tesla AI director, OpenAI founding member, and deep learning educator (cs231n, "Neural Networks: Zero to Hero"). Recent focus has shifted from model internals to knowledge management and agent workflows

Dario Amodei — CEO of Anthropic and one of the most prominent voices on both AI's transformative potential and its existential risks. Co-founder who left OpenAI in 2020 with a cohort of safety-focused researchers. Known for four major essays

DHH (David Heinemeier Hansson) — Creator of Ruby on Rails, co-founder of 37signals (Basecamp, HEY), and one of tech's most opinionated voices on software craft for two decades. Did a public 180-turn on AI coding tools over the 2025 winter break

Lenny Rachitsky — Creator-operator who built a top-tier product newsletter into an empire (1.2M newsletter subscribers, 500K YouTube subscribers, ~25 pieces/month)

Peter Steinberger — Vienna/London-based developer, ex-PSPDFKit founder (PDF SDK, exited 2021), now building AI-native developer tools at extreme velocity. Creator of OpenClaw and 50+ open-source repos. Self-describes as "polyagentmorous builder"

Writing Craft — Specificity is the single most powerful lever in writing