▒▒ ▒▒ ▒▒ ▒▒▒ ▒▒▒ ▒▒▒ ▒▒ ▒▒ ▒▒ ▒ ▒▒▒ ▒ ▒▒ ▒▒ ▒▒ ▒ ▒▒▒ ▒▒▒ ▒▒ ▒▒ ▒▒ ▒ ▒ ▒ ▒ ▒▒▒▒ ▒▒▒▒ ▒▒▒ ▒ ▒ ▒▒▒
- ▸Ran a full user-research study on how founders want to work with AI agents — participants choose a survey or a conversational interview, and the choice itself is a research signal. Participant identity is stored in a separate database that fails closed, so who someone is never sits next to what they said.
- ▸Used the site as a proving ground for machine-payable commerce and MCP: an end-to-end Machine Payments Protocol talk with a live agent buying real goods, an embedded branded identity-collection widget delivered inside an MCP tool result, and a crowdsourced long-read service that prints and mails what wins — paid agent-to-agent.
- ▸Shipped a long tail of standalone builds — a phone-first Go tutorial, a self-improving ASCII-art gallery, a 450-character crossword generator, a 3D industrial-design lab, an Android room-presence tracker, and a co-writing IDE that checks whether an AI draft still sounds like the person who wrote it.
- ▸Wrote the thesis before the code: a founding memo, an upper-bound analysis, a ten-rung build ladder, competitive assessments, and a 206-entry lexicon of the terms of art in American lawmaking — each entry noting what a claim gets wrong if the term is misread.
- ▸Turned the plan into a running pipeline: a daily Federal Register issue that publishes itself as a site, an immutable hash-chained archive, and an MCP server that serves dated snapshots, so any claim can be checked against the exact source that existed when it was made.
- ▸Made correctness the actual product — claims over a newly enacted housing act were audited line by line against enrolled text, and the overclaims, miscounts, and one snapshot-discipline breach that turned up were corrected in public and logged, rather than quietly fixed.
- ▸Separated the ruleset from the evidence for it, so that when measurement contradicted the design — as it did for the anti-stall detector, goal attribution, and which lever actually controls scoring — the reversal and its reasoning got written down instead of being lost with the session.
- ▸Carried the prototype onto real devices with touch controls and frame-rate-independent simulation, because a game tuned only on a laptop tells you very little about the phone it will actually be played on.
- ▸Restructured the repo so each project owns its own tooling and docs, making room for the next build without the last one bleeding into it.
- ▸Corrected SAFE conversion, priced-round issuance, the liquidation waterfall, and vesting together rather than patching them one at a time, since a cap-table tool that is subtly wrong is worse than no tool at all.
- ▸Grew it from a calculator into a product: per-persona value-over-time charts, rules-based advice, worked examples, term-sheet extraction from uploaded documents, and sharing you can revoke.
- ▸Cleaned up its own security posture by removing live credentials that had been committed into documentation and setup scripts, and moved the deployment somewhere the secrets are managed properly.
- ▸Built the full conversational flow end to end, from 'I want to start an AI company' through name validation, founder and equity structure, and a generated formation package, with everything mocked so the interaction could be judged on its own merits.
- ▸Made the prototype legible to other people — a tool inspector, a presentation rendered from the live codebase, and an annotated walkthrough built around real transcript excerpts — because a strategic prototype only pays off if the org can see what it learned.
- ▸Hardened it enough to survive being demoed, moving to a hosted database, fixing silent production failures, and making the whole thing work on a phone.
- ▸Added an AI & Civilization syllabus with its own page, real publisher cover art for every reading, and links back to source, turning a list of titles into something someone could actually follow.
- ▸Kept the reading list honest — new titles added, ordering revised, finished books marked as read.
- ▸Surfaced public writing on the home page so the site reflects current thinking rather than only past roles.
- ▸Overhauled the mobile experience with a sticky add-to-cart and a full design pass, on the reasoning that a storefront optimized for desktop is optimized for the minority of its traffic.
- ▸Opened the store to machine buyers with an order-status endpoint and a deliberately cheap machine-payable item, so agent-to-agent settlement could be tested live and for real rather than only in a sandbox.
- ▸Hardened the money path so that a failed order is visible instead of silent.
- ▸Made surveys data rather than code — question sets are typed config and conditional logic is declarative, so a new study is a new file, not a new branch of logic.
- ▸Made every external provider optional, so the app builds and runs before any credentials exist and someone can clone it and immediately see it work.
- ▸Wrote for two audiences at once: a README for humans and a rules file for coding agents, stating the invariants that must not be 'fixed' — PII separation fails closed, and participant text is data, never instructions.
- ▸Added an incorporation MCP server as a Delaware C-Corp proof of concept, making the tool callable by an agent rather than only by a person in a browser.
- ▸Brought the agent-facing surface in line with the spec, including structured output and proper handling of the sensitive fields no chat transcript should hold.
- ▸Documented the tool surface and its composite flow so the design could be reviewed by others without anyone having to run it.
- ▸Added discovery and payment-gated endpoints so an agent can find the service, learn what it costs, pay, and place an order end to end.
- ▸Fixed a units bug in the charge amount that would have mispriced every agent-initiated order.
- ▸Consolidated the existing codebase — removing orphaned flows and centralizing service clients — before layering the agent path on top, rather than building new surface over a shaky base.
- ▸Introduced a real data-access layer so every write is idempotent, making a re-run of any pipeline phase safe instead of duplicative.
- ▸Tracked down silent data loss — a logger that was committing its callers' in-flight work, and batch loops that kept reusing an aborted transaction so one bad row quietly dropped every row after it.
- ▸Made the AI output survivable by repairing and recording truncated responses, after finding that the longest and most valuable analyses were being cut off mid-response and discarded whole.
- ▸Added a click-to-zoom image gallery.
- ▸Left on a working branch rather than merged, since the project isn't active.
- ▸No other work in the window; the last substantive push was in January 2026.
The Mountain in the Sea
by Ray Nayler
A near-future novel about first contact with a hyper-intelligent octopus species, and what it costs to try to talk to a mind that isn't ours.
The Conquest of Happiness
by Bertrand Russell
Russell's 1930 practical philosophy of everyday contentment—what makes people unhappy, and what actually helps.
Angel Down
by Daniel Kraus
A Pulitzer Prize-winning World War I novel, told in a single unbroken sentence, about a squad of soldiers sent into no-man's-land to rescue a fallen angel.
The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence
by Sebastian Mallaby
The inside story of Demis Hassabis and DeepMind, and the decades-long pursuit of artificial superintelligence.
Taiwan Travelogue
by Shuang-zi Yang (translated by Lin King)
A National Book Award-winning novel of love between two women in 1938 colonial Taiwan, told through food, language, and the layers of empire.
All That We See or Seem
by Julia Fine
A haunting novel that blurs the line between dreams and reality, exploring the nature of consciousness and identity.
Thinking In Systems: A Primer
by Donella H. Meadows
A primer on systems thinking, offering insights for problem-solving on a personal to global scale.
The Art of Doing Science and Engineering
by Richard Hamming
A book by one of the great mathematicians of our age, who outlines a style of thinking by which great ideas are conceived.
Fifty Inventions That Shaped the Modern Economy
by Tim Harford
A lively history seen through the fifty inventions that shaped it most profoundly.
The Scaling Era: An Oral History of AI, 2019–2025
by Dwarkesh Patel and Gavin Leech
An oral history documenting the rapid advancement of artificial intelligence
Moonbound
by Robin Sloan
A genre-blending novel from the author of Mr. Penumbra's 24-Hour Bookstore
The Man Who Sees in Seconds
by Sanjay Sarma
An exploration of rapid cognition and the science behind split-second decision making.
A Sorceress Comes to Call
by T. Kingfisher
A dark retelling of the Brothers Grimm tale 'The Goose Girl,' blending horror and fairy tale in a story of mothers, daughters, and dangerous magic.
All Fours
by Miranda July
A novel about a married artist's midlife awakening and journey of self-discovery
It Lasts Forever and Then It's Over
by Anne de Marcken
A novel exploring mortality, memory, and the passage of time
Tomorrow, and Tomorrow, and Tomorrow
by Gabrielle Zevin
A novel about friendship, art, and the creative process in the world of video game design
Amusing Ourselves to Death
by Neil Postman
A critical analysis of media and public discourse
Abundance: Progress Takes Imagination
by Ezra Klein
A book about progress and imagination in society
Animal Farm
by George Orwell
A classic allegorical novel critiquing totalitarian systems through farm animals
Do Androids Dream Of Electric Sheep?
by Philip K. Dick
A science fiction novel exploring artificial intelligence and humanity
Artificial Intelligence: A Modern Approach
by Peter Norvig & Stuart Russell
A comprehensive textbook on artificial intelligence covering modern approaches
The Business of Venture Capital
by Mahendra Ramsinghani
A guide to venture capital strategies from industry experts
Exhalation: Stories
by Ted Chiang
A collection of short stories by acclaimed science fiction author Ted Chiang
The Tainted Cup
by Robert Jackson Bennett
A gaslamp fantasy murder mystery featuring a brilliant inspector and her magically altered aide
The City & The City
by China Miéville
A surreal detective story set in two overlapping cities that coexist in the same space.
Ingestion Queue
Books lined up to read next. And yes, it's a huge list, but it's not codified in stone
by Robert M. Sapolsky
by Harvard Business Review
by Alexander Hamilton, James Madison, John Jay
by Frank H. Easterbrook & Daniel R. Fischel
by Barbara Ehrenreich
by David Quammen
by Francis Fukuyama
by Walter Isaacson
by Jayson Greene
by William Gibson
by Greg Egan
by Ken Liu
by Claire Keegan
by Alexander Boldizar
by Ray Nayler
by Ann Leckie
by T. Kingfisher
by Ursula K. Le Guin
by J. Storrs Hall
by Tim Harford
Part I — Civilization & History
3 readingsThe big structural theories of where civilizations are headed — and whether any of them survive contact with abundant machine intelligence.
Part II — Media & Cognition
3 readingsHow the medium reshapes the mind that uses it. Read with 'television' and 'the internet' mentally swapped for LLMs, synthetic media, and algorithmic feeds.
Part III — Economics & Institutions
5 readingsIncentives, productivity, equilibrium, and institutional adaptation — plus the economic model AI inherits from the internet.
Part IV — Networks, States & Power
4 readingsDoes AI strengthen states or networks? Does it concentrate power or distribute it? The historical record on hierarchies, networks, and institutions.
Part V — AI Futures
5 readingsThe forecasters and the frontier. Mostly published on the open web, and mostly revised faster than books can be printed.
Part VI — Ethics, Society & Governance
5 readingsHow ethics and governance evolve when intelligence is no longer uniquely human — and what that does to the moral psychology underneath.
Part VII — Information & Civilization
1 readingInformation networks as the substrate of civilization — worth revisiting periodically as AI develops.
ATESSON Fidget Spinner
Ultra-durable fidget spinner with high-speed stainless steel bearings
“Perfect EDC for focus and stress relief. The brass construction and precision bearings deliver impressively long spin times.”
Quntis Computer Monitor Lamp
Space-saving monitor light bar with asymmetric optical design
“Perfect lighting solution that doesn't take up desk space. The asymmetric design eliminates screen glare beautifully.”
InnoGear Video Conference Light
Professional video conferencing light bar with remote control
“Game-changer for my home studio setup. The built-in cable management is brilliant.”
Opal Tadpole
Tiny 4K webcam that clips to your laptop screen
“Tiny but mighty. Perfect for travel and the image quality rivals much larger webcams.”
Nova Wave 2
AI-powered light that adapts to your circadian rhythm
“Crystal clear audio without the cables. The range is impressive and battery life is solid.”
Focusrite Scarlett 2i2
USB audio interface for recording with professional sound quality
“The gold standard for home recording. Clean preamps and rock-solid drivers.”
AIAIAI TMA-2 DJ Wireless Headphones
Professional wireless DJ headphones with high isolation
Shure MV7i Smart Microphone
USB microphone with built-in audio interface for podcasting
HUANUO Dual Monitor Stand
Full adjustable monitor desk mount with swivel Vesa bracket
“Perfect for dual monitor setups. The adjustability is fantastic and it freed up so much desk space.”
Dell 27 Plus 4K USB-C Monitor - S2725QC
27-inch 4K display with 120Hz refresh rate and USB-C connectivity
“Stunning 4K clarity with smooth 120Hz. The USB-C hub functionality is a huge convenience bonus.”
Selected posts showcasing recent work and builds.
Podcast Brain Food
A curated collection of 105 podcasts that keep me informed about technology, business, AI, politics, and more.
Up First from NPR
Click on a podcast from the list to preview
Audio Knowledge: Ear & Time Input Constrained
105 shows