Preview · what your reports look like
The AI tells you what it thinks of you, and what it has learnt about you: where you spike, what drives you, the patterns you fall into, and how you actually work. Here is what that looks like, shown for John, a sample profile.
Shared only if you choose. The one report you can publish — you approve every line before it goes anywhere.
John
A solo founder who has spent ten months and four pivots pursuing one question: how to change what society measures so people keep developing and finding opportunity after AGI.
Based in San Francisco, California.
Background
Solo founder · past 10 months. Four pivots toward human development and opportunity after AGI. Currently building Meridian Labs out of Mox; previously part of South Park Commons and Founders Inc.
Technical Staff · Cinder Labs, Jan–Mar 2026. Built RL environments for computer-use agents and solo-built an agentic coding harness that anchored the company's largest frontier-lab contract; the company was acquired within the year.
Machine Learning Researcher · Microsoft Research, May–Jul 2022. Built a controllable video-captioning algorithm that became a filed patent.
Cornell University · MS Computer Science, 2023–2024. GPA 3.98/4.0.
Georgia Tech · BS Electrical Engineering and Computer Science, 2019–2023. GPA 3.97/4.0; one NeurIPS workshop paper.
Putnam Competition: top 150 of roughly 4,000 entrants.
National top 1% in the Physics, Chemistry and Astronomy Olympiads.
Goldwater Scholarship for undergraduate research.
One NeurIPS workshop paper and one filed US patent in AI research.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
What he is doing now
Currently building Meridian Labs: paid, expert-judged two-week work trials that produce verified skill records and connect overlooked talent to jobs, collaborators and communities.
He works solo and is looking for collaborators. The long ambition is to build what gives humans purpose and opportunity after AGI.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
General Reasoning
Uses ChatGPT as a place to reason across company-building, education, relationships, markets, philosophy, science, technology and art rather than staying inside one field.
- Over eight weeks: company-building appeared on 35 days, education and talent on 29, relationships on 27, the Gita, Vedanta or meditation on 9, and consciousness on 5. Topics overlap.
- In markets, he explored why capital intensity, vertical integration and scale advantages push electric vehicles, commercial space and automated manufacturing toward oligopoly.
- In neuroscience and biology, he compared ultrasound with EEG and fNIRS for non-invasive BCI, and asked whether ageing is an accumulated-damage problem or a biological clock that could be reset.
- Other threads ranged from projection versus knowledge of a partner, to whether art aimed at a predetermined outcome can discover anything, to whether universes might model their own possible futures.
See more· private to you
Cold on IIT, re-derived three serious published objections (incl. Aaronson's reductio), built a two-axis synthesis importing Deutsch's 'universal explainers,' and forced the theory to its most uncomfortable commitment — full generative+critical loop (route a).
“I think state temporal continuity is so much more elegant of a definition”
Split consciousness into two clean orthogonal axes that survive counterexamples and carry to new cases (XOR grid, boxed transformer), then overturned his own first disqualifier for the deeper one.
“the real disqualifier for the transformer isn't intermittency — it's autonomy”
Out-reasoned the tool on a hard modeling question — rejected the additive estimate for multiplicative, then caught that the gates were correlated; the AI conceded 'they're not independent gates.'
“No you're calculating it wrong - make it multiplicative based on all the things panning out”
Named the single hinge his whole company rests on and re-derived mechanism cold; caught the fatal flaw in his own moat.
“This all hinges upon if different memory scaffolds and RL architectures needed for different people or not.”
Usually identifies the decision that matters, rejects attractive options when they fail the larger objective and changes direction when evidence turns against the current plan.
- Killed Ledgerline despite finding a real market gap. The ceiling was limited and Airtable controlled the distribution.
- Ended a productive cofounder trial. Several months of prior investment did not outweigh the long-term fit concerns.
- Moved from teaching to talent measurement. Several education products showed that learning without a trusted outcome would not hold attention.
See more· private to you
A string of high-stakes personal calls under maximal pull, landing right nearly every time, with explicit loss conditions and a checkpoint discipline most people never write down.
Hurt by a disclosure, refused to let emotion set the call — structured it as a clean conditional and landed calibrated rather than reactive.
“if anything my trust in her has increased that she told me”
Resisted the seductive surface read of a flaky co-lead ('high agency') and cut him against sunk-cost pull.
“What is the likelihood that I am wrong in cutting him”
Frequently raises objections to his own position before another person does.
- Audited five failed products. He concluded that optimism had repeatedly overridden evidence about incentives, behaviour and distribution.
- Caught confirmation-seeking during the Ledgerline decision. When the AI agreed with both sides, he stopped the discussion and demanded an argument that could change the decision.
- Questioned a developing superiority belief. He deliberately looked for evidence of other people outperforming him rather than protecting the belief.
See more· private to you
Attacked his own most ego-protected estimate (a deeply personal forecast), named the direction of his own distortion, and demanded the structurally harsher model rather than the flattering one.
“No you're calculating it wrong - make it multiplicative based on all the things panning out”
Audited the DIRECTION of his own bias on someone he loves, built explicit anti-motivated-reasoning machinery, and acted on it.
“Decision rule — when in doubt, always trust prior data over current change theory.”
Voluntarily dismantled a flattering self-comparison to expose that his own bet is riskier, and caught a fatal flaw in his own moat unprompted.
“I cannot learn from other people's data that would be super bad for me”
Conceded his own ego-distortion on spiritual specialness within the same breath, generating the disconfirming evidence himself.
“ahhh there is not a lack of attachment to praise, need to compare, subtle superiority... maybe i am not very clear”
His strongest taste is in messaging and product presentation: deciding what a page must communicate, replacing vague praise with proof and removing elements that weaken the central idea. There is less evidence about visual design.
The same pattern appears across landing-page copy, public-report structure, demos, launch material and his personal website.
- Rewrote the landing page around the actual promise. When the generated opening sounded like ordinary AI marketing, he replaced it with: “A resume says where you sat. Two judged weeks of real work say what you can do.” He paired it with “Proof, not polish” and used the product’s terminal aesthetic instead of a generic SaaS page.
- Turned the trial report from praise into evidence. The first version was a long free-form profile with more than eight top-percentile badges. He gave it fixed sections for craft, judgment, speed and collaboration, required concrete moments beneath every score, and limited the opening to the three rarest strengths so readers could judge the person rather than trust promotional copy.
- Replaced impressive-sounding language with facts a stranger could understand. He rejected lines such as “world-class strategist who thinks natively in systems and incentives” and required each sentence to stand alone using named books, products, decisions, counts or direct examples.
See more· private to you
Curatorial diet in a single session spans Don Norman (canonical) → Dieter Rams Ten Principles → the Phaidon Rams monograph by Sophie Lovell (a substantially more specific artifact requiring genuine design-history literacy beyond surface-level exposure) → George Nelson (non-mainstream American modernist) → Paul Graham's 'How to Do Great Work' — a discriminating, historically-grounded, non-mimetic design canon assembled without prompting
“Dieter Rams: As Little Design as Possible free pdf”
Creative philosophy notes articulate an independent, sophisticated insight about the failure mode of instrumental art ('Can't make art with an end in mind') — correctly naming why art-made-to-achieve-an-effect fails aesthetically — plus a minimalist aesthetic held as a named operative principle and a precision-about-gestalt quality judgment that a person with merely emerging taste would not reach independently
“A deliberate approach to constrain meaning is really bad. It constrains does not help. Can't make art with an end in mind.”
A classical-music thread sustains demanding aesthetic refinement across multiple turns — refuses mediocre imagery and articulates the structural reason ('Sky is lazy imagery. That's stock-photo thinking.'), makes fine distinctions (one mode as pure and gentle, another as sensual romance), and actively shapes register and voice, showing a feel for what is good that is reliable and reasons-backed, not just reactive
“Sky is lazy imagery. That's stock-photo thinking.”
Ideas he has been thinking about
Human development after AGI. If intelligence becomes abundant, money and professional achievement stop being reliable measures of human worth. That led him to measurement: the SAT shaped schools, the bar exam shaped law school and Leetcode reshaped computer-science education. Change what society rewards and people develop toward it.
Consciousness and meditation are separate questions. He has repeatedly explored what makes conscious experience rich, what might make an AI conscious, and theories such as integrated information theory. Meditation, the Gita and Vedanta are a different inquiry: how to deepen his own experience and pursue extreme ambition without making outcomes or destiny the basis of inner stability.
Exposure may matter more than formal education. He believes books and self-directed exploration contributed more to his development than Georgia Tech, Cornell or learning on the job. He repeatedly returns to whether institutions mainly select capable people rather than develop them.
Other thoughts and hypotheses
- AI autonomy: if an aligned superintelligence consistently makes better decisions than humans, what principled reason remains for overruling it—and is removing human agency the destruction of the thing alignment was meant to protect?
- AI governance: he has explored whether a superintelligence with granular economic information could coordinate society better than governments or markets, and whether democracy can move quickly enough for the coming rate of change.
- Institutions: selecting good people is insufficient; durable institutions need opposition pathways, contestability and limits on concentrated power because good incumbents cannot be assumed to remain good.
- Civilization’s bottleneck: he has questioned whether more individual intelligence stops helping once coordination and information flow become the binding constraints.
- X-risk and progress: he keeps asking when faster technological progress expands human possibility, when sufficiently powerful AI permanently removes human agency, and what remains worth building under that uncertainty.
- Company-building: a company is worth his life only when its absence leaves a meaningfully different world; this has led him to reject profitable products that another founder would predictably build.
- Founder optimism: after five product post-mortems, his rule became pessimism about incentives, distribution and stated user behaviour—but optimism about engineering and his own learning speed.
- Probabilistic thinking in practice: he uses explicit probabilities in company bets, cofounder questions and relationship decisions, while returning to Superforecasting, The Black Swan, Antifragile and Charlie Munger.
- Learning sequence: direction and intuition should precede detail. A ten-minute intuitive explanation unlocked technical derivations that rote coursework had made inert; this now shapes how he thinks education should be built.
- Meaning beyond one life: because individual experience ends, he locates durable meaning in humanity’s cumulative experience and in decisions whose effects persist after the person making them is gone.
- Spiritual ethics without certainty: “divinity in everyone” remains useful as an ethical rule across Vedanta, Christianity, Judaism and Islam even if none of the underlying metaphysics can be proved.
- Simulation as survival research: he has considered whether civilizations might run faster, lower-resolution universes to search for solutions to their own long-run survival before their stars die.
The analyzed corpus contains 33 retained ideas, general thoughts and epiphanies. These are the clearest public examples; many more remain in the private evidence.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
Agency & Work
Chooses work using his own impact criterion, including when the safer or more legible option is available.
- Rejected profitable and already-working directions. A finance product, vertical SaaS ideas and products with users were stopped when he believed someone else would build substantially the same world.
See more· private to you
Field-scale mission held up against two hard rejections and relocation pressure, with the safe path explicitly named and refused
“There is substantial odds that there is no business in what I'm working on. But if there is the whole field changes and ends up in benevolent hands”
Rejected the legible, family-advantaged vertical (a field his family gave him a head start in) on the counterfactual test
“Realized I am much more missionary than mercenary. I cannot work on something which does not have the potential to change humanity, or one where someone else would build it the same way I would.”
Conventional path structurally ruled out, derived as duty from first principles
“A job is not even an option for me.”
Uses civilization-level impact as a live filter for what to build, rather than as a description added after the decision.
The same filter appeared in 10 independent conversations across 13 months and four startup phases.
- The mission survived the products. Brio, Ledgerline, Briefwire and the earlier education products ended; the question of human development after AGI remained.
Moves quickly once a direction is clear, especially in product and startup work.
- Built six products in eight months. New directions followed the evidence from the previous product rather than remaining strategy documents.
See more· private to you
Stood up an entire cold-outreach stack from cold in one session — Apollo account, sequence, templates, domain email — and activated it same-session, driving the tooling himself
“Bro, can you actually help me create an Apollo account, want to do outreach to series A, B and C companies, I already have a list. Use agent mode”
Reports completed actions rather than intentions across onboarding and compliance execution
“I have alre...”
Decision crystallizes to action in a single sentence with no deliberation lag
“You're right, ok let me apply asap”
Kept returning to human development after AGI through failed products, distribution problems and collaborator changes.
- Six products, one mission. An AI tutor, Briefwire, mentored cohorts and AI-supervised work trials each failed for a different reason; he wrote the post-mortems and used them to narrow the next attempt.
- The fifth failure became the premise of the sixth. Teaching alone did not hold attention without a trusted outcome, so Meridian Labs moved the work toward evidence, credentialing and opportunity.
- Then built the new direction into a working system. The current production platform spans trial matching, expert-review tooling, verified records, payments, OAuth and deployment across 65 recorded coding sessions.
See more· private to you
Six genuinely distinct product theses attempted, five post-mortemed, going harder on the sixth at maximum urgency — the empirical signature of bouncing back every time
“Project 1 — Long term trend is here, plan seems right, pain seems acute. Let's execute really fucking hard on this now.”
Generated nine distinct routes against a block that would not move, then surfaced an unincorporated data point from his own memory that materially changed the read — persistence as search, not repetition.
Arrived at the business's root-cause before the AI and immediately generated the institutional route past it
“Biggest issue is that universities do not have incentive to get people jobs... Maybe make an institution that has incentive to get people jobs”
These are representative examples from 78 conversations graded for agency. The remaining concrete moments stay in the private dropdowns.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
Growth & Information Intake
Constantly thinks about how he can get better on his own, then names the specific pattern getting in the way.
- Separated motivation from ambiguity. Work often restarted as soon as the next action became concrete; the problem was not a lack of ambition.
- Found where optimism was overriding evidence. Five product post-mortems traced failures to incentives, behaviour and distribution rather than luck.
- Stopped treating isolation as a virtue. He noticed that daily intellectual company changed both work quality and emotional stability, and began treating a people-rich environment as infrastructure.
- Named the maximizer loop in personal decisions. Keeping another plausible model alive was sometimes preserving uncertainty rather than improving the choice.
See more· private to you
Diagnoses his own motivational architecture at the mechanism level and meta-diagnoses his own bias mid-spiral, codifying it as a forward rule — observer watching the observer.
“Root pattern: I overweight whatever is hardest to get. Naming it is the only thing that breaks it.”
Catches motivated reasoning running in REVERSE — bias AWAY from the thing he doesn't want to commit to — more sophisticated than standard bias detection.
“i dont want the memory thing to be a big company sinec im not really working on it.... so im afraid im also biased in my thinking”
Real-time system-level catch of a perverse reinforcement loop he himself created, named before the AI weighed in, and refuses the flattering euphemism for his own state.
Six-failure post-mortem naming the actual cognitive error each time (idealism-over-priors vs positioning vs market-structure), plus catching a subconscious superiority drift before the AI named it.
Reads across books, essays, blogs and feeds, then uses AI dialogue to question, connect and apply what he finds.
His public record names 37+ books. In the indexed eight-week ChatGPT window alone, at least 12 long-form works appeared across more than 20 threads, alongside 10+ named essays or blog series and 60+ substantive idea explorations.
See more· private to you
Interdisciplinary deep processing as the same motion — re-derived Aaronson's objection to IIT independently and synthesized an original two-axis consciousness model while running religion/moral-philosophy in parallel.
“IIT's fatal-feeling problem (which you re-derived as Aaronson's objection): dumb high-Φ systems like XOR grids come out 'highly conscious.'”
Explicit habitual near-daily intake plus live application and cross-domain synthesis (AI safety, music theory, decision theory, education) in a single session.
“Through books, I learned about relationships, economics, AI, builders and their mistakes, critical thinking, cosmology, philosophy, and sales. This has been, by far, the highest-ROI education of my life.”
Deep mechanism-level extraction from a book (BATNA math, delta/1+delta bargaining, signaling/screening) translated into personal directives and an original product extension.
“Convert books into reasoning - could be HUGE.”
Two years of voracious, explicitly interdisciplinary intake going to source material on outliers, with internalized applied learnings.
“I have been reading voraciously over the past 2 years, biographies, mental models, sales, marketing, anything that can help me get better at the game.”
Information footprint
- Judgment and economics: Poor Charlie’s Almanack, The Black Swan, Antifragile and Superforecasting.
- Founders and company-building: The Mind of Napoleon, Titan, Elon Musk, Traction, Founding Sales and Dealers of Lightning.
- Philosophy and the future: the Bhagavad Gita, Ashtavakra Gita, Meditations, Life 3.0 and The Singularity Is Near.
- Current writing: Ben Kuhn, Sam Altman, Paul Graham, Hamming, LessWrong and Situational Awareness; Twitter adds daily AI, funding, biotech, longevity, safety and rationality.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
Interpersonal
Care for specific people remains visible when a relationship becomes inconvenient, uncertain or cannot become what he wanted.
- Did not make care conditional on getting the relationship he wanted. He continued protecting the other person’s interests after accepting that the romantic outcome might not happen.
- Preserved another person’s opportunity after ending a work trial. He recommended the former co-lead to an evaluator who controlled his own next opportunity.
- Made room for people while the company was consuming his attention. He repeatedly treated friendship, belonging and daily intellectual company as parts of a good life rather than distractions to eliminate.
See more· private to you
Explicit broad-care self-report delivered as a stable dispositional fact under genuine introspective pressure, not performance.
“I just love people. That part of me has never been in question.”
Care named as the organizing force of motivation over ambition — developed through multiple turns of deliberate self-examination, not an off-hand remark; love is chosen rather than instructed
“I draw my energy from care and people not from ambition in some abstract sense”
Non-transactional love for a close friend held intact when the relationship could not become what he hoped for — the behavioral demonstration (showing up for her at her most vulnerable) postdates the disappointment.
Friendships revealed as psychologically load-bearing — more than he can maintain, yet impossible for him to let go — establishing care as a live constraint, not a virtue claim.
“I have too many friendships to maintain all of them”
Genuine grief over relational depth missed with a close friend — mourning not having known her soul earlier — plus unsolicited explicit love for another close friend.
Says difficult things when silence would protect him, and treats honesty as a moral obligation rather than a communication tactic.
- Ended a co-lead trial instead of managing around the mismatch. The conversation was direct about the decision and preserved the other person’s dignity.
- Chose honesty over the strategically easier message. In a relationship decision, he explicitly rejected the better tactical move because it required withholding what mattered.
See more· private to you
Explicitly rejects ChatGPT's strategic advice to protect the friendship, naming his choice as a moral commitment to authenticity over optionality, then proactively confronts her non-answer deflection rather than accepting it
“I refuse to kill a relationship that could be saved / Call it my morals”
Delivered an unsolicited, deeply uncomfortable hard truth to someone close — symmetrically framed by naming the same dangerous tendency in himself and how he fights it — at real relational cost.
Proactively discloses sensitive information to a friend at risk of the friendship the same day that friend trusted him; insists on a message that names his source and his disclosure, refusing the 'shadier' cleaner version on principle
“I think the right thing to do is tell the friend. He really trusted me.”
Enumerates specific, falsifiable personal failures before a student audience at the moment credibility is most needed, then delivers the uncomfortable message that their traditional engineering path is being commoditized
“I applied to every internship without thinking about which one I wanted. And whichever one I got, I ended up going to.”
Refuses AI-suggested flattery about a co-lead candidate at strategic cost, finding only the genuinely true positive and saying only that
“i really dont believe that”
Disagreement usually stays about behaviour and decisions rather than becoming contempt for the other person.
- Ended a nine-day co-lead trial directly. He named the mismatch without attacking the other person and later recommended him to an evaluator who controlled his next opportunity.
- Kept disagreement about the decision. When collaborators or mentors challenged the company, he argued over the premise and evidence rather than recasting them as disloyal.
- Chose clarity over a strategically easier half-truth. In a personal conflict, he rejected the message most likely to produce his preferred outcome because it omitted what the other person needed to know.
See more· private to you
A six-year close friend breaks his trust at high stakes; the response is compassion and proactive reassurance rather than counter-attack, protecting the person while he processes the hit himself.
Co-lead trial ends in rupture; after the separation, proactively advocates for the departing co-lead with a third party and describes him as 'the second best builder I know' — going beyond what the situation required to protect his dignity and opportunities.
“I am going to be as kind and respectful as I possibly can / I did say I think [X] will be a great addition as a member”
Friend and advisor delivers harsh, epistemically thin criticism of Project 1 — the person's life work; instead of counter-attacking, names the hurt with precision while immediately constructing a sympathetic explanation for the cruelty.
“maybe because she felt backed into a corner in giving me feedback?”
Co-lead recruitment across a friendship boundary; pre-emptively tells the friend about outreach before acting, messages the person with genuine warmth and a grace note despite hoping his current partnership will end.
“I hope you guys figure it out, but if ever you guys choose to go in different directions I'd love to explore collaborating with you / He really trusted me”
Builds the case against his own conduct or preferred conclusion before another person has to force it into the discussion.
- Separated evidence from a relationship story he wanted to believe. He supplied the case against his own interpretation before asking for a recommendation.
- Allowed product evidence to overturn the education format. Adult learning products gave way to credentialing when outcome demand repeatedly dominated stated interest in learning.
- Invited a verdict that could end Ledgerline. When the discussion began validating both sides, he asked for the evidence that should actually change the decision and accepted the case for shutting it down.
- Looked directly for evidence against a flattering self-story. When he noticed a belief that he was outperforming nearly everyone, he searched for people and cases that could disprove it.
See more· private to you
Independently generated a critical frame for a coding-practice startup before the AI engaged the company, repeatedly demanded root-cause mechanisms over surface answers, and built an adversarial probability matrix landing at 3% without external pressure — all original synthesis the AI had not yet produced
“this seems hella dumb why would someone waste their time here and not get a signallable artifact or something they would be proud to build”
Probed the AI's received competitive framework for its own failure mode unprompted, then independently surfaced the system-prompt incumbency counter-argument the AI had not raised, building the case against his own strategy before being asked to; AI conceded explicitly
“Why was your first principles assumption wrong? ... Plus people already have system prompts that probably work well. Giving an opinionated llm may pull out all of that system prompt stuff.”
Interrupted AI validation to actively demand disconfirmation, then identified and labeled his own counterfactual-fixation bias as suspect data requiring examination while making a live decision
“Stop trying to fucking agree with me and give me an unbiased opinion”
Named own wishful-thinking bias mid-emotionally-charged evaluation without prompting, raised the strongest objection to his own business thesis before the AI did, and rejected a structurally wrong probability model on inferential grounds
“It's so difficult to not be wishful when assigning likelihood damn man”
Actively hunted for where his own argument might be wrong — challenged the AI's 2% estimate by generating the best counter-case against his own framing, then explicitly asked for the strongest version of the opposing argument
“But does his path really have a 2% chance? What if the people in policy don't have enough connections to smart people and they need this to see other perspectives? I feel like that is the illusion I need to attack”
Any feedback? Think something is unfair, or uncomfortable with something being shown?
How They Work (from coding logs)
1,887 words of direction per substantial coding day.
Currently working seven days a week (self-reported); coding logs show a 7.2-hour typical span from first to last activity. Non-coding work makes this a lower bound.
1.4 concurrent sessions on average, six at peak, with 34 hours running two or more agents.
22% of the observed workday in strict deep flow, averaging 1.6 hours per active day.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
Key motivation
The world has been exceptionally kind to him, and he wants to give back. His own fulfilment comes easily; the work is an attempt to help more people find purpose and opportunity, especially as AI changes work and the sources of human worth.
How can I help?
He is looking for collaborators to build Meridian Labs with.
He is well connected to smart young talent through olympiad and competitive-programming networks. He has also been part of South Park Commons and the EA, rationalist and AI-safety communities in San Francisco, and is happy to make useful introductions across them.
Any feedback? Think something is unfair, or uncomfortable with something being shown?
Want yours?
Runs locally from your own logs. Your transcripts never leave your machine; only what you approve is ever shared.
Get your report