"Illness is normal", they said. "Let's learn to live with it."
"Death is normal", they said. "Let's… er…well, you know what we mean."
"Pessimism is normal."
Perhaps, but it shouldn't be. Because it is false.
Not knowing something is never in itself a reason not to do something. It's the things we know about the environment or premise of that 'thing which we don't know' that actually influence the decision. Think opening a surprise gift from friends vs cutting a random wire inside a live bomb.
I have no view of the underlying project, the reasoning comes with a flawed premise.
Market feedback on a decision can take years. Patterns only help when the thing creating them does not change.
The only reliable way to learn is to create a system, an internal feedback loop.
It is most useful when it:
- Maximises the error corrected per iteration
- Minimises the time per iteration
This maximises the rate at which reliable knowledge is created.
Another episode of 'The Bilbel Capital
Podcast'
This month's topic: How does an investor search globally for extreme mispricing?
You may watch the podcast here:
https://t.co/wTlBvokgiN
I hope you enjoy it, and find it useful in some way. I would appreciate your thoughts in the comments.
@bilbelcapital
I think that the greatest competitive advantage on a human-vs.-human level is a person's relative 'want' to invest effort into something.
There are many problems where a person's natural response is to defer or avoid them. Over a long period of time, I think the most common and consequential reason why one person will perform better than another at a specific problem set or field is a greater natural tendency to solve that problem set.
Many feel threatened, and hence experience pain, when trying to explain why different people perform better or worse at different things, and I think it's just this simple. I also think that this is the reason why I have found a person's free-time activities to be so insightful in understanding what they're good at.
Finally, this is something I have noticed much more ever since AI has become tremendously useful. There are many ideas that others and I have observed where we thought something was obviously a great idea, yet we did not actually bring it into reality ourselves, even though the amount of effort required was very little (especially due to AI). So the 'want' to do something really is a competitive advantage.
@TradeIdeasFlow VM browser using a residential IP. Muse's browser till now seems to work. The raw solution is just having an old pc run a browser inside a virtual machine or something and having it be connected to your home Wi-Fi.
LinkedIn's value has increased a lot since AI agents became a thing. Whenever I identify a group of people that I'd like to connect with, my AI agent logs into my LinkedIn and checks for secondary or direct connections who can introduce me. Given that LinkedIn serves as a contact book explicitly for more formal and business-oriented relationships, it's been quite valuable for this task. This makes me want to use LinkedIn more to connect with people in the future. A good example of a service that becomes more valuable the more data people give it, creating hard to replicate 'inertia'.
Meta's Muse is some pretty significant stuff for consumers. As some may know I and a few friends of mine have been pretty ahead in terms of experimenting with AI generally, and things like OpenClaw or Hermes.
At least 90% of what I've been able to do with the set-up that I've put together you can easily do with Muse and with at least 10x better UX.
On the economics side, it's an interesting product. My continually growing context base has acted very sticky. I move mountains to find away to integrate a digital workflow/information source into my agent. This is pretty powerful behaviour. I'm not sure who wins the 'race', but a product like Muse's finally drives the point that many should have picked up earlier, which is that a lot of the value does not live in the capability of the models (beyond a certain degree), but in the integrations and context base that you can connect the model to.
On the 'strategic' side this move also makes incredible sense for Meta. The number of adverts that I consume have fallen off a cliff since I've began interacting with the same digital services through my agent instead of doing so directly. Rarely does an advert understand my intent as much as my agent. There are still a few problems I have in my theories/explanations for what the value of Muse is. But I am unable to exclude that it may be one of the most valuable products ever.
Another interesting thought is what will happen to these very important 'digital services'. Substack is a great example. I never open the app. It's always my agent that jumps on there and looks for things according to my intent. If Substack closed tomorrow, I would be significantly worse off. So there is still strong value there (and hence monetization opportunity). The current idea I have is that agents will interact with these services through a 'pay per result' model. So, you still use data aggregators, marketplaces, etc... because their value has nothing to do with their user interface, it's just that some of their current monetization might, and hence that piece will have to change. Amazon is another example, today I do not imagine going on there and manually browsing myself. Nonetheless I would be willing to pay for the wide variety of options that Amazon's marketplace brings about.
I think anyone that wishes to preserve agency over their future must be fundamentally understanding this technology, comparing it to the internet is a no brainer (just as comparable even in terms of the amount of meaningless extrapolations that are made especially with regards to the economics of different ventures).
Not sure if this works to reject 'context' as a reason. It's like rejecting 'content flywheel' because YouTube has things that are not relevant to you. I think agents have a great flywheel in this regard, even more so if they can use others' solutions to solve your problems.
What I've been wondering is whether or not the device (or rather OS) makers possess a significant advantage because of their ability to create seamless integrations with hardware capabilities (such as microphones, Bluetooth, cameras, and general phone apps, such as photos, files, etc.).
Agreed with this:
"I think that is part of the opportunity so many people see the decline in paying users and click run write off
$BMBL "
and I also think this is correct:
"[referring to the significant decline over 2025] this is controlled and part of a quality over quantity reset"
The piece that is unclear to me is the following:
"Also ask yourself how much the stock would re rate if say renewed marketing spend, new user experience and a focus on quality grew paying subscribers one Q. To me it’s asymmetric"
I'm going to assume that you don't mean that you just want to bet on one good-looking quarter here.
I don't think marketing spend is what makes the difference in getting more users, Hinge has 'demonstrated' this to be the case, they are not 'winning' because of marketing spend.
A new user experience makes sense, that is in fact the 'success factor' that worked for Hinge. I was still unable to understand the nature of this 'change'. Change in itself does not create value (I was also concerned that the speed at which users may leave the app would erode the platform's value as a launching pad for a new product). That being said, Whitney did return because she thought that there was a pivotal opportunity for 'change', and she chose Bumble specifically to pursue that change. She has more than enough money and social capital to pursue an opportunity in 'dating' outside of Bumble if she thought that Bumble did not bring some value to the table for that 'opportunity for change' (possibly this is the 'most favourable' arguement I have for impactful change coming about here).
Would love your thoughts.
Remix theory posits that all creation and knowledge stem merely from the recombination of existing elements. As an epistemological framework, this is fundamentally flawed. If all knowledge is just a combination of prior knowledge, one must eventually ask: where does the base knowledge come from? Recombination requires starting variables, and these premises serve an axiomatic role within a given deduction. Without a mechanism to generate these foundational premises, they remain unaccounted for.
Conjecture and criticism are the mechanisms that actually create this base knowledge. However, I previously assumed this meant that all problems must be solved through conjecture and criticism. That is incorrect. What we have seen AI do, even before tackling complex mathematical proofs, is successfully solve problems for years through pure recombination. A new problem does not inherently require the creation of new explanatory knowledge. Many problems can be solved simply by recombining existing knowledge within an established framework. I vastly underestimated the number of unsolved problems whose solutions are directly derivable and therefore subject to recombination.
Therefore, human effort is not best used on combinatorial problem solving. While it is entertaining to watch the best chess players navigate finite rule sets, the human role will shift as AI continues to improve. Our focus will be on defining and altering the provisional "axioms" by providing the hard to vary conjectures that set the boundaries for AI computation. If the computational cost of recombining these premises keeps trending to zero, the leverage we gain from the genuinely new knowledge we create increases exponentially. This is truly fascinating for the future of knowledge creation.
@JoeValue I think the solution should be to just look for an explanation, within the thesis, for why the 'Cheap vs peers' is relevant. It would just be a bad thesis if someone left out that the peers were all overvalued.