1,025 Matching Annotations
  1. Jul 2023
    1. TikTok is an interesting new player in social media because its default feed, For You, relies on a machine learning algorithm to determine what each user sees; the feed of content from by creators you follow, in contrast, is hidden one pane over. If you are new to TikTok and have just uploaded a great video, the selection algorithm promises to distribute your post much more quickly than if you were on sharing it on a network that relies on the size of your following, which most people have to build up over a long period of time. Conversely, if you come up with one great video but the rest of your work is mediocre, you can't count on continued distribution on TikTok since your followers live mostly in a feed driven by the TikTok algorithm, not their follow graph.

      Glasp needs to prioritize the algorithmic feed over the social feed so that insightful content can more easily be distributed to user for whom the content is most relevant.

    1. Users are encouraged to follow and friend each other to assemble their own graph one connection at a time. The challenge with that is that it’s almost always a really slow build, and you have to provide some reason for people to hang around and build that graph, often encapsulated by the aphorism “come for the tool, stay for the network.” Today, it’s not as easy to build the “tool” part when so much of that landscape has already been mined and when scaled networks have learned to copy any tool achieving any level of traction.

      A few points to address here. Assembling a social graph is tough. User's shouldn't have to. While the user is waiting to build out a critical mass people to follow the need some single user utility to entice them to stay. It is fairly easy to build the tool part in areas that haven't been as thoroughly investigated. For one, there are several applications that alloy for the creation of annotations but few if any provide any functions to review and make later use of those annotations.

    2. Within a larger social network, even subcultures need some minimum viable scale, and though Bytedance paid dearly to fill the top of the funnel, its algorithm eventually helped assemble many subcultures surpassing that minimum viable scale.

      Is there a lower cost alternative to filling the top of the funnel. Perhaps that is where cryptocurrencies come in. You could create a PoW protocol that issues a finite amount of tokens in the early days of the network to users that contribute most.

    3. Prior to TikTok, I would’ve said YouTube had the strongest exploit algorithm in video,The exploit versus explore conundrum is sort of a classic of algorithmic design, usually mentioned in relation to the multi-armed bandit problem. For the purposes of this discussion, think of it simply as the problem of choosing which videos to show you. An exploit algorithm will give you more of what you like, while an explore algorithm tries to broaden your exposure to more than just what you’ve shown you like. YouTube is often described as an exploit algorithm because it tends to really push more of what you like, and then before you know it, you’re looking at some alt-right video that’s trying to redpill you. but in comparison to TikTok, YouTube’s algorithm feels primitive (the top creators on YouTube have long ago figured out how to game YouTube’s algorithm’s heavy dependence on click-through rates and watch time, one reason so many YouTube videos are lengthening over time, much to my dismay).

      Always need to consider how the algorithm might be gamed and contribute to changes in the content on the platform. This example contrasts with Spotify's algorithm which registers that the user has listened to a song after 30 seconds causing songs to shorten over time (not a new phenomenon. this also happened when music moved to the radio)

    1. The Discover Page features hashtags. By the very act of featuring a hashtag, they signal to creators that if they create using that hashtag, they will get the distribution boost of that hashtag being featured on the Discover Page. Which raises the age-old conundrum, which came first, the Discover Page hashtag placement, or the hashtag's trending? The answer is yes. It's circular, an ouroboros of virality.

      Need to signal to curators which topics are highly desired.

    2. Now imagine trying to make a TikTok from scratch if you're older than, say, 19. The creative bar is high, you don't know how to dance, you're not up on the latest memes or popular music. Even if you're a teen, it's not easy to come up with a 0 to 1 TikTok.↳ But the beauty of TikTok's FYP algorithm and the Discover page is that you don't have to create a TikTok from scratch. The vast majority of TikToks are riffs on memes and trends that other users originate. It's no shame to be a 1 to n TikToker. Many on the platform achieve their first viral hit riffing on an existing meme.

      I think this is why highlights are a good basis for social network because there is no creative bar. You can rely on recognition/discernment which everyone has some capacity for rather than creation which most of the population is inept.

    3. How many truly original ideas are there in Silicon Valley? Very few. Most have been tried umpteenth times in the past. Much of finding product-market fit in tech is context and timing. And people always underestimate the market side of product-market fit. When something fails, people tend to blame the product, but we should blame the market more often. The pull of the market is usually as important, if not more so, than the push from a product.↳ One day, the conditions are finally right, and an idea that has failed ten times before suddenly breaks out. Sometimes it's a tweak in execution, maybe it's an advance in complementary or enabling technology, sometimes it's a cultural shift.

      Is there away to effect cultural shifts?

    4. Until TikTok came along, there wasn't an easy way to do reaction videos to other videos and have them make sense unless the original video had so much distribution that it was common knowledge.

      What is the equivalent of a reaction video for annotations? Just he replies?

    1. There’s a reason swipe right and swipe left have become shorthand slang for signaling approval and disapproval, generally. Tinder came up with what feels like a design primitive on a touchscreen UI for binary voting.

      Tinder is a great example of an alternative implementation of the paginated feed.

    2. Videos on TikTok are so short that even in a brief session, TikTok can gather a lot of feedback on your tastes.

      Short form allows for greater quantity of training data.

    3. Some forms of friction are good.

      Need to consider what other forms of friction are beneficial to the broader system/algorithm.

    4. TikTok doesn’t have an explicit downvote button, but by serving you just one video at a time, they can infer your lack of interest in any single video based on whether you churn out of that video quickly A quick swipe up before a video has completed is like swiping left on Tinder. The best TikTokers have an intuitive sense of the narrative pace that is appropriate for that platform. How long can you drag out the punching or payoff without losing the viewer, how do you have a set up that keeps the user involved. Using a music cue that has already been co-opted into a meme helps because the bass drop or musical payoff foreshadows when the punchline of the video will drop; a viewer knows how much longer before they reach the payoff. Also viewers may stick around just for the pleasure of hearing that musical resolution. and by which positive actions you don’t take.

      Glasp or any other social annotation platform could use scrolling past a top highlight and its associated document as a negative signal and saving the content for later or saving the highlight to review later as a positive signal. The interaction might incorporate a tinder like swipe left and right.

    5. Even on a story a user does engage with, judging sentiment is a challenge. Most apps only have positive feedback mechanisms, most typically some form of a like button. Since apps like Facebook, Instagram, and Twitter are built around social graphs, it’s obvious why they might opt not to offer dislike buttons.

      The major weakness of incumbents that provide content discovery through a feed is that they rely on social interaction that prevent the introduction of negative feedback mechanisms like downvoting or rating that would provide better training data. Even youtube has removed the like button. One way to get around this issue is to not show the rating of content until it has reached some positive threshold. You could also refrain from showing user profile images within the feed.

    6. On the other hand, maybe you wouldn’t mind reading one tweet at a time if they were better targeted, and maybe they would be better targeted if Twitter knew more about which types of tweets really interest you. And maybe Twitter would know more about what really interested you if you had to give explicit and implicit positive or negative signals on every tweet.

      In this way, the infinite scroll design belies the fact that the majority of the content is not relevant and therefore worth attending to. You need a way to quickly move past all the shitty posts and ads.

    7. The default UI of our largest social networks today is the infinite vertically scrolling feed (I could have easily used a screenshot of Facebook above, for example). Instead of serving you one story at a time, these apps display multiple items on screen at once. As you scroll up and past many stories, the algorithm can’t “see” which story your eyes rest on. Even if it could, if the user doesn’t press any of the feedback buttons like the Like button, is their sentiment towards that story positive or negative? The signal of user sentiment isn’t clean.

      Paginated feed creates greater legibility for the algorithm by limiting the scope of attention. In a sense the attention of the user and application are shared/aligned.

    8. The algorithm can also see what TikTok already knows about you.

      He adds this comment almost as an aside but this seems a very important area to explore.

    9. Vision AI also does a pass on the video, and to the extent it can, contributes what it sees. Some of TikTok’s camera filters are designed to track human faces or hands or gestures so vision AI is often invoked even earlier, at the point of creation.

      LLMs might also be used in the classification of text based content. The classification should be confined to terms within a controlled vocabulary.

    10. But let’s step back even earlier, before you’re even watching the video, and understand how the TikTok algorithm “sees” the video itself. Before the video is even sent down to your phone by the FYP algorithm, some human on TikTok’s operations team has already watched the video and added lots of relevant tags or labels.

      This function of the operation team can be outsourced to users either the entire base of contributors or some subset of lead users that are designated as curators.

    11. Everything you do from the moment the video begins playing is signal as to your sentiment towards that video.

      Paginated feed with single item allows for the generation of better data both about the content quality and user preferences. There is no ambiguity as to what the user is attending to or ignoring. The interaction model makes user attention explicit.

    12. TikTok fascinates me because it is an example of a modern app whose design, whether by accident or, uhh, design, is optimized to feed its algorithm as much useful signal as possible. It is an exemplar of what I call algorithm-friendly design.

      User friendly (ease of use) => algorithm friendly (frictions/work)

    13. But what if the key to serving your users best depends in large part upon training a machine learning algorithm? What if that ML algorithm needs a massive training dataset? In an age when machine learning is in its ascendancy, this is increasingly a critical design objective.

      You need to recruit your users to provide that training data, first acquiring/creating it and then labeling.

    14. This, then, is the magic of the design of TikTok: it is a closed loop of feedback which inspires and enables the creation and viewing of videos on which its algorithm can be trained. For its algorithm to become as effective as it has, TikTok became its own source of training data.

      The same process can be applied to other interest based content distribution networks (e.g. Glasp and other social annotation platforms with a feed).

  2. Jun 2023
    1. There is a format for recording information based around connections called a concept map. Concept maps are spatial representations of knowledge, in which a concept is represented as a circle (node) and a connection/statement is represented as a line connecting two circles (edge). But concept mapping is limited in its current forms. Making a concept map is messy and often when handwritten. It is very difficult to have enough foresight in making a concept map as to prevent lines (representing edges) from crossing each other all over the page. Moreover, though concept maps have advantages over bullet points and other linear note-taking techniques in their ability to store and model nonlinear knowledge, existing software that is meant for creating concept maps is meant more for presentational purposes, e.g., creating a diagram that might be included in a report, than for recording notes in real time and understanding, organizing, or manipulating notes.

      Concept map software could definitely be improved. It would be nice if a index of concepts could be derived from ones notes or highlights and then the user could select index concepts to drag into a canvas and begin to map out the connections. Selecting a concept could then surface all related notes and highlights or other concepts that are tangential.

  3. Sep 2022
    1. Tags are just backlinks to pages that don’t exist.

      Tags are bit more than just backlinks.

  4. Jul 2022
    1. “In simple terms, this means that a 50 basis point increase will be among the choices on the table when we next meet,” said BOE Gov. Andrew Bailey in a speech Tuesday. “50 basis points is not locked in, and anyone who predicts that is doing so based on their own view.”Mr. Bailey also said the BOE would lay out a plan for selling some of the government bonds it bought under a series of stimulus programs known as quantitative easing. He said sales could start as early as September, and total between £50 billion and £100 billion in the first year, equivalent to $60 billion and $120 billion.
    2. Economists expect the annual rate of consumer-price inflation to increase further. The Bank of England has said it should top out at around 11% in the final months of the year. The U.K. sets a ceiling on home energy prices twice yearly, and the next adjustment is due in October, when a further rise of 50% is expected.“The inflation outlook remains grim,” said Sanjay Raja, an economist at Deutsche Bank. “Our updated projections now show CPI peaking at 11.3%.”
    1. Potential government measures include taking tariffs on imports off, lowering pharmaceutical prices, improving energy policies and lowering the budget deficit, he said.“There’s a lot we can do to contain or control inflation,” he said. “But if we continue with the kind of ostrich policies we had in 2021, there’s going to be much, much more pain later.”
    1. Investors, though, have typically been cautious about anticipating such a pivot. In the mid-1990s, when the Fed last tightened policy at something close to its current pace, the central bank raised rates by 3 percentage points in total. But investors were ready for it to go much further than that, said Mr. Caron, who was trading at the time.The Fed, in that episode, raised rates for the last time in February 1995 and then lowered them in July. But it wasn’t until May that short-term Treasury yields suggested investors were preparing for a cut.
    1. In separate research, strategists at Morgan Stanley counter-intuitively found that US bond yields tended to fall during the previous episode of quantitative tightening as prices rose — a conclusion they said was “not a typo” and attributed in part to economic growth patterns. “This is clearly a complicated story, where balance sheet change is just one of many factors which impact cross-asset performance,” they said, adding that analysis is “hamstrung by a severe shortage of data”.
    2. Still, Fed officials and researchers have offered up rough estimates. According to a June study published by the Fed, a $2.5tn drawdown in the balance sheet over the next few years would be roughly in line with just over a half-point increase in the benchmark US policy rate — a range in keeping with Fed vice-chair Lael Brainard’s assessment that the central bank’s plans amount to “two or three additional rate hikes”.
    3. The ramp-up in scale and range of QE in 2020 compared to 2008 was “completely insane”, says Tatjana Puhan, deputy chief investment officer at French asset management group Tobam.Over the course of two years, the Fed snapped up some $3.3tn in US government bonds and $1.3tn in agency mortgage-backed securities. As of March, that left the US central bank owning a quarter of all outstanding Treasury debt and a third of agency MBS.The ECB and BoE each own just shy of 40 per cent of their government bonds, while the Bank of Japan, which is unique in having no intention of stopping its purchases, already owns nearly half of Tokyo’s outstanding government debt.
    1. The Affordable New York tax provision, commonly known as 421a, offered a property tax exemption for housing projects in New York City that include a percentage of units earmarked for lower-income renters. Nearly 70% of rental housing built over the past decade used the tax abatement, according to New York University’s Furman Center. But the 51-year-old program expired in June when state lawmakers ended their session without renewing or replacing it. That has left plenty of New York City developers in a bind, scrambling to complete rental projects they started before the provision lapsed or switching gears and building other types of properties.
    1. The see-and-point principle states that users interact with the computer by pointing at the objects they can see on the screen. It's as if we have thrown away a million years of evolution, lost our facility with expressive language, and been reduced to pointing at objects in the immediate environment. Mouse buttons and modifier keys give us a vocabulary equivalent to a few different grunts. We have lost all the power of language, and can no longer talk about objects that are not immediately visible (all files more than one week old), objects that don't exist yet (future messages from my boss), or unknown objects (any guides to restaurants in Boston).
    2. The critical research question is, "How can we capture many of the advantages of natural language input without having to solve the "AI-Complete" problem of natural language understanding?" Command-line interfaces have some of the advantages of language, such as the large number of commands always available to the user and the rich syntactic structures that can be used to form complex commands. But command-line interfaces have two major problems. First, although the user can type anything, the computer can understand only a limited number of commands and there is no easy way for the user to discover which commands will be understood. Second, the command-line interface is very rigid and cannot tolerate synonyms, misspellings, or imperfect grammar. We believe that both these deficiencies can be dealt with through a process of negotiation.
    3. The basic principles of the Anti-Mac interface are: The central role of language A richer internal representation of objects A more expressive interface Expert users Shared control
    4. If the user is required to be in control of all the details of an action, then the user needs detailed feedback. But if a sequence of activities can be delegated to an agent or encapsulated in a script, then there is no longer a need for detailed and continuous feedback. The user doesn't have to be bothered unless the system encounters a problem that it cannot handle.
    5. The problem with WYSIWYG is that it is usually equivalent to WYSIATI (What You See Is All There Is). A document has a rich semantic structure that is often poorly captured by its appearance on a screen or printed page. For example, a word may be printed in italic font for emphasis, as part of a book title, or as part of a quotation, but the specific meaning is lost if it is represented only by the fact that the characters are italicized. A WYSIWYG document shows only the final printed representation; it does not capture the user's intentions.
    6. The Anti-Mac principles outlined here are optimized for the category of users and data that we believe will be dominant in the future: people with extensive computer experience who want to manipulate huge numbers of complex information objects while being connected to a network shared by immense numbers of other users and computers. These new user interface principles also reinforce each other, just as the Mac principles did. The richer internal representation of objects naturally leads to a more expressive external representation and to increased possibilities for using language to refer to objects in sophisticated ways. Expert users will be more capable of expressing themselves to the computer using a language-based interface and will feel more comfortable with shared control of the interface because they will be capable of understanding what is happening, and the expressive interface will lead to better explanations.
    7. WYSIWYG assumes there is only one useful representation of the information: that of the final printed report. Although we are not arguing against a print preview function, even when the goal is to produce a printed document, it may be useful to have a different representation when preparing the document. For example, we may want to see formatting symbols or margin outlines, or it may be useful to see index terms assembled in the margin while we are composing.
    1. The One Hypertext done right, I think, is a proto-Metaverse. If we build a Web that feels like a place designed for life, the services and sites on the Web can stretch into the horizons of this hypertext Metaverse. We could meet people on the Web, find entertainment, connect with partners, and build a living online. But aren’t we already building pieces of this Metaverse today? To propel this trend into the Metaverse of 90’s science fiction, I think we need someone thinking more intentionally about building a Web experience from the perspective of an architect, rather than a user interface designer. The Web is increasingly a home more than it is a tool, and our design needs should change to reflect it. I think this work of building a Metaverse requires something different from the iterative engineering the Web industry has engaged in for the last decade. We need more first-principles design, more imagination in what kinds of places the hypertext Metaverse could contain in the future, the way an architect might dream of their next stadium or skyscraper.
    2. I think we’ve barely begun to tap the potential of designing the Web as as built environment and a work of architecture around our digital living spaces. When we design the One Hypertext for people, not just for information, the Web becomes something more than a resource. It becomes the Metaverse.
    3. Today, we find a different set of metaphors for the Web. We don’t go on the Internet as much, or log in and log out anymore. Instead, we’re online or offline, connected or disconnected. “Online” is a state of being, not a place to be. (When was the last time you closed your web browser?) We spend most of our time on the Web not browsing or exploring, but subjecting ourselves to the flow of information that the Internet now levies at our attention.
    4. the Web has lost a sense of place that it used to have. Today’s Web is a condition of being – being online, being connected, being subject to the flow of the feeds. A sense of place is what allows humans to gather and meet and have conversations, and with fewer places on the Web feeling like real spaces we can enjoy, I think we find our conversations pushed out into the few places that retain that metaphor of place.
    1. Twitter’s main user interface, the algorithmic timeline, is bad for using Twitter as a learning tool. It feels like sticking a straw into a firehouse and hoping you’ll suck out enough interesting insights to be worth the effort.
    2. For me, Twitter serves two purposes. First, it’s a learning tool. There are lots of smart folks talking to each other and sharing what they’re thinking about on Twitter from software to economics to writing, and I can find on Twitter opinions or perspectives I can’t find on blogs or books. Second, it’s a place for me to share whatever I’m working on on my blog or on my side projects with my audience. Lucerne is designed around these two primary workflows: learning and sharing.
    3. With Lucerne, I can search for interesting conversations happening on Twitter by experimenting with these filters. When any filter seems particular useful, I can save it to check again later, by adding it to the left sidebar with a name. As I use the app, I end up curating an ever-changing personalized collection of these channels in my sidebar that provide multiple different views onto the firehose of Twitter.
    4. The biggest change I’ve noticed from using the client is that it turns Twitter from a consumption experience into an exploratory experience.
    5. Lucerne isn’t meant to be a Twitter replacement. Twitter’s web app is still great for writing and following threads, for example, and I don’t want to have to re-create something that’s already fine for my use. But for my two main workflows of learning and tracking my progress on Twitter, Lucerne works better for me.
    1. What if curating things for your audience was radically easier? A company like Pocket or Instapaper, who already have a substantial user base clipping and collecting reading material they like, may be in a good place to launch an experiment in this space. But those most avid readers and curators may also already have started newsletters, and may not want something lower maintenance. Another minimal viable product may be a kind of a filter for Twitter that aggregates just the links and good reads that people you follow have shared. People are already curating and sharing – a good first step may be to simply slide into where people are already curating, and make the processes of curation and discover easier and higher-reach.
    1. If we were to build a medium for better thinking on top of the web browser, it’s reckless to expect the average user to manually connect, organize, and annotate the information they come across. Just as the early World Wide Web started out manually-curated and eventually became curated by algorithms and communities, I think we’ll see a shift in how individual personal landscapes of information are curated, from manual organization to mostly machine-driven organization. Humans will leave connections and highlights as a trail of their thinking, rather than as their primary way of exploring their knowledge and memory.
    2. However, to build an enabling medium that’s more than a single-purpose tool, it isn’t simply enough to look at existing workflows and build tools around them. To design a good creative medium, we can’t solve for a particular use case. The best mediums are instead collections of generic, multi-purpose components that mesh together well to let the user construct their own solutions.
    3. The current state web browsers is particularly damning from this perspective. Web browsers have access to such a treasure trove of valuable, often well-structured information about what we learn and how we think, what interests we have, and who we talk to. Rather than trying to take that information and let us build workflows out of them, browsers remain a strictly utilitarian tool – a rectangular window into documents and apps that play dumb, ignorant of the valuable information that transits through them every day.
    4. The vision of the web browser that excites me the most is one where the browser is a medium for creativity, learning, and thinking deeply that spans personal and public spheres of knowledge. This browser will be fast and private, of course, but more than that, this browser will let me explore the Web from the comfort of my own garden of information. It’ll break the barriers between different apps that silo our information to help us search and remember across all of them. It’ll use a deeper machine understanding of language and images to summarize articles, highlight important ideas, and remind me what I should remember. It’ll let me do it all together with other people in a way that feels like real presence, rather than just avatars on screen.
    5. Most existing tools and browsers treat web pages and pieces of notes like complete black boxes of information. These tools know how to scan for keywords, and they have access to the metadata we use to tag our information like hashtags and timestamps, but unlike a human, most current tools don’t try to peer into the contents of our notes or reading materials and operate with an understanding of our information. With ratcheting progress in machine understanding of language, I think we have good high-quality building blocks to start building thinking mediums and information systems that operate with some understanding of our ideas themselves, rather than simply “this is some text”.
    6. So, what are the building blocks of a powerful thinking medium that can actually help us think, more than just recall? For a tool that has such broad access to information like a web browser, I think a critical piece of the puzzle is better machine understanding of language.
    7. If we want to organize information that flows through our lives, we simply can’t restrict our design space to be a single product or app. No matter how great a note-taking app is, my emails are going to live outside of it. No matter how seamless the experience in my contacts app, my text conversations are going to live outside of it. We should acknowledge this fundamental limitation of the “note-taking app” approach to building tools for thought, and shift our focus away from building such siloed apps to designing something that lives on top of these smaller alcoves of personal knowledge to help us organize it regardless of its provenance. If we want to build a software system that can organize information across apps, what better place to start than the one piece of software that has access to it all, where most of us live and work nearly all the time? I think the browser is a rich place to build experiments in this space, and my personal experience building Monocle and Revery support this idea so far.
    8. In the browser of the future, the boundary between my personal information and the wider Web’s information landscape will blur, and a smarter, more literate browser will help me navigate both worlds with a deeper understanding of what I’m thinking about and what I want to discover. It’ll remind me of relevant bookmarks when I’m taking lecture notes; it’ll summarize and pick out interesting details from long news articles for me; it’ll let me search across the Web and my personal data to remember more and learn faster.
    9. Despite the renewed focus I see in the community of people and companies trying to build better tools for thought, I think much of our work is still confined to tool-making. That is, most of our efforts are about creating more automatic, more efficient ways to do what we already know how to do – spaced repetition, Zettelkasten, journaling, and so on. We are busy making more effective command-line apps for thought, rather than dreaming up graphical interfaces.
    10. Designing a medium for thought requires that we discover what these primitive components of a thinking medium should be. Should there be some sense of geometry and space? How important should text be, against drawings and images? How should people collaborate and share their thoughts? I propose that the solution to these questions are not an opinionated tool with a “Share” button and a rigid way to use an image in a project, but something with a collection of capabilities that happen to include inserting and positioning text and images, sharing and collaborating on those objects on the page, and connecting ideas.
    1. Participation inequality plagues the internet. Only 1% of people on any given platform create new content. 99% only consume.Many think that's just what happens when human communities scale. But maybe it's just what happens in an internet built for advertising. Consider that:All of the internet's interfaces—social feeds, search bars, news sites—are optimized for consumption.Interfaces for creating new content, particularly knowledge, are antiquated. Word-processors look like they did forty years ago, disconnected from the internet and any content you might write about. Which means: writing requires hours of searching and sorting. Knowledge creation is painful for the people best at it, and inaccessible to most others. What would it take to make writing accessible? Maybe: a totally new kind of interface. Ideally: a word-processor that pulls in the information you need as you type. And what would that take?Unprecedented NLP to make connections as you type,A word-processor redesigned around links, andA highly technical team focused on a non-technical market.If achieved, it would:save writers hours,make knowledge production accessible to anyone who knows how to type, andlay the groundwork for a mainstream knowledge economy.
    1. So we’re starting out with a tool focused on those two core jobs we have: capturing ideas in the moment, and making sense of them as you grow your web of ideas.
    1. But both of these issues are trivially solved if we simply begin with today's lightly hyperlinked documents, and let the reader's computer generate links on-demand. When I'm reading something and don't understand a particular word or want to know more about a quote, when I select it, my computer should search across everything I've read and some small high-quality subset of the Web to bring me 5-10 links about what I've highlighted that are the most relevant to what I'm reading now. Boom. Everything is a hyperlink.
    1. One interesting observation about these markets is that the upper-limit for value creation of individual ventures here, one creator, one community, etc, is smaller than startups. There are only a handful of trillion-dollar communities in the history of humanity, and no trillion-dollar influencers or entertainers. Scale will come more from breadth, not just making one good bet, as is the case in startups.
    2. The next Y Combinator will identify an industry where most people are under-valuing a certain class of talent in an industry poised to grow rapidly, gather a community to commoditize tribal knowledge, and over time scale it into a self-sustaining flywheel that helps them grow their gravitational field. As for which industry, my guess is as good as anyone else’s – unpredictability is a core element of disruption, almost definitionally – but I think there’s a good chance it’ll come from one of these areas: The creator economy – people who can independently grow an audience and monetize them sustainably. Entertainment, as an old, bureaucratic industry adjacent to this space, is also an interesting target. Communities – I’ve written extensively about my bullishness on communities elsewhere. Higher education – what replaces Harvard and Stanford?
    1. For the past two decades, Evergrande has taken advantage of cheap credit to massively expand its real estate empire. The company reported $108 billion in annual sales and some $356 billion in assets. Its portfolio is massive: Evergrande owns over 1300 real estate projects across 280 cities in China, has a property services management arm involved in almost 2800 projects across over 310 cities in China, has other ventures across unrelated industries, and directly employs over 200,000 people.
    2. In a page out of the WeWork playbook, Evergrande fancied itself as more than a loss-leading, debt-burdened real estate business. The company turned its health care business unit into an electric vehicle company despite never making a single car. In July, Evergrande was considering an IPO for Evergrande Spring, its bottled water business (you read that right). At the same time, Evergrande was also considering an IPO for its tourism business and an incoherent theme park featuring fairytales from across the world. Guangzhou FC, China's most successful soccer team, is owned by Evergrande (with an Alibaba stake) for some reason.
    3. The situation has grown so desperate for the cash-strapped firm that it demanded employees loan Evergrande money at a high interest rate in order to keep their bonuses, only to stop paying back those loans soon after. The situation has sparked protests led by home buyers waiting on unfinished homes and employees who have lost their life savings.
    4. Research firm Capital Economics estimates that some $200 billion of Evergrande's debt is actually cash put down for 1.4 to 1.6 million unfinished properties, and a huge number of justifiably furious purchasers of these unbuilt homes.
    1. China's debt-laden local governments are also facing grim prospects. Declining income from land sales because of the crisis in the real estate sector and falling tax receipts are expected to cause a 6 trillion yuan shortfall, roughly $900 billion, in local government revenues this year. Local government financing vehicles that have borrowed heavily from banks or issued bonds will have great difficulties servicing their debt.
    2. Weak supervision, poor risk management and corruption that likely drove the small rural banks in Henan into insolvency are systemic among the country's nearly 4,000 small and medium-sized banks with nearly $14 trillion assets.
    3. Large banks in China are in trouble as well. They have lent tens of billions to poor countries as part of China's ambitious Belt and Road Initiative. A significant portion of their credit portfolio is likely to become nonperforming as their borrowers are unable to service the debt due to the global economic downturn.
    1. In retrospect, there's a good case that the reason the bursting of the housing bubble seemed to matter so much was that it was masking an underlying problem of secular stagnation. The same might be said of Japan's real estate-stock bubble of the late 1980s
    1. Steep payments on international sovereign bonds, which comprised nearly 40 percent of the country’s external debt, put Sirisena’s government in dire fiscal straits almost immediately. When Sirisena took office, Sri Lanka owed more to Japan, the World Bank, and the Asian Development Bank than to China. Of the $4.5 billion in debt service Sri Lanka would pay in 2017, only 5 percent was because of Hambantota. The Central Bank governors under both Rajapaksa and Sirisena do not agree on much, but they both told us that Hambantota, and Chinese finance in general, was not the source of the country’s financial distress.
    2. In 2012, Sri Lanka borrowed another $757 million from China Eximbank, this time at a reduced, post-financial-crisis interest rate of 2 percent. Rajapaksa took the liberty of naming the port after himself.
    3. This was in 2007, six years before Xi Jinping introduced the Belt and Road Initiative. Sri Lanka was still in the last, and bloodiest, phase of its long civil war, and the world was on the verge of a financial crisis. The details are important: China Eximbank offered a $307 million, 15-year commercial loan with a four-year grace period, offering Sri Lanka a choice between a 6.3 percent fixed interest rate or one that would rise or fall depending on LIBOR, a floating rate. Colombo chose the former, conscious that global interest rates were trending higher during the negotiations and hoping to lock in what it thought would be favorable terms. Phase I of the port project was completed on schedule within three years.
    4. The prime example of this is the Sri Lankan port of Hambantota. As the story goes, Beijing pushed Sri Lanka into borrowing money from Chinese banks to pay for the project, which had no prospect of commercial success. Onerous terms and feeble revenues eventually pushed Sri Lanka into default, at which point Beijing demanded the port as collateral, forcing the Sri Lankan government to surrender control to a Chinese firm.
    5. There was also never a default. Colombo arranged a bailout from the International Monetary Fund, and decided to raise much-needed dollars by leasing out the underperforming Hambantota Port to an experienced company—just as the Canadians had recommended. There was not an open tender, and the only two bids came from China Merchants and China Harbor; Sri Lanka chose China Merchants, making it the majority shareholder with a 99-year lease, and used the $1.12 billion cash infusion to bolster its foreign reserves, not to pay off China Eximbank.
    6. Our research shows that Chinese banks are willing to restructure the terms of existing loans and have never actually seized an asset from any country, much less the port of Hambantota.
    7. China’s march outward, like its domestic development, is probing and experimental, a learning process marked by frequent adjustment. After the construction of the port in Hambantota, for example, Chinese firms and banks learned that strongmen fall and that they’d better have strategies for dealing with political risk. They’re now developing these strategies, getting better at discerning business opportunities and withdrawing where they know they can’t win.
    8. Over the past 20 years, Chinese firms have learned a lot about how to play in an international construction business that remains dominated by Europe: Whereas China has 27 firms among the top 100 global contractors, up from nine in 2000, Europe has 37, down from 41. The U.S. has seven, compared to 19 two decades ago.
    1. After nearly one million Chinese people were unable to access their bank deposits in central China’s Henan province earlier this year, residents in east China’s Shanghai, south China’s Shenzhen, north China’s Dandong, and central-east China’s Jiujiang reported the difficulties they faced when trying to withdraw cash from their bank accounts. Some banks only serve a limited number of customers per day, some banks limit each client’s withdrawal to no more than 1,000 yuan and others have closed their branches. Even the ATM machines are empty. Bank runs have been happening in the world’s second-largest economy for over a week, which is unusual in China because most of the banks are state-run. “The reason why the bank run issue hasn’t been solved is that China’s economic system is in crisis and the Chinese regime doesn’t have the ability to solve it,” Wang He, U.S.-based China affairs commentator, told The Epoch Times. Zheng Yongnian, one of the economic advisors to Chinese president Xi Jinping, published an essay on June 1, in which he pointed out that China’s economy is facing critical challenges, including over half of the foreign investments, have left China, and China’s private businesses are struggling for survival due to a supply chain crisis and lack of cash. Zheng’s essay was removed from China’s internet soon after it was published.
    1. The protest involved more than 100 delayed projects as of July 13, up from 58 projects just one day earlier. The frustrated buyers accuse the developers of misusing sales proceeds and the banks of failing to safeguard their loans.
    2. But this mortgage strike isn’t entirely unpredictable. Homebuyers have every reason to be angry. Most of the projects were begun by developers who have defaulted. China Evergrande Group led the pack, accounting for an estimated 35% of the total projects that faced mortgage revolts, data compiled by CLSA shows. One such project in eastern Jiangsu province was launched before the Covid-19 pandemic. Construction has been suspended since last August, while property values in its neighborhood has come down by about 10%.
    1. As far back as the spring of 2007, U.S. banks began to underperform financial institutions in Asia. By now, everyone knows most of the subprime mortgage credit risk was inside the USA with domestic banks more exposed than other banks around the world. Notice above, Goldman Sachs (green above) 5 year CDS (the cost of default protection on the bank), began to meaningfully divergence from Standard Chartered. Standard Chartered PLC is an international banking group operating principally in Asia, Africa, and the Middle East. The company has far more credit risk exposure to China – Asia than U.S. banks. It is clear above, more than 12 months prior to Lehman´s failure, banks in the USA were dramatically underperforming from a credit risk perspective. In other words, in 2007 – the cost of purchasing credit default protection on Goldman Sachs was far more expensive than the bank´s Asian peers. Indeed, elephants leave footprints – when large hedge funds see credit risk – they start placing bets months if NOT years before a credit event. The credit market sniffed out Lehman´s demise months BEFORE equity investors got the joke.
    2. Fast forward to 2021, Evergrande headlines are all the media rage, especially with the Lehman, the lucky 13th anniversary this week. But, what are credit markets telling us this time? As you can see above – far right. Credit risk is calm on Asia banks with exposure to China, no difference to speak of. Central bank liquidity is so abundant, there is NO way Lehman would have failed today.
    3. Always with an important lens – our friend, Jens Nordvig reminds us – “foreign involvement is small in China. It is true that the high-yield bond market has a sizable USD component (mostly foreign). But relative to the US, where subprime exposure was sold around the world, it is a much more local (controllable) system.” It has been clear for months, there is Evergrande credit contagion – it’s just inside China at the moment.
    1. Evergrande is the most-indebted property developer in the world. Its on-balance-sheet liabilities amount to nearly 2 percent of China’s annual GDP, and its off-balance-sheet obligations add up to as much as another 1 percent.
    2. For the past several years, Chinese regulators have worked hard—if unsuccessfully—to reduce the economy’s overreliance on debt. As part of this regulatory push, regulators implemented what became known as the three red lines for property developers last year. These consist of hard limits on a company’s debt-to-asset ratio, its debt-to-equity ratio, and its cash-to-short-term-debt ratio.
    1. Evergrande group has more than $300bn of liabilities, about $20bn of which are offshore dollar-denominated bonds. The Chinese government has focused on completing work on its hundreds of projects, where homes have typically been sold to ordinary buyers before completion.
    1. A weaker euro makes Europe’s exports cheaper while helping to lure overseas tourists to the beaches and resorts of Greece and Spain. That export-boosting effect is being eaten up by a large increase in the price of the continent’s imports, especially energy and raw materials, many of which are priced in dollars, analysts say. Those price increases are driving up inflation across the currency bloc. (function () { var adOptions = {"options":{"adActivate":true,"adId":"wsj-body-AD_UNRULY","adRequestOnRemount":true,"adSize":[[2,2]],"adSizeMap":{"at4units":[[2,2]],"at8units":[[2,2]],"at12units":[[2,2]],"at16units":[[2,2]]},"adTargeting":{"adlocation":"UNRULY","circ":"","msrc":"","news_id":"","psg":""},"adUnitPath":"/2/interactive.wsj.com/markets_foreignexchange_story","checkIfRendered":false,"collapseAdBeforeFetch":true,"hideAd":false,"isMetaTag":false,"isObserve":true,"isTemplate":false,"isUtagData":true,"label":"","labelPosition":"top","moatEnabled":true,"noWrapper":false,"reserveInitialHeight":false,"rootMargin":"0px 0px 500px 0px","shouldUpdate":true,"staticHeight":{},"threshold":0,"triggerAdBidding":true,"triggerApstag":true,"triggerPrebid":true,"labelClasses":"","location":"L","responsiveContainer":false,"adLocation":"UNRULY","pageId":"markets_foreignexchange_story","params":{},"trackingKey":"interactive.wsj.com/markets_foreignexchange_story","wrapperStyles":{"width":"100%"},"observeFromUAC":true},"content":{}}; window.adslots = window.adslots || {}; (window.adslots.adIds = window.adslots.adIds || []).push('wsj-body-AD_UNRULY'); window.__ace('uac', 'renderAd', [adOptions]); })(); “The extreme price increases in import and producer prices overshadow any profit that exporters can book for themselves due to a weaker currency,” said Sonja Marten, head of foreign exchange and monetary policy research at DZ Bank in Frankfurt.
    1. To be sure, if gasoline prices stay at elevated levels but don’t rise further, eventually that will translate to lower annual inflation. “What we’re really trying to do is just get [commodity prices] to stop going up. If it leveled off…then the inflationary effects go away. We don’t need them to actually fall back down,” Fed Gov. Christopher Waller said last week.
    2. According to the PCE index, consumer prices rose 6.3% in May from a year earlier. Core PCE prices rose 4.7% in May from a year earlier, down from a peak of 5.3% over the 12 months ended in February. Core PCE inflation has slowed to a 4% annualized rate over the February to May period, the lowest four-month rate since March 2021.
    1. In Somalia, Ethiopia, South Sudan, Yemen and Afghanistan, nearly 900,000 people already face starvation and death. That is a more than 10-fold increase from 2019—and, by some estimates, could result in more people dying from hunger in 2022 and 2023 than in any years since the 1960s and China’s disastrous Great Leap Forward agricultural policies.
    2. The World Food Program says that increases in the cost of food and fuel since March have pushed an additional 47 million people into acute food insecurity, when a person is no longer able to consume enough calories to sustain her life and livelihood, taking the total to 345 million people world-wide. Of those, some 50 million are living on the edge of famine.
    1. The United States was once a global leader in both solar innovation and manufacturing — we invented photovoltaic technology in the 1950s to power satellites and spacecraft. And we retained our undisputed leadership in solar for decades. But lately we have seen our solar industry get crushed as a direct consequence of China’s huge state subsidies to its manufacturers beginning in the 2000s.Since then the U.S. share of solar component shipments worldwide fell to less than 1 percent in 2021 from 13 percent in 2004. China’s share of the production of solar components has increased over the past two decades from virtually nothing to nearly 85 percent today. For some solar components, that could rise to 95 percent in the coming years, according to a report this week from the International Energy Agency. Simply put, the United States did not provide enough support to its solar companies, which could not compete in the face of China’s state-directed strategy to control the market.
    1. Inflation is the biggest reason for the diverging positions of the two countries’ central banks. In April, overall consumer prices in Japan rose 2.5% from a year earlier. The index excluding volatile fresh-food and energy prices rose just 0.8%. In the U.S., meanwhile, consumer inflation reached an 8.6% annual rate in May as energy and food prices surged.
    2. A cheap yen is typically a boon for the Japanese economy, which is driven by exports such as automobiles. Japan’s exporters, however, are starting to show concern over the yen’s latest slide because it is happening at the same time as rises in commodity prices and supply shortages.
    1. The South Asian nation’s wilting economy has seen Sri Lankans endure months of double-digit inflation, rolling power blackouts and severe shortages of fuel and medicine. Sri Lanka’s foreign reserves are depleted to the point that it can no longer afford to pay for essential imports and the country defaulted on its debt for the first time in its history in May.
    1. Government employers were the only category to report a decline in payrolls, with a seasonally adjusted loss of 9,000 jobs in June.
    2. The continued strength in the labor market was led by employers in the category of professional and business services, a broad group of white-collar industries that added 74,000 jobs last month, the Labor Department said Friday. Those gains were concentrated in management roles, software development and office administrative services.Such white-collar employers had 880,000 more jobs on their payrolls in June than in February 2020, before the Covid-19 pandemic hit the U.S. economy, despite a series of recent high-profile layoffs at companies such as Redfin Corp. and Robinhood Markets Inc.
    1. Kansas City Fed President Esther George, the only voting policy maker (out of 11) to dissent from the 75 basis-point hike -- she preferred 50 basis points -- had no company among the FOMC's seven nonvoting members: The minutes said only “one participant,” referring to George, favored a half-point increase.
    1. Fed researchers updated the index in April to include additional indicators of inflation expectations. In a June 15 note to clients, Goldman Sachs economist Ronnie Walker said the revised measure puts greater weight on short-term inflation expectations and on households’ attitudes to prices than previously. That tilts the index higher, he wrote. 
    1. “The index of common inflation expectations at the board has moved up after being pretty flat for a long time, so we’re watching that, and we’re thinking, ‘This is something we need to take seriously,’” Powell told reporters on June 15 after the Fed raised interest rates by 75 basis points, the biggest increase since 1994.
    1. The Royal United Services Institute report noted that Ukraine’s paucity of skilled infantry and armored-vehicle operators limit its abilities to launch serious counteroffensives. Russia’s artillery is also successfully targeting the Ukrainian military so that it is unable to mount attacks, the report said.
    2. In a report this week, the U.K.’s Royal United Services Institute think tank said Ukraine needs long-range artillery systems and electronic warfare equipment to counter Russia’s own advanced systems.The report said that Ukraine’s allies are capable of making up the shortfalls. But, in a nod to the complications of operating too many different weapons systems, the report said success “cannot be achieved through the piecemeal delivery of a large number of different fleets of equipment, each with separate training, maintenance and logistical needs.”
    3. Moscow’s success in the east recalls some of Russia’s previous wars. As in Chechnya in the 1990s, Russia is offsetting the shortcomings of its ground forces by using massed firepower.Ukraine is trying to impose high costs on the advancing Russians with fierce resistance, while also preserving troop strength by slowly falling back to more defensible positions as it awaits flows of heavy weapons from the West. It took Russia two months to take the city of Severodonetsk.
    1. During strong dollar environments (based on various fiscal and monetary policies), liquid capital flows more towards US equities, whereas during weak dollar environments, it flows elsewhere. This can create a self-perpetuating cycle in one direction or another until a number of catalysts and policy changes lead to a reversal. So, US equity valuations became very high in 2000 (strong dollar period), and then emerging market valuations became very high in 2007 (weak dollar period), and in recent years US equities have once again been highly-valued in a strong-dollar period.
    2. If you expect interest rates to remain near zero for a multi-decade period, then you can infer that fast-growing companies would likely remain extraordinarily highly-priced for a long time. The growth/value forward performance gap would likely narrow, however, because interest rates would likely not keep getting persistently lower (as they’d have to go into negative yields and then keep digging down into lower and lower negative yields to do so). On the other hand, if rates rise to, say, 3% or more, that could put considerable downward pressure on equity valuations, particularly highly-valued stocks with little or no current earnings but large expected forward growth.
    3. Notice that, during the same reduction in the discount rate from 12% to 8%, the growth stock had a notably larger percent increase in its fair value than the value stock (61.8% vs 44.7%). This is because the growth stock has a larger percentage of its expected cumulative cash flows occur far into the future compared to the value stock that is more front-loaded, so a larger portion of the growth stock’s expected cumulative cash flows are subject to the multi-year compounding effect of the discount rate. The growth stock is more rate sensitive, in other words. Whenever we have a big shift to lower interest rates, and the market expects those low rates to persist for a while, then it’s not surprising to see a surge in valuations of growth stocks during that transition from higher to lower rates.
    4. This current period in the 2020s is more difficult for people because despite the fact that stock valuations are so high, bond valuations are so high as well, and investors are stuck between a rock and a hard place. They have to venture into things like value stocks, commodities, or international equities if they want something resembling a historically normal price for things.
    5. The lowest that this version of the equity risk premium ever reached was in 2000, at -4%. That’s why the Dotcom bubble was so silly; you could get a 6.5% yield on a Treasury note, but investors were paying a cyclically-adjusted earnings yield of less than 2.5% for the S&P 500. Then, equities crashed. Treasuries were clearly the better buy there.
    1. Whenever valuations are reasonable, I think investors should consider maintaining some oil and gas exposure, in the form of low-cost producers in diverse jurisdictions, energy transporters with solid balance sheets, and/or long-dated oil futures. Being long uranium holding companies or uranium miners is also a nice diversifier along these lines.
    2. While I do think solar and wind will become a larger percentage of the global electrical grid in an additive way, they’re not very well-suited to actually replace prior energy sources, unless we make some absolutely massive (not merely incremental) breakthroughs in energy storage or turbine/panel longevity. This sets up a decent asymmetric investment opportunity in my view. I think this gap between engineering viability and public/investor perception is providing a long-term opportunity for oil and gas investors, as well as uranium investors. We’re embedding a lot of assumptions into the future energy mix, without necessarily investing in the capex required for that vision, or fully considering some of the technical challenges associated with that vision.
    3. A lot of ESG mandates, current valuations, and capital allocation practices in general, are assuming a rather rapid phase-out of fossil fuels in favor of wind and solar energy. They’re betting on being able to phase out prior energy sources and to reduce the power density of our primary energy sources, for the first time in human history, as though it’s a given. Many of them also appear to be underestimating the technical limitations or true “greenness” of solar and wind power, as it relates to the energy return on investment, recycling issues, reliance on fossil fuels for manufacture, reliance on China to make them cheaply, and so forth.
    4. In a slow-growing country, it’s possible to gradually replace a portion of prior energy sources with new energy sources, at least for the grid. In a fast-growing country, it’s nearly impossible to phase out old energy sources, or even stop growing them.
    5. We find that the lost nuclear electricity production due to the phase-out was replaced primarily by coal-fired production and net electricity imports. The social cost of this shift from nuclear to coal is approximately 12 billion dollars per year. Over 70% of this cost comes from the increased mortality risk associated with exposure to the local air pollution emitted when burning fossil fuels. Even the largest estimates of the reduction in the costs associated with nuclear accident risk and waste disposal due to the phase-out are far smaller than 12 billion dollars.
    1. This issue remains in the newly proposed pair model. To counteract the long-skew and make GLP more delta-neutral, GMX could introduce a negative borrow fee (similar to a funding rate) for the underutilized side. This is common practice for perp exchanges when the perp price deviates from the spot price. GMX doesn’t have perp prices, so the funding rate mechanics need to be adapted.
    2. To account for different risk profiles of LPs, it makes sense to create a variety of GLPs with different token holdings and weights. This ultimately leads to more liquidity and raises the IO ceiling. As this could result in liquidity fragmentation, the GMX team proposed to restructure GLP based on pairs
    3. As of January 4, leverage tradoooors made a loss of -$6.2m. LPs earned $8m in fees, resulting in 14.2m net profit.
    4. There’s a 1.5% price change limit in place for positions that have been open for under 24 hours to avoid bots from taking advantage of the lag in oracle pricing. This limits trading volume from arbitragoooors.
    5. GMX takes a fee of 0.1% for opening and closing positions plus a dynamic borrow fee based on utilization rates. Swap fees are also dynamic, related to whether a swap improves the weights of assets in the pool towards the target or away. The same applies to GLP minting and redemption fees. LPs can burn their GLP to receive any of the assets in the pool. In case an LP redeems GLP for an over-weighted asset, fees are reduced.
    6. GMX is a mix of a perp and spot exchange for traders + an index fund for LPs (GLP).Perp: Traders can long/short utilizing assets from the liquidity pool (GLP) with up to 30x leverage. Perp trading is responsible for ~85% of protocol revenue at the moment.Spot: Traders can swap tokens with zero price impact. GMX uses Chainlink oracles together with FTX and Binance market data to price assets in the pool, instead of an AMM formula. As a result, users won’t incur slippage.GLP: LPs hold a basket of assets (ETH, BTC, USDC, …). Therefore, they are exposed to the price of the underlying assets, adjusted by the positions of traders (as LPs automatically take the other side of each trade). LPs basically rent out the upside of assets to traders. For this, LPs earn 70% of protocol revenue from fees as well as GMX rewards. Holding GLP can be seen as a long-term strategy compared to the short-term view of traders.
    7. Supply-side: All the potential demand needs to meet supply on the other side. On GMX, liquidity providers (LPs) provide capital for traders to long/short assets with up to 30x leverage. LPs receive 70% of protocol revenue, 30% go to GMX stakers. GMX currently generates ~$112k in daily fees, or ~$41m annualized. $29m go to LPs, which equates to 28% APR paid in ETH, based on protocol fees alone. For the basket of assets (called GLP) that LPs hold and the risks involved, this is a competitive offering and acts as a first indication that the protocol could be sustainable. This would allow liquidity to scale with demand without relying on token rewards too heavily. However, LPs take on risks that could especially materialize in a bear market. Therefore, further improvements are required to ensure better performance for LPs during difficult market conditions, which the team is aware of.
    8. Demand-side: GMX offers traders decentralized derivatives, a highly-demanded product. BTC and ETH perpetual futures contracts have a monthly trading volume of ~$2T. In comparison, spot volume is roughly 7x smaller, with ~$300B per month. dYdX, the leading perp DEX currently facilitates ~$3B in perp trading volume daily (~5% of total volume). GMX stands at ~$70m on Arbitrum, where the protocol only recently launched in September. A narrative shift towards L2 solutions more generally could act as a catalyst for GMX. Arbitrum has the highest TVL of all L2s. GMX is ultimately chain-agnostic, expanding to Avalanche on January 6, further increasing the target market. With the recent crackdowns on centralized exchanges to add KYC and reduce the amount of leverage offered to traders, demand will continue to shift over to decentralized applications. GMX is run by an anon team, which puts it at an advantage over its direct competitor dYdX in this regard. As a result, GMX is in a pole position to capture a share of the (growing) demand for decentralized derivatives.
  5. entrepreneurshandbook.co entrepreneurshandbook.co
    1. As we know, derivatives trading volumes should be orders of magnitude higher than spot volumes. Whichever few DEXs comprise the natural monopoly will earn beaucoup fees. That is the future bull case for the category itself, and from what I have seen so far, GMX is the best out there. However, I don’t believe it is elegant enough … yet … to truly capture the fee wallet of DeFi leveraged traders en masse and catapult volumes over spot DEXs such as Uni and Sushi.
    2. In the derivative DEX space, dYdX dominates the field. That said, I have a very puritanical objection to the dYdX model, which is that dYdX is not truly a DEX. It is a centralised orderbook hosted on a dYdX machine, and only settled trades are posted on chain for finality. But more importantly, when it comes to its P/E ratio, it is markedly higher than the protocol I’m most impressed by, GMX.
    3. Compared with the CEX behemoths, the derivative DEX average daily trading volume numbers are quite low. But that only gives more upside to the faithful. Just to match the market share that spot DEXs possess vs. their CEX counterparts at 10% would be a 5x improvement in trading volumes. I’ll take that.
    1. The funding rate is how the price of a perpetual swap is kept close to the price of the underlying asset. It works by sending periodic payments between long and short traders. This is critical: a poorly designed funding rate makes perpetual swaps riskier, more volatile, and costlier.
    1. To be sure, though perpetuals have become super popular, the broader market is still mired in a downturn. Open interest -- or the amount of open contracts -- tends to be high during bullish periods because rising prices attract speculators to the market. But OI has declined roughly 40% since the fall. Aggregate futures trading volume has also declined -- during the first half of 2021, it was typical to see trading volumes between $70 billion to $80 billion per day. As of the end of April, that had dropped by roughly 60% to around $30 billion daily, according to Glassnode, though it ticked up amid the last few weeks’ volatility.
    2. Bitcoin perpetual contracts -- which, unlike traditional calendar futures, don’t expire -- have long been a crypto retail-investor favorite, with a bigger number of professional and institutional traders also shifting toward them and away from traditional calendar futures. Such contracts have grown to account for around 66% of open interest and roughly 93% of trading volume, according to data from Glassnode. Traders say these instruments mimic the spot market, and are easy to use and access on exchanges. And their dominance has ballooned as they’ve become the preferred source of leverage, according to the researcher. 
    1. Yet, much of this year’s selloff wasn’t about recession risk. To see this we need to distinguish the direct and indirect effects the Fed has on prices of stocks and bonds.The direct effect is to push up bond yields and push down valuations of stocks with profits far in the future, which means those with high valuations such as Big Tech. This is what dominated until June, with bond yields soaring and growth stocks crashing, while cheap “value” stocks were basically fine. Exclude the technology sector to strip out the bulk of this effect and economically-sensitive cyclical sectors of the stock market had only slightly underperformed defensives by June 7.
    2. While investors are at last focused on recession uncertainty, risks elsewhere in the world could hit U.S. investors, too. Japan might finally be forced to relent and allow bond yields to rise, which would suck back cash the country’s investors had poured overseas. In Europe, the central bank has promised a new plan to support Italy—but we’ve seen this show before. If it follows the pattern of too little, too late, we could see a return of the eurozone debt crisis, something markets are not prepared for.
    1. Since there have been some questions on the long-term sustainability of the project we feel it is appropriate to address them here. The team currently has ~1.8 million held in USDC, this is sufficient to last 18 months till March 2024.
    1. The next thing to notice is that the write throughput of my prototype database approaches the same long-term rate with or without read threads. However, the support for multi-threaded reading allows my database to sustain over 120,000 operations per second while performing over 30,000 random inserts per second while the state is larger than RAM.
    2. The Green line represents an apples-to-apples comparison of my new database's single-threaded insert against LMDBX and std::map. As you can see it is over 2x faster than LMDBX and std::map while everything is in RAM. We will zoom in on the disk-based performance in a later chart. This means it is likely more than 2x faster than chainbase as used in eosio and hive.
    3. As a point of comparison, I chose one of the highest-rated databases I could find (LMDBX) because it supported reading past states while advancing. Past experience with RocksDB and months of testing and optimization let me know that it would not serve us well. Some Ethereum implementations use LMDBX because of its performance over Rocks. I also chose to compare against the default key/value store native to all C++ code, std::map. std::map isn't a proper "database" and only has to concern itself with maintaining a sorted tree of key/value pairs. It has no other overhead. Its performance starts out high, being in-memory, and eventually forces the operating system to start using virtual memory to swap to disk. At this point, we are simulating the upper limit of chainbase, which was used by Hive and is used by EOS. Chainbase has extra overhead and has historically always been slower than pure std::map.
    1. Typical voting processes have a pre-defined ballot which limits the choice of the people. Using fractal voting there is no need for a pre-defined ballot and all members of a fractal democracy have equal opportunity to be heard. This occurs by randomly grouping members into groups of 5 or 6. Each group must reach a consensus among its own members on who best represents their interests. The chosen representatives of each group are then randomly grouped and the process is repeated ƒractally (in a fractal manner). The end result is the integration of opinions of the entire population without any filtering by the media or political parties. A fractal voting system is what enables fractal democracies to truly empower the people while preventing the covert capture of power by the media, the political parties, or even the military.
    1. “We would have expected major rebalancing flows by now to help equities” after a bad second quarter, El Erian said, and wondered whether markets “are being upset by outflows? Or is it that investors are less willing to rebalance in favor of risk assets?”
    2. “On my radar screens are three sets of potential -- I want to stress, potential -- liquidity strain,” he said. “One is in peripheral markets that somehow contaminate the main markets. So far crypto hasn’t, so far EM hasn’t. The second part of risk is simply the inability to raise funding at any cost. We’ve seen high yield go through this but high yield is less important than if it migrates up the quality ladder.”
    3. Junk-bond issuance dropped to $25 billion in the last quarter, the lowest second-quarter supply since at least 2006, according to data compiled by Bloomberg. June volume was $10 billion, the slowest for the month since 2010, the data show.“We need to look at issuance and make sure that that doesn’t freeze up,” El-Erian said.
    4. “The one risk I’m really worried about is liquidity risk,” El-Erian told Bloomberg Television’s The Open on Friday. “We’re starting to see markets locked out of funding. Issuance in June was very low. Companies either were unwilling or unable to refinance themselves.”
    1. “For rate cuts, the first cut is the deepest,” said Tom Orlik, chief economist at Bloomberg Economics. “For quantitative easing, the biggest bang comes from the first buck. The swing back to net asset purchases by major central banks is a pivotal moment. The second time around, though, quantitative easing will struggle to gain the same traction as it did the first.”
    2. The Federal Reserve’s decision to stop shrinking its balance sheet from Thursday means the era of quantitative tightening by major central banks is over less than a year after it started.
  6. Jun 2022
    1. Some Ukrainian officials have said that Ukraine is currently losing from 100 to 200 men a day, while other Ukrainian officials say that figure is exaggerated.
    1. “This is a case where GDP revisions alters the view of the first quarter,” said Alex Pelle, US economist at Mizuho Financial Group Inc. “Instead of accelerating in the first quarter versus the prior two quarters, consumption actually moderated.” window.__bloomberg__.ads.enqueue("box-TM6cTNG"); {"contentId":"RE8O87T0AFB401","position":"box","dimensions":{"mobile":[[300,250],[3,3],[1,1],"fluid"]},"type":"Mobile Body Box Ad","positionIncrement":1,"targeting":{"position":"box1","positionIncrement":1,"url":"/news/articles/2022-06-29/us-personal-consumption-revised-sharply-lower-in-first-quarter"},"containerId":"box-TM6cTNG"} By the time the government releases its third estimate of GDP and the underlying components, the numbers don’t typically change much. The large downward revision to consumer spending -- paired with a sizable upward revision to inventories -- is quite unusual.
    1. Some may argue that such models of coordination are replicable without crypto— stock agreements and bonus rewards for aligned actors are already present in most centralized organizations. However, crypto-economic protocols unlock five key benefits:Rapid Scale — Permissionless and borderless protocols can grow everywhere in the world in parallel across many legal jurisdictions.Credible Neutrality — Credibly-neutral networks give their stakeholders guarantees that the rules can’t arbitrarily be changed from underneath them.Collective Ownership — A system that is owned by its users creates loyalty, aligns incentives, and drives growth.Frictionless Payments — Blockchains allow for peer-to-peer micropayments that the legacy payments system cannot support.Integration with DeFi rails — DeFi rails are useful for bootstrapping liquidity via automated market makers. Overtime, we also expect proof-of-physical-work networks to leverage other composable DeFi-native tools, including NFT marketplaces, social tokens, derivatives, and more.
    2. The vast majority of crypto-innovation to date has been focused on coordinating digital communities and economies; however, tokens also create opportunities for innovation in capital formation and human coordination that extend beyond the digital world and into the physical. We refer to this thesis as “proof of physical work.” Protocols that fit this thesis incentivize people to do verifiable work that builds real-world infrastructure. Relative to traditional forms of capital formation for building physical infrastructure, these permissionless and credibly-neutral protocols:Can build infrastructure faster—in many cases 10-100x fasterAre more attuned to hyper-local market needsCan be far more cost effective
    1. This is bad news for the big companies in the S&P 500, whose foreign earnings will now be worth less in dollars. But a strong US currency also tends to inhibit earnings for everyone else. Over time, as this Absolute Strategy chart shows, surges in the dollar tend to be followed by falls in global earnings. The pressure a stronger dollar exerts on all those who need to buy dollar-denominated commodities or repay debt denominated in the currency makes this an inevitability
    2. From the top-down point of view of the asset allocators who take BofA’s survey, then, it seems obvious that profits are coming down. Higher prices will eat into demand and revenues, higher rates will increase financing costs, wage demands will tighten margins, and so on. However, that is not the view of the brokers’ analysts who Bloomberg surveys to produce profit estimates. S&P 500 earnings expectations have risen 3% so far this year. Admittedly, this is mostly thanks to energy companies. Exclude them and forecasts are down minimally.
    3. Valuations could certainly go lower still, but the bulk of the valuation-led part of the selloff is over. Now, the question is whether the earnings expectations on which those multiples are based are accurate. If expectations are over-optimistic and need to be cut, then there is room for share prices to fall further without much multiple compression. The BofA survey found serious bearishness about the profit outlook. Only in the immediate aftermath of the Lehman Brothers bankruptcy in 2008 has a greater proportion of fund managers expected global profits to fall
    1. Traders are monitoring a summit of the Group of Seven leaders, who are discussing the viability of a price cap on Russian oil and adopted a declaration pledging to support Ukraine “for as long as it takes.” President Volodymyr Zelenskiy joined the summit by video link from Kyiv and said he wants the war to be over by the end of the year, according to officials familiar with his remarks.
    1. Canada is among the world’s largest producers of natural gas but it lacks export infrastructure on its eastern coast. Building a new LNG terminal could take a decade to get through the regulatory process and would likely face fierce opposition from environmental groups.However, there is an import terminal owned by Spanish firm Repsol SA that could be converted to an export terminal relatively quickly. If the company decides to proceed, Canada’s Natural Resources Minister Jonathan Wilkinson has said the terminal could start shipping gas in three to four years.
    1. “We could switch some production from gas to oil if needed, but it would be five-times less efficient,” Hagen Pfundner, head of the German operations of Swiss drugmaker Roche Holding AG. “That would not be a durable solution.”
    2. Europe’s increased demand for liquefied natural gas will also hit poorer nations around the world as they struggle to compete for cargoes. 
    3. Germany’s vice chancellor drew a parallel between the gas squeeze and the role of Lehman Brothers in triggering the financial crisis. If energy suppliers continue to pile up losses by being forced to cover missing Russian supplies at high prices, there’s a risk of a broader collapse.
    4. The latest figures show that it would take 115 days to reach the government’s target of filling gas reserves to 90% capacity by November. That time frame assumes flows remain at the current level, which is unlikely given the Kremlin’s increasingly aggressive posture toward Europe in retaliation for sanctions imposed over Russia’s war in Ukraine. 
    5. “Companies will move production to where there’s competitive pipeline gas, and this won’t be in Germany,” said Wolfgang Hahn, managing director of Energy Consulting Group GmbH. “You can’t correct 20 years of policy errors in two or three years.”
    6. In Germany, some industrial furnaces have been running without interruption for decades. If they cool down suddenly, the molten materials harden and the system breaks. That’s the kind of concern sweeping through Europe’s largest economy as it faces an unprecedented energy crisis.What started as vague foreboding over reduced supplies of Russian gas is now very real. After President Vladimir Putin slashed flows on the main link to Europe by 60%, experts in Chancellor Olaf Scholz’s administration this week worked out the scenarios and none of them led to sufficient reserves to make it through the winter.  
    1. As of Sunday, production has dropped more than 50% below the average 520,000 barrels a day Ecuador extracted from its Amazon territory before the protests began. A total 1,176 oil wells have been forced to stop pumping.
    1. Supplies could fall further as the Nord Stream pipeline linking Russia to Germany is due for a scheduled maintenance closure on July 11. The closure would normally be for just over 10 days but analysts and officials are concerned the pipeline may not reopen at all this time.
    2. Under German law, strategic gas reserves must be 80% full by October and 90% by November—a scenario now becoming very unlikely to be met. When the government triggers the third level of the plan, known as the “emergency phase,” the country’s energy regulator can begin to ration gas.
    3. Berlin triggered the second of its three-step plan to deal with gas shortages after the Kremlin-controlled energy giant Gazprom, the country’s biggest gas exporter, throttled delivery via the Nord Stream pipeline by around 60% last week. Germany’s gas reserves are at 58% capacity, and the government now expects a gas shortage by December if supplies don’t pick up, Economy Minister Robert Habeck said.The second step, dubbed the “alarm level” is a prerequisite for the government to enforce some of the gas-saving measures it announced at the weekend, including substituting coal to gas for power generation and creating financial incentives for companies that consume less gas.Rationing, which would come in the third step, would focus on industry and could severely impact companies that use gas as fuel or as a raw material for production, likely pushing Europe’s biggest economy into recession, economists and company executives have warned.
    1. The World Bank, in a 2019 report, noted that governments also ban exports when prices are high and encourage exports when prices are low, amplifying price swings in both directions. In 2010-11, such policies contributed to a jump in wheat and maize prices that tipped 8.3 million people into poverty, it estimates.
    2. But governments today are often doing the opposite, according to the World Bank. Their “policies so far have taken the form of tax cuts and fuel subsidies, especially for gasoline…Such measures actually increase demand and put further upward pressure on the prices of crude oil and other petroleum products.”
    1. Managers figure that if gas supply stays above 50% of Ludwigshafen’s maximum demand, they can continue to operate by reducing the load and using substitutes. If gas supply falls significantly below that over a sustained period, they would have to stop production, the company said. The threat of gas rationing is growing and Russia is likely to continue to curtail gas deliveries, German officials and analysts say.
    2. “To put it plainly: There is no short-term solution to replace natural gas from Russia,” BASF Chief Executive Martin Brudermüller said in April.
    3. At BASF’s Ludwigshafen site—a city within a city with over 60 miles of roads, some eight restaurants and a wine cellar—natural gas is fed into an intricate system of pipes and spigots to reach plants making ammonia and acetylene, a compound used in plastics and pharmaceuticals. The site is responsible for as much as 4% of German gas demand.
    4. The threat isn’t just to BASF and its 39,000 employees in Germany. Because BASF and other chemicals companies sit at the beginning of most industrial supply chains, their disruption would reverberate well beyond the sector, threatening Europe’s economy at a time of high inflation and slowing growth. A throttling of BASF’s ammonia output, a key ingredient in fertilizers, could exacerbate the world’s growing food crisis, analysts say.
    5. Today, dwindling Russian gas supplies are proving a threat to the company’s vast manufacturing hub here—the world’s largest integrated chemical complex spanning some 200 plants. Earlier this month, Russia started throttling back its supply of gas to Germany and other European countries. In response, company executives are doing what was unthinkable just a few months ago: considering how to potentially shut down the complex if gas supplies fall further.
    1. A third, and I’ll just mention it. A third thing, which is very important in manufacturing, is tradable. You don’t have to develop a whole industrial complex. You can import inputs and then export the outputs. You don’t require domestic demand to take off. You don’t require an economy-wide productivity revolution in order to have consumers to whom you can sell your output. You can simply sell it on world markets.
    2. The second thing, which I think is the microeconomics of why it is that manufacturing is so special is that manufacturing, standard, traditional manufacturing like making cars or garments or toys or wigs, has the feature that you can absorb a lot of very unskilled workers into that. Being a production worker, even in an auto factory, certainly in a footwear factory, doesn’t take a whole lot of skill. That means that you can absorb a lot of people off the countryside in a much more rapid way than you could in a lot of other activities where technology and skill are actually highly complementary.
    3. One is that manufacturing technology is much more easily transportable across international borders. It’s much easier to adopt and adapt manufacturing technology. You can take a textile and clothing plant. You still have to tinker with it, but it’s much easier to transplant, unlike many agricultural technologies.
    4. I think what matters here is probably less the quantitative side, how rapidly you’re growing, but the qualitative things that are happening during the growth process. You have to bear in mind that there are archetypal, successful industrialization growth, liberal democracy kind of countries in fact, never experienced the kind of growth rate that Japan and South Korea or China did. In the aftermath of the Industrial Revolution, Great Britain was growing at rates that we would scoff at today for any emerging market. So, it wasn’t the growth itself that enabled the development of the middle class and the emergence and the spread of liberal values and the creating of the liberal democracy, but the transformation of the economy and its social structure in a particular kind of way: if you will, the spread of bourgeois liberal values, the restraints placed on the state in terms of how much it could do and under what kind of circumstances.
    1. If Lithuania had not imposed this blockade when it did, it is not clear the European Union would have forced it to do so. In this sense the blockade was a strategic decision on the part of Lithuania. One view of Lithuania’s strategy is that it is forcing or inducing its allies to affirm their support, at a more rather than less favorable moment in the hostilities.
    2. A nested game is what it sounds like -- a game within a game. It recognizes that the actors in most real-world settings are not unified and have conflicting motives. The classic example is the failure of a proposed Middle East peace deal because some hard-line faction sponsored a terror attack or an assassination. The agreement was never just the two sides dealing with each other; each side was also dealing with its own internal conflicts.
    1. Moscow envisions sending up to an additional 50 billion cubic meters annually to China through a second pipeline that is being negotiated. Russia’s need to pivot to Asia is likely to give it a weaker hand in those negotiations.
    2. During Mr. Putin’s visit to China this year, China agreed to buy another 10 billion cubic meters of gas annually from Russia.Even so, those numbers are dwarfed by the 155 billion cubic meters of gas that the European Union bought from Russia in 2021—representing 40% of its gas consumption.
    3. China won’t completely stop buying LNG from the U.S. Most of its purchases of U.S. LNG have until now come from the spot market. Going forward, more will come from fixed contracts that are just starting to kick in. Even as fighting rages in Ukraine, some Chinese firms have continued negotiating such deals with U.S. producers.
    4. For now, U.S. LNG exporters are protected from the Chinese shortfall by the plentiful demand from Europe. Down the line, competition from Russia for China’s market poses potentially big complications for billions of dollars of planned LNG infrastructure on the U.S. Gulf Coast, investments that assume China will be a huge buyer of America’s gas for many years to come.
    5. The changes haven’t been limited to gas markets. China’s imports of Russian oil surged 55% in May compared with a year earlier, with Russia overtaking Saudi Arabia as China’s top supplier. Russian oil has been selling at a steep discount as some countries have cut imports because of the war.
    6. Between February and April, China’s imports of LNG from the U.S. dwindled by 95% from the same time a year earlier, Chinese customs data showed, while its buying of Russian LNG grew by 50%. Data for May indicated a moderate rebound in China’s demand for America’s LNG, but it was still far below last year’s levels.
    1. For months, Sri Lanka has lacked the foreign currency to buy all that it needs from abroad. Shortages of food and fuel have caused prices to soar. Inflation is now running at 30%.
    2. Sri Lanka owes $50bn (£40bn) to foreign creditors but says it cannot pay them. It is asking for a loan from the International Monetary Fund (IMF).
    1. The European Union announced a ban on insurance for ships carrying Russian oil, which will come into effect in December. About 80% of ships carrying the crude to Indian ports belong to EU shipowners, according to analysis from the Helsinki-based Centre for Research on Energy and Clean Air.
    2. India’s decision is a commercial one: The price of Russian crude tumbled after the Ukraine invasion, with a popular grade known as Urals falling as low as $37 below the Brent benchmark. It edged up in recent days to a discount of just below $34 by Monday, signaling a recovery in demand, according to analysts. 
    3. India has increased imports of Russian oil by more than 25-fold since the start of the war, buying an average of 1 million barrels a day in June, compared with 30,000 in February, according to Kpler data. That is equal to more than a quarter of Europe’s imports of Russian crude and crude products, according to International Energy Agency data. 
    1. When creating your Anki cards, try to find an image for most of your cards. It doesn’t always have to be exactly related to the card, either. If the topic of the card makes you think of something unrelated, find an image of it. While this may take upfront time and investment, over the long term it greatly reduces your learning time. Be generous with inserting images into Anki. Go on Google images, search for something relevant, and quickly copy/paste it or screenshot into Anki. Knowing your shortcuts will save you loads of time here.
    2. Your memory is much more efficient at retaining visual than text based information. It makes sense – we’ve evolved over millions of years and only had written language for a small percentage of our existence.
    3. If you’re finding it difficult to stick with the minimum information principle, then Cloze deletions are a great tool to break your bad habits. They’re also incredibly efficient to create, as you can copy text from your powerpoint or notes and create Cloze cards in just seconds.
    4. The Minimum Information Principle reminds us that simple is easy, and that simple cards are easier to review and schedule. Consider this. If a single card has two sub-items, you need to keep repeating the card to keep the more difficult item in your memory. However, if you split this single card into two separate cards, each can be repeated at their own pace, ultimately saving you time in the long term.This is arguably the biggest offense of most students when they first begin using Anki. Most students make very complex cards that could be broken down into a dozen or more sub-items. I was guilty of doing this for much of medical school.
    5. A surprising number of students succumb to the mistake of trying to memorize something that they don’t comprehend. There is little utility in memorizing a string of information if you are not able to adequately conceptualize and place it within a mental scaffolding.
    1. The companies in the sector don’t get rewarded by investors for reinvesting in growth anymore; they get rewarded by strengthening their balance sheets, being disciplined with drilling, producing free cash flow, and returning a lot of that cash to shareholders.
    1. To summarize the scale of the global ex-USA dollar network, foreigners own $39 trillion in U.S. assets and have $12 trillion in dollar-denominated debts, although the asset-holders and debtors are often not the same regions (e.g. Switzerland is a big asset holder with a surplus, and Turkey is a big debtor with a shortage). Meanwhile, the U.S. is running annual fiscal deficits of well over $1 trillion, and is exporting about $600 billion in dollars annually via its trade deficit, or $3 trillion gross from the U.S. import figures alone.