These questions repeatedly circle three concerns: making better judgments, making good decisions easier to carry out, and turning knowledge into experiences worth participating in.
This catalogue retains the original 21 directions, the first 15 additions and nine questions from a second expansion, for 45 questions in total. It is organized around what is worth investigating, not the assumption that ambition, novelty or a working webpage establishes effectiveness.
Five suggested starting points
The initial shortlist. This is an editorial suggestion, not a commitment to launch these projects.
01 / 45Does an AI explanation teach, or just feel clear?
A research need expressed by the author
After an AI explains a chess principle, can learners make better independent judgments in unfamiliar positions?
Suggested test and important limits
Smallest useful test:Compare explanations with answers alone on fixed new problems, using independent and delayed tests. Verify explanations before any appropriately approved recruitment.
Boundary:Immediate success is not retention or transfer. Inclusion here does not start recruitment or outreach.
02 / 45Can investment skill be measured like chess strength?
An idea directly expressed by the author
How can we separate prior judgment, risk, position discipline and eventual luck?
Suggested test and important limits
Smallest useful test:Use authorized historical records or fictional cases to record available information, probability estimates, positions and outcomes separately. Test whether scores reflect decision quality.
Boundary:Calibrate the measure first. Profit does not establish skill, and loss does not establish a bad prior decision. A scenario score is not a certification of investment ability.
03 / 45A spending ruler: what did the money buy?
An idea from the author, including combining model judgments
Can evaluating purchases reveal better alternatives rather than merely attach scores to spending?
Suggested test and important limits
Smallest useful test:For ten voluntarily recorded purchases, separate prior expectations, actual use and later feelings. Compare AI suggestions with simple self-assessment.
Boundary:Averaging two models does not create objective truth. Art, pleasure and relationships cannot all be reduced to money saved.
04 / 45An opportunity map for choices about where to live
An idea directly expressed by the author
What comparisons can CFPS, CHFS and regional indicators support, and what can they not tell us?
Suggested test and important limits
Smallest useful test:Start with one city. State sources, years, geographic coverage and missing data, then ask readers what they infer from the map.
Boundary:Respect data-use permissions. Regional proxies are not individual longitudinal estimates, and correlations are not the causal gains from moving.
05 / 45Make jazz, food and movement into crafts in a fantasy world
An idea directly expressed by the author
Could music, food, exercise and pharmacology shape a character’s actions as concretely as sword-making can shape a story?
Suggested test and important limits
Smallest useful test:Write an original short story in which a meal or a piece of music changes a character’s choice. Check whether the knowledge actually affects the plot.
Boundary:The first test need not establish commercial value. Distinguish fictional effects from medical facts; this is not a medication guide.
Investment and decisions
Separate judgment, execution and outcomes before asking whether an advantage repeats.
06 / 45A rule-locked wallet: can constraints reduce impulsive choices?
An idea expressed by the author in 2024
An earlier concept used AHR999 thresholds to constrain buying and transfers. The prior question is whether rules improve execution, not whether the indicator makes money.
Suggested test and important limits
Smallest useful test:Simulate rules, temptations to break them and emergency exits without real funds. Record discipline, costs and returns separately.
Boundary:Indicator validity, emergency access and custody security are distinct issues. This list does not recommend locking assets or automated trading.
07 / 45Clock: only two actions in four years
An idea directly expressed by the author
Could an extremely low-frequency product save attention? Accurate market timing is a separate question.
Suggested test and important limits
Smallest useful test:Compare checking frequency, effort and departures from a simulated plan. Test the timing hypothesis independently.
Boundary:Fewer actions do not imply lower risk. Two actions in four years is a product concept, not a validated investment pattern.
08 / 45What mechanism could explain Bitcoin cycles?
A question directly asked by the author
Could halving, liquidity, leverage or a deeper mathematical structure explain apparently repeating price paths?
Suggested test and important limits
Smallest useful test:Freeze peak and trough definitions before comparing competing explanations with simple baselines, retaining failures and out-of-sample tests.
Boundary:There are few independent cycles and many opportunities for hindsight. Existing public work did not find a reliable bottom clock; an open question is not a forecast.
Related published article ↗
09 / 45From Deep Research to automatic trading: what is missing?
An idea directly expressed by the author
What timing records, benchmark comparisons and risk checks should a research signal pass before execution is considered?
Suggested test and important limits
Smallest useful test:Use offline or simulated tests only. Retain every signal, abstention, creation time, contemporaneously available input and baseline result.
Boundary:Do not select only successes or backfill signals. Inclusion does not authorize live trading or restart earlier experiments.
10 / 45Investment software, independent games and interaction
An idea directly expressed by the author
Could participation improve understanding of risk without making trading more exciting or frequent?
Suggested test and important limits
Smallest useful test:Create a short interactive story and compare understanding of risk and uncertainty before and after it.
Boundary:Validate one experience rather than building investment, gaming and social platforms at once. Use fictional situations without real trades.
11 / 45Autopilot: build, invest or acquire?
A concept in the author’s 2023 memo
When an investment software business is split into technology, acquisition and compliance, where are the barriers and missing capabilities?
Suggested test and important limits
Smallest useful test:Review the old assumptions against current public evidence on customer acquisition, differentiation and compliance.
Boundary:An old memo is not a current valuation. No historical personal financial figures are published; this is not an investment or acquisition recommendation.
12 / 45The FT Lex headline-judgment experiment
A model task set by the author in 2023
Does classifying a headline as positive, negative or uncertain add predictive information beyond a simple baseline?
Suggested test and important limits
Smallest useful test:Use a complete, appropriately licensed historical signal set with a frozen evaluation window, benchmark and cost definition. Retain uncertain answers.
Boundary:A few successful examples cannot establish effectiveness. A headline judgment is not a full company analysis.
13 / 45Bitcoin short-term reversal: a signal or just holding exposure?
A strategy research topic discussed by the author
Would returns come from a reversal signal or simply from holding Bitcoin and taking risk?
Suggested test and important limits
Smallest useful test:Compare the same period and data with a holding baseline, transaction costs and notification delays, then test out of sample.
Boundary:No personal trading results are published. Available material does not establish an effective strategy.
Products, life and creative work
Make one concrete experience work before expanding it into a product.
14 / 45A parenting simulator for exploring family trade-offs
An idea directly expressed by the author
How could economic conditions, random events and family choices create a game that helps players understand trade-offs?
Suggested test and important limits
Smallest useful test:Build a ten-minute scenario. Check whether players can explain the costs of different choices and want to replay.
Boundary:Do not turn genetic risk into numerical destiny or treat game outcomes as real parenting advice.
15 / 45Serendipity: unexpectedly finding a worthwhile question
A product and naming idea from the author in 2023
Could an AI investment product go beyond a news feed to reveal connections worth pursuing?
Suggested test and important limits
Smallest useful test:Manually curate a small set of unexpected connections and observe whether readers generate new, concrete questions.
Boundary:Novelty is not information value. Inclusion does not commit the Lab to a new name or brand.
16 / 45Theatre owner, playwright and actor
A creative ambition expressed directly by the author
Which aspects of running, writing for and performing in a theatre could be experienced through one small work?
Suggested test and important limits
Smallest useful test:Stage an original reading or short prototype, noting what writing, rehearsal and audience responses each reveal.
Boundary:Make a work before forming a company. Adapting private experiences requires separate decisions about disclosure and the people involved.
17 / 45A nearly frictionless inbox for handwritten ideas
A need directly expressed by the author
With an iPad and Apple Pencil, is the main obstacle capturing, organizing or finding an idea again?
Suggested test and important limits
Smallest useful test:Use existing tools for a week. Record when ideas are lost, retrieved or revisited before deciding whether new software is needed.
Boundary:More notes do not guarantee better reflection. Validate the problem before building another system to maintain.
18 / 45Does my taste predict the market’s taste?
A question from the author, reframed here as a taste experiment
How does liking a work differ from predicting that the wider market will like it?
Suggested test and important limits
Smallest useful test:Record personal liking and market forecasts separately before outcomes, with a predefined review date and success criterion.
Boundary:An unvalidated questionnaire should not be used to label people as harbingers of failure.
Longevity and science
Keep the ambitious questions and the gaps between animal studies, human evidence and individual use.
19 / 45What would let a human body remain repairable?
Pursued by the author; authorship of the initial wording is unconfirmed
What damage, repair capacity, tissue reserve and information loss would need to remain controlled for function to keep recovering?
Suggested test and important limits
Smallest useful test:Create a conceptual inventory separating measurable variables, existing experiments and unsupported hypotheses.
Boundary:The English question is not claimed as the author’s original coinage. Decomposition is not a longevity formula or a treatment plan.
Related published article ↗
20 / 45Can dynamical systems help us understand aging?
The author connected mathematics with lifestyle
Could a model of external forces, changing states and recovery capacity produce better questions than a single indicator?
Suggested test and important limits
Smallest useful test:Begin with a clearly labeled toy model, specifying its variables, assumptions and falsifiable predictions.
Boundary:A mathematical analogy does not establish an escape from death or justify self-directed drug combinations.
Related published article ↗
21 / 45Language, music and future health: can they inform preparation?
An idea directly expressed by the author
Could behavioral signals provide reliable health information and then improve care and financial preparation?
Suggested test and important limits
Smallest useful test:Narrow the idea to short-term prediction on authorized data. Check calibration and external validity, then test decision value separately.
Boundary:It cannot reliably estimate an individual’s disease probabilities decades ahead. Prediction, diagnosis and financial decisions require distinct validation.
Related published article ↗
Nine additional directions
Drawn from additional questions and ideas reviewed in this round. The experiments are proposed next steps.
22 / 45When does adding a human make AI performance worse?
A question directly asked by the author
With equal time, information and tools, can human–AI collaboration reliably outperform AI alone? Does the person add information, selection or error correction?
Suggested test and important limits
Smallest useful test:Use chess or code tasks with checkable answers. Compare human, AI and combined conditions, recording accuracy, time and correct answers changed to wrong ones.
Boundary:Historical chess results cannot substitute for tests of today’s models. This differs from asking whether AI teaches people.
23 / 45Can a personal AI workbench cost less without losing quality?
The author expressed the need; AI developed the workbench proposal
Could routing tasks by difficulty across models, memory, tools and checks reduce the cost of completing real work?
Suggested test and important limits
Smallest useful test:Freeze twenty tasks and acceptance criteria. Compare a fixed strong model with routing on completion, cost, delay and human rework.
Boundary:Rebuilding a workflow is not copying a closed model’s capabilities. Multiple calls may cost more; set budgets and stopping rules.
24 / 45What musical structures might help ideas emerge?
An idea directly expressed by the author
Could rhythmic repetition, harmonic change and surprise affect association, reading and exact reasoning differently?
Suggested test and important limits
Smallest useful test:Use original or authorized audio and randomized crossover comparisons with silence. Separately assess idea count, blindly rated novelty and reasoning accuracy.
Boundary:Acoustic properties alone do not establish cognitive effects. A small personal test cannot support universal therapeutic claims.
25 / 45What does a concert ticket actually buy?
A question directly asked by the author
Are audiences buying music, belonging, identity, shared memories or an evening away from routine? Which needs could another service address?
Suggested test and important limits
Smallest useful test:Interview eight to ten willing attendees of one show about expectations, valued moments and disappointments. Test one recurring need on a small scale.
Boundary:High ticket prices do not establish willingness to pay for companion software. Demand remains to be tested.
26 / 45An event radar for opportunities people can still act on
An explicit need from the author, generalized into a product question here
Could an alert distinguish a newly found event from a still-actionable opportunity, with enough notice and verified ticket availability?
Suggested test and important limits
Smallest useful test:Cover one activity type in one city for two weeks, recording announcement and event dates, availability, lead time, false alarms and missed opportunities.
Boundary:A page’s update date is not the event date. Timeliness and usefulness need continuing verification.
27 / 45A time loop where each player holds a different truth
An idea from the author; AI generated the specific story outline
How can time loops, fragmented memory, factions and hidden identities interact rather than merely coexist?
Suggested test and important limits
Smallest useful test:Create an original thirty-minute paper prototype for four to six players, using one central loop and one major information reversal. Test fairness and participation.
Boundary:An outline cannot guarantee depth or sales. Use original characters, wording and puzzles when drawing on genre mechanisms.
28 / 45A darkly comic documentary short about ordinary city life
An idea directly expressed by the author
How could observation, narration and editing turn everyday streets into an absurd but revealing story?
Suggested test and important limits
Smallest useful test:Pick one small how-to theme, capture ten observational shots and edit three minutes. Ask viewers who did not know the theme what they remember.
Boundary:Develop an original perspective rather than copying a creator’s distinctive expression. Respect privacy and obtain appropriate consent for identifiable people, private places and sensitive situations.
29 / 45Ten parallel endings for a choice about the future
The author requested the format; AI generated the specific endings
Could warm, absurd and unsuccessful endings reveal value conflicts hidden by a single forecast?
Suggested test and important limits
Smallest useful test:Start with three endings for one public technology topic. Check whether readers can identify the key condition behind each branch before expanding.
Boundary:Label science fiction and alternative histories clearly. Do not present fictional events involving real people as facts or predictions; use no private experiences here.
30 / 45How would matching software prove it helps people connect?
The author proposed this product type; the evaluation question is a new framing
Should a matching product optimize profile similarity, clicks or both people’s willingness to keep talking?
Suggested test and important limits
Smallest useful test:Start with original fictional profiles to compare explanations and obvious mismatches. Any real-person trial needs willing adults, informed consent, privacy and an exit route.
Boundary:Working software does not establish network effects or matching quality. Do not copy brands, code or profiles, or infer sensitive traits from private chats.
AI-proposed extensions
These six are not claimed as the author’s original ideas. One was subsequently quoted and pursued by the author.
31 / 45Do several AIs correct each other or lead each other astray?
An AI-proposed extension, not claimed as the author’s original idea
At equal cost, which is more reliable: one model reasoning longer, independent votes or a discussion among models?
Suggested test and important limits
Smallest useful test:Freeze a task set and retain answers before and after discussion. Count correct answers that become wrong through persuasion.
Boundary:This is an untested research question, not an established result.
32 / 45Can AI improve individual writing while making works more alike?
An AI-proposed extension, not claimed as the author’s original idea
How do AI-first drafting and author-first ideas with later AI critique differ in quality and distinctiveness?
Suggested test and important limits
Smallest useful test:Compare both workflows on one theme with blind ratings of quality, memorable elements and similarity across works.
Boundary:This is an untested research question, not an established result.
33 / 45When AI seems to understand: personalization or general wording?
An AI-proposed extension, not claimed as the author’s original idea
Without identity cues, can people distinguish genuinely personalized descriptions from ones that fit many people?
Suggested test and important limits
Smallest useful test:Use authorized nonsensitive material from willing participants. Separately rate specificity, falsifiability and subjective fit in a blind test.
Boundary:This is an untested research question, not an established result.
34 / 45See your own problem from an observer’s perspective
An AI-proposed question subsequently quoted and pursued by the author
Could third-person reflection improve consideration of uncertainty, other viewpoints and long-term consequences?
Suggested test and important limits
Smallest useful test:Use fixed reflection prompts and crossover first- versus third-person versions, scored blindly against predefined criteria.
Boundary:This is an untested research question, not an established result.
35 / 45Why does a good twist feel surprising and inevitable?
An AI-proposed extension, not claimed as the author’s original idea
Did viewers actually notice the clues, or reconstruct their memory after learning the ending?
Suggested test and important limits
Smallest useful test:Record predictions and noticed clues before the reveal, then collect recall afterward and compare the records.
Boundary:This is an untested research question, not an established result.
36 / 45Does telling someone a prediction change the outcome?
An AI-proposed extension, not claimed as the author’s original idea
When does a prediction fulfill itself, and when does it prompt actions that undermine it?
Suggested test and important limits
Smallest useful test:Start with a rule-based game without real-world consequences, comparing behavior when predictions are disclosed or withheld.
Boundary:This is an untested research question, not an established result.
Second expansion: nine new questions
Added after individual review, retaining attribution. Related but distinct questions remain separate.
37 / 45Can an AI character remain consistent throughout a story?
A proposal in a historical file, with original authorship unconfirmed; the test is refined here
Could role-play, chat, text and audio form a story where characters remember events, follow world rules and respond to player choices?
Suggested test and important limits
Smallest useful test:Build one ten-minute scene with two original characters and three world rules. Compare ordinary chat with state-tracked characters on memory errors, conflicting motives and willingness to replay.
Boundary:Use original characters and authorized materials. Narrative consistency does not imply that a character has a real personality or personhood.
38 / 45Which impossible triangles are actually impossible?
A question directly asked by the author; the audit method is developed here
Can popular trade-offs be separated into mathematical results under stated assumptions, current resource constraints and untested slogans? Which boundaries move when conditions change?
Suggested test and important limits
Smallest useful test:Choose five familiar triangles. Specify their three goals, assumptions and counterexamples, mapping jointly improvable regions and identifying rigorously proven claims.
Boundary:Different domains do not share one automatic theorem. Good, fast and cheap is not itself a mathematical proof.
39 / 45What kind of question actually changes the next step?
A question directly asked by the author; the evaluation design is developed here
Beyond sounding profound, can a question distinguish competing explanations, expose important unknowns and change a choice?
Suggested test and important limits
Smallest useful test:Write five questions about the same material and ask readers who do not know their authorship to answer. Record new findings, testable judgments and changes in decisions.
Boundary:A question’s value depends on goals and prior knowledge. Avoid a universal profundity score.
40 / 45When does setting a problem aside help?
The author explored related activities; reframed here as a testable question
After engaging seriously with a problem, might walking, simple manual work or rest help more than continuous focus? Which tasks improve, and which merely feel better?
Suggested test and important limits
Smallest useful test:Rotate continuous thought, low-demand activity and quiet rest with fixed preparation time. Later assess problem solving and creative quality, keeping condition labels from evaluators where possible.
Boundary:Feeling inspired is not the same as being correct. Both an incubation effect and its explanation require evidence rather than subjective impressions.
41 / 45What environments make intuition worth trusting?
A question directly asked by the author; the calibration design is developed here
Does an intuitive judgment reflect learned regularities or mere familiarity? Could environmental predictability and feedback quality help distinguish the two?
Suggested test and important limits
Smallest useful test:On tasks with checkable answers, record a quick judgment and confidence before allowing deliberation. Compare accuracy, overconfidence and gains from revision in familiar and unfamiliar domains.
Boundary:Do not use this to label someone’s ability or psychology. Small samples do not generalize to every life decision.
42 / 45When does honesty create a lasting advantage?
A question directly asked by the author; the simulation is proposed here
How do repeated interaction, verifiability and reputation change the effects of honesty, withholding information and declining to answer on cooperation?
Suggested test and important limits
Smallest useful test:Simulate repeated interactions among fictional agents while varying the chance of meeting again, misinformation costs and verification effort. Track sustained cooperation, not only one-round payoff.
Boundary:Distinguish truthful answers, refusal to disclose and deception. Payoffs do not replace ethical or legal judgments. Do not conduct unconsented deception experiments on real people.
43 / 45How much did one person actually change the world?
A question directly asked by the author; the counterfactual framework is developed here
Without a particular contributor, how much later might a discovery have appeared? How should alternatives, dissemination, lasting benefits and harms enter the evaluation?
Suggested test and important limits
Smallest useful test:Build public-source timelines for three historical technologies and identify alternative contributors. Use several explicit measures to describe scenarios and their sensitivity to assumptions.
Boundary:Avoid a falsely precise single ranking. Do not equate one person’s contribution with an entire team’s or institution’s work.
44 / 45Can a brilliant analogy create an illusion of understanding?
An AI-proposed extension in this round, not an original proposal from the author
Does a cross-domain analogy make a concept more memorable, or encourage people to mistake a comparison for a mechanism?
Suggested test and important limits
Smallest useful test:Compare a direct explanation, an analogy and an analogy with explicit limits. Test transfer to unfamiliar situations and collect confidently wrong conclusions.
Boundary:Analogies can generate questions but cannot replace evidence for a mechanism. State their scope and the ways they could mislead.
45 / 45What kind of imitation helps an original work emerge?
The author made cross-domain connections; reframed here as a research question
Does imitating a work’s appearance, structure or creative process best support original work in a new subject?
Suggested test and important limits
Smallest useful test:With one short-story prompt, compare surface imitation, structural analysis and reconstructing the process before changing subjects. Ask readers to rate independence, effect and similarity without knowing the condition.
Boundary:Use authorized or otherwise lawfully usable material. Do not treat unverified quotations or attributions to famous people as established facts.
If choosing just three small attempts
One possible combination is a chess-learning test, a small spending-ruler trial and an original theatre or fantasy piece. The first two test judgment tools; the third makes an imagined world available to read, watch or participate in.
This is an editorial suggestion, not a fixed schedule. The list supports a more informed next choice and leaves room to revise, combine or set aside a question after testing it.