entertain-vestment + family office

Z / Runner 🏃🏻‍♂️

🏃🏻‍♂️ coherence: I think of it as an "internal story" or self-identity. It doesn't change every day. If I say I like tofu and likes to run and tend to invest mostly in tech, the day-to-day is the same. This is something, we as founders are guilty of cheating and not following. If we want to build a lego zoo, we at times will wake up and turn it into a spaceship then a car and then forget where all the lego pieces are (b/c we get motivated by the shinny new ideas or market headlines). So, in this example coherence = remembering what we're building and keeping it going. In another way, coherence is the continuity of intent. This has been the hardest thing to solve, we all have ADHD. Why this matter is b/c context window & memory (RAG included) are all machine side optimization, no AI tools today is looking out for the human-side coherence.

• 🏃🏻‍♂️ actionable = get users to state, re-state, re-state their top most important goals (intents)... we need to win this battle. Even as much as I use GPT/Gemini/Manus, I don't state my intent, I only prompt immediate tasks. (90% certainty, we can win).

• 🏃🏻‍♂️ 100%, but the cool & emerging trend is that NEWS are being repackaged as opt-eds, reviews, research opinion pieces. Historically, news were extremely expensive to product (AP, CNN, BBC, sending 100,000 reports to war zones, local HQs), $1 in vs $1.01 out. Now days, Ben, Lex, JRE, all just read the raw news and repackage. I Believe this is the "Bespoke" content we can deliver. (To be clear, not mixing up generative content w/ bespoke content recommendation).

• 🏃🏻‍♂️ Someone like li is looking for a daily dose of "buffet level insights".

• 🏃🏻‍♂️ do you think FX health rev model is worthwhile for us to replicate? (i'm hinting, they effectively put a wrapper on top of QD [quest diagnostics] and had a good mouth piece (Mark) to talk about it). Obv/obv, i am over-simplifying their success but the logic is the same. They (FX health) told me something that I value and want to know, but didn't really deliver any actionable insights (beyond the superficial wkipeida level analysis) or manage intervention (direct capital deployment).

• 🏃🏻‍♂️ i still monitor, but now all my cousins and sister's co-workers are jumping on the bandwagon, it is a growing faster and faster. Again, the danger of sample size of 1. We are always early, but we rarely capitalize on the trend b/c we have ADHD / coherence problem… the world is so interesting and so rich… for us.

• 🏃🏻‍♂️ i think use cases #4 & #5 will have billions of rev potential, but the struggle is how do we get there?

• 🏃🏻‍♂️ i rather win 100,000 DAU w/ $1M ARR in the next 180 days vs. 1,000 users w/ $1M wealth commitment. Bc the EVA tilts the former; we have control of the former, the latter has more to do w/ luck.

• 🏃🏻‍♂️ For example, if our thesis is that our system enables you to buy into latest middle-east rare earth enrichment companies or gain access to Unitree IPO, then someone like RJ, Leo, Hohchung, Kai… are likely to jump on it b/c it is easy money and the sharp is obvious. But telling them information w/o the immediate actionable path won't be enough.

• 🏃🏻‍♂️ only if we can re-create what made FTX did (kimchi premium). If our thesis is we have alternative means to buy into markets that are not available in the US market, then YES. it is a huge win. That's why Sam did in early days and I saw how Hochung was flying back to Korea. But are we offering this? For example, Unitree IPO is coming… how do you think our system can help? [Stephen Tse](mailto:tse.stephen@gmail.com)--- this is probably the biggest thought experiment… (I put this as the same category of investment as Buffet's BYD (20x return w/ $230M), or Masa's alibaba ($20M to $170B). We are talking about structural gain.

• 🏃🏻‍♂️ i reached the same conclusion as you, hence, i put wealth management as #4.

• 🏃🏻‍♂️ actionable = get users to review continuous activities 3x per week, are they happy w/ the results we are generating. If I want to learn more about XYZ, are the results comparable to "single-shoot" prompting? (50% confidence level, TBD on the actual implementation).

• 🏃🏻‍♂️ do you define result as in "gain vs. loss"?

• 🏃🏻‍♂️ if true (Chamath showed a chart that even warren had below market return after Reg D was made illegal, meaning, warren had "insider" info that were not available to us. Specifically, from 1981-2000 = 24% return, but post Reg D illegal, his return flattened to 10% (same as the index). That's why i think wealth management is a weak entry point for us… b/c we don't have any info that's not part of the public domain.

• 🏃🏻‍♂️ YES, let's double click on this, if we can use the same playbook Manus did w/ the token arbitrage, getting ppl like me or aaron to overpay for usage, then #4 is doable. Scaling the random walk by paying for access…

• 🏃🏻‍♂️ we have these thing going so far 1). We have a strong hypothesis that getting 60s voice inputs per day is possible (voice + incentives & packaging it as a daily ritual, at least for the first 100,000 users) 2). Continuous task management in the background is rare (no strong competitors).

• 🏃🏻‍♂️ sample size of 1 (so always w/ grain of salt), but this past 2 weeks have been bombarded w/ friends & family chat about going on ski trips (would be nice, not having to monitor bookings & snow condition, ski pass, rental gears)... it just basic, boring tasks that should be a background activity.

• 🏃🏻‍♂️ people don't use AI for this type of workflow b/c the white collar workers use it for productivity and family use it for the one-off research. It is a $1 Billion problem, too small for most big tech companies, but too big for startups.

🏃🏻‍♂️ win big on DAU (yes, targeting the king). OpenAI stats: per GDM internal comments & openAI (2025 year end summary); total 800-900M. WAU = 10%; 5% are paying users. 80% of users sent less than 1,000 in 2025 [🏀 39 per day!]. for context, i sent about 14,000 in 2025, and I still don't think I'm using enough.

• 🏃🏻‍♂️ hence, if we can get 1,000 users to have 90% DAU, then that's a strong start for us.

• 🏃🏻‍♂️ YES, the tech side of using nanoclaw gives us that, but equally important is the daily loop. I shared the pulse feature not as a news (or look, new sexy squirrel), i know it is the problem that big 5 model companies are trying to solve. Google is winning the DAU battle b/c they just shamelessly plug gemini into every product and Meta is betting on their top 3 social platforms (whatsapp, IG, FB). If we can deliver the perfect DAU thesis, I think a $10B outcome is possible. That to me is the simple and almost too good to be true idea.

• 🏃🏻‍♂️ doubtful, b/c the trust level is sooooo high. When I present the use-case, people who have the means to invest, generally think very highly of themselves (smart, motivated, and high achievers). They bet on themselves vs. following AI advice. Put it bluntly, the best SWEs are the most resistant to use full AI vibe coding b/c they think they can do better. Do you see why I struggle w/ gradient #4 (wealth) b/c the time to earn trust is too expensive.

• 🏃🏻‍♂️ i think we can just bank on "manus" the paid wall content and substack to feed into the system…

• 🏃🏻‍♂️ Most ppl pay for morgan b/c of the "trusted brand" not b/c of the quality of research. That's why someone like Li is the perfect patient 0.

• 🏃🏻‍♂️ is this "pro humanity" :-)

Stephen / Basketball 🏀

• 🏀 i'm (philosophically) aligned – but these points are not very actionable?

• 🏀 but the paradox of getting to the same goal (maximize wealth) vs user daily restating the obvious? so, news and events must be present in the loop?

• 🏀 true, hence the explosion of podcasts / substacks / twitter threads. but, without the 10% performance fees of $100k wealth / family office deposits, our revenue won't fly – vs sponsor / patreon / ads models of the human (news) anchors / hosts.

• 🏀 exactly, personalized "entertain-vestment".

• 🏀 exactly, but ditto for the 100 ai stocks or pre-ipo or foreign investment. been off longevity and fitness quests for months now.

• 🏀 not the deficiency of attention but the paradox of choice.

• 🏀 possible also for #4 (buying a house), but pre-commit deposit of $500k – much like our $100k for #3 (wealth management). but much less appealing, as 10x less frequently and 5x more deposits. much worse for #5.

• 🏀 no acronyms! "expected value analysis". if not $100k, i don't know / use a (new) product that i'd pay $100 per month.

• 🏀 again, information daily but $100k pre-commit. pre-ipo and off-shore much later – first: rebalance weekly of 10 diversified ai public stocks.

• 🏀 enticing gains, but not about ai agents (systems) at all.

• 🏀 which #4? don't use shorthand without defining it or i definitely thinking of the same reference.

• 🏀 of all the 5 use cases you shared on "Gradient Outcome for Ghost" (and probably all we ever discussed): if combining "daily usage" + "changing result", then only possible with ever-changing stock performance.

• 🏀 actually, more than the portfolio profits, it's the gradient updates on the investment thesis / world view that may be like warren buffet's daily iteration.

• 🏀 i analyzed and saw the flat line, too – across all angels, funds.. even (quant) models. hence, it's more of an entertainment / art play (more than singularly utility and objectively maximizing) of wealth: each person engages daily and decides his own journey of "walking down the random path" in the (stock) market.

• 🏀 use case #4 as in buying a house?

• 🏀 YES. fully aligned with these two success criteria, but no overlap of use cases between us – satisfying both criteria and our forever-ongoing needs.

• 🏀 but not every hard problems or hair on fire is easily monentizable / immediately market fit.

• 🏀 exactly, must ignore all 99.9% of them for now.

• 🏀 YES. news as prompt + voice as journal are our obvious but proven points. let's take it home! are you open to shorter-term investment, hence monthly / quarterly trading + balancing of a portfolio? still the strongest use case of end-to-end.

• 🏀 ha, paradoxes and dilemmas, indeed and everywhere! hence, instead of agents placing the trades or even recommending the advice, we position as "wealth pulse": ..

🏀 #full-automony #wealth-agent

• morgan stanley's private wealth management + family office
• astor (yc s2025, $39/month)
• 🏀 plus 100k youtube channels + 1m twitter..
• 🏀 some (me + jacy) will do everyting to avoid contact with human agents, the quintessential closed-source and byzantine.
• 🏀 per sanity!

🏀 z, define "coherence" in the common terminology of agents and in the way i can fully align? is it simply context and memory management?

Unattributed

The four definition-quote bullets at the end of Day 90 (sub-bullets under Stephen's "z, define coherence" question) carry no markers:

• "Coherence becomes computational, not cognitive."
• "The hard part isn't generating output — it's stitching tasks and context together in my head. I call this the coherence problem."
• "Most AI today is trained to respond to every prompt with maximum verbosity. A ghost is for coherence."
• "UNIX solved this coherence problem decades ago by separating the transient user from the persistent system."