THE IDEAS REGISTER · vault updated 17 Nov 2026 · 39 ideas
Every idea, kept. Take one.
Each week the show throws off ideas, and I do not build them. I write them down with the
cheapest experiment that would tell me something, and I leave them here. Generating is free.
The mistake is generating with nowhere to revisit. If one of these is yours to run, run it, and tell
me what happened. Once a month I read the outcome column back, and that is an episode.
RULE 1
Ideas are unlimited
Nothing here is deleted. "Dropped" means tried or considered, and here is what happened.
RULE 2
Builds are one
Exactly one project holds the build slot at a time. An idea does not have to become a project to survive.
RULE 3
Every idea has an experiment
The cheapest thing that would tell me something. Seconds to write, usually an evening to run.
RULE 4
Statuses are results
open 37 · open — promised on air 2026-09-11 1 · prototype 1
39 of 39
AI futures — a source-linked experiment
9 Oct 20261 ideas
AI-01
AI Futures Atlas: from competing futures to independent tests
connect observed benefits, documented harms, and plausible scenarios to their mechanisms, requirements, safeguards, and evidence. An independent experiment inspired by the AI futures discussion with Emad Mostaque; exact episode quotation and timestamp still need verification.
The experiment
Explore the working prototype: six cases, including space autonomy. Challenge one pathway, inspect its sources, and identify the independent test that would change your mind.
NAME ON ITprototypefor Gonçalo
outcome · Prototype created 9 October 2026. Proposed safeguards remain untested. Includes a video walkthrough and exportable evidence cases; no claim of endorsement or a complete AI futures map.
A visual interpretation of what connects across episodes
pull what was said on tape (claims, numbers, people, positions) and show how it links across five, ten, fifty episodes: who said what first, what came back, what changed. The vault graph is the seed; this is the version built from the tape itself
The experiment
Too early to build. First: tag the per-episode decompositions consistently (entity · claim · date · speaker) for four more episodes, then see whether the cross-episode links are worth drawing. If a plain table already shows the pattern, the visual can wait
TOOL TO LEARNopenfor Gonçalo
outcome · Parked on purpose, 2026-09-11: "probably too early… there will be nodes of this."
model-based systems engineering applied to the Krav syllabus @1:07:37. "Why isn't my curriculum agentic in the sense that everyone can consult it, can talk with it, can view it? … The same way you'd introduce a change to a car — where are the parts that it touches?"
The experiment
Take one belt level, write it as a system with dependencies, and ask one instructor a question the document should be able to answer. An evening, not a month
NAME ON ITopenfor SD4ALL / Gonçalo
outcome · ⚠️ This is a build and the slot is held by Do Not Die On The Mat → ACTIVE. It stays here until the slot is deliberately swapped
@1:22:31 — refurbish public libraries into AI education places. He proposed it independently; the library is already in CNN's list of what Quincy's data-centre tax revenue built
The experiment
Content first, and it is the best segment on the tape. Pair it with Drammen (Norway) and Barreiro (Portugal), which he already knows
CONTENTopenfor Gonçalo
outcome · ⭐ The precedent was inside the episode he was reacting to. That is the whole show in one beat
built from the dating-matchmaking segment @1:10:16 — same market-awareness approach, pointed at helping clients find the right person to connect with ⚠️(see README)
The experiment
Take 10 current AJC clients, hand-match them by the criteria the segment describes, ask Ana whether the matches are ones she'd have made
@NASAAdmin states policy weeks before it becomes a solicitation with a deadline
The experiment
Log every @NASAAdmin policy post and the date the matching SAM.gov notice appears. After a month you have the lead time — that is a publishable number nobody else has
NAME ON ITopenfor Gonçalo
outcome · The measurable version of "follow Isaacman"
He is "very open" to more prize competitions. Watch America COMPETES prize authority announcements; a new prize is the launch event for how to win a NASA challenge
BUSINESSopenfor Gonçalo
outcome · Direct line from the administrator to the course