22 Matching Annotations
  1. Last 7 days
    1. Independent research has found that some models write weaker code depending on who it's for. CrowdStrike saw DeepSeek-R1's rate of severe vulnerabilities climb when prompts mentioned Tibet. Booz Allen found three of four Chinese coding models wrote more vulnerable code for a government customer, with Qwen3-Coder roughly 130 percent worse. Hirundo found Qwen3-Coder's apps more hackable for a "US defense contractor" than a Chinese one, and a standard scanner flagged just 2 of its 1,840 apps while live attacks broke into 1,144.

      Adverse open weight AI .... or problems with open weight AI

    1. Gaming used to be done in arcades for 25c a play*.Over time NeoGeo realized they could sell you the hardware to play at home (“on-prem”) at arcade quality albeit at a price out of reach to most people. Alongside them, Nintendo and Atari gave you 80% of the fun at 20% of the cost. History will view games arcades as a flash in the pan compared to gaming at home.Namco, Capcom, and Konami went into selling software instead of hardware but it feels hard to imagine buying a home copy of Fable** for $X,000 any time soon.⸻* That’s for video games. In fairground arcades you’d literally pay for tokens then use tokens for play. The analogy almost writes itself.** The model, not the video game.

      How models will evolve eventually to be run at home as the cost of the hardware decreases a lot over many many years

    1. My dad was a newspaper reporter in the early 2000s, so there were several dinner table conversations about this dilemma.The buyouts were taken by superstars and low performers. The superstars moved on to other things, like working in other forms of media.Dad didn't take several rounds of buyouts and was fine.After suffering a stroke, he was "put out to pasture", which for a journalist means he was sent to cover city hall meetings in some suburb. Eventually, he retired with some much smaller payout than what the last buyout would have given him. The sense in the family was that the collapse of the industry denied him a dignified last decade of work. (When the industry was flush with cash, older journalists were sort of like "elder statesmen" in the newsroom. Post-internet, no longer.)

      how jobs change in prestige

  2. Jan 2026
    1. Imagine if we could bring world-class healthcare, education, finance and creative services to anyone for almost zero cost. People able to afford these services don’t doomscroll looking for more stuff. They goal seek to flourish more.

      Not only is production at a low cost but distribution is now at the point and available hyper targeted.

    1. Tempo* launches testnet

      Which is a fast internet scale payment focused decentralized blockchain. So again, it's a rail infra for any stable coin - and new companies can start creating their own customer support systems and regulation systems around it. Not easy because of regulations and laws, but it makes the infra of money movement easy since you move it within the wallet. And since you move it on a recognized and trusted blockchain, customers already have a sense of trust with you.

    2. The Year of the Fintech Compounder0102030405FinTech �TrendsAgentic CommerceStablecoins BanksNeobanks06070908PaymentsPrediction MarketsOpen FinanceConsumer �Health10WrapThe big names keep getting bigger. Fintech hyperscalers had a year, with Nubank over 127m customers, Klarna at 114m and Revolut snapping at their heels at 65m. The banks have their mojo back. The IPO market is open. And yet there’s a lingering sense of doom.This was the year of Agentic Commerce, Stablecoins, and prediction markets. Trends that seem destined to reshape the industry, but are still in their early development. Agentic commercce volume is small. Stablecoins are still cross border, and prediction markets are mostly sports. Will that all change?ST

      Also available here:

      https://web.archive.org/web/20260108011954/https://docs.google.com/presentation/d/e/2PACX-1vRKyMF0WR4uY9OtvfnnWGlZZoAyMgDMJjRBOCr7-WN3PcOWDusklFbXZYo_bAD6qiw02_eJN3J6hdZd/htmlpresent

    1. Here, the authors extend the argument to a more speculative level. If AI stabilizes societies and extends lifespans (or even immortality), there will be fewer “shocks” (wars, deaths, disasters) that traditionally reset inequality. Without those resets or widespread redistribution, the existing wealth concentration could persist and deepen indefinitely. Hence, they argue that a global tax on capital might be the only effective check on rising inequality in a fully automated, AI-driven world.

      Because labour is free and the only thing now remaining is capital.

  3. Oct 2025
    1. I've seen startups killed because of one or two "influential" programmers deciding they need to start architecturing the project for 1000TPS and 10K daily users, as "that's the proper way to build scalable software", while the project itself hasn't even found product-market fit yet and barely has users. Inevitably, the project needs to make a drastic change which now is so painful to do because it no longer fits the perfect vision the lead(s) had.Cue programmers blaming the product team for "always changing their mind" as they discover what users actually need, and the product team blaming developers for being hesitant to do changes, and when programmers agree, it takes a long time to undo the perfect architecture they've spent weeks fine-tuning against some imaginary future user-base.

      In the initial stages you want to be nimble. Scalable and complicated architecture makes that difficult.

      But you also need for burst patterns as explained in one of the sub-comments.

  4. Oct 2023
    1. perfect compatibility is impossible.

      'is impossible with cambria' This is especially true with mapping where data is lost between translations - because then more data is required to do a reverse translation, and cambria doesn't have that support

    2. Users should be able to collaborate on a document regardless of the software version they’re running. New clients should be able to open documents created by old clients, and vice versa.

      So this method is mainly for evolving schema so that clients on different versions are still able to communicate (without optional parameters)

  5. Nov 2022
  6. May 2022
    1. AnthonBerg 16 hours ago | parent | next [–] Sunlight’s effects on body chemistry also have a damping effect on inflammation. It’s a funny thing.Sunlight has an impact on many chemical processes in and on the body. One example: Our skin is coated with a bunch of chemicals. There’s a number of different acidic chemicals. These are known as the “acid mantle”: https://en.wikipedia.org/wiki/Acid_mantleOne of these substances is urocanic acid. The body manufactures it as the trans-isoform. UV light – as in sunlight – causes it to change into cis-urocanic acid. (Ultraviolet-induced isomerization.)Cis-urocanic acid fits a certain serotonin receptor – type 5-HT2A. Cis-urocanic acid is a 5-HT2A agonist. That receptor is known to be profoundly immunomodulatory.It’s super interesting! There are some recent papers on it like “ Cis-urocanic acid, a sunlight-induced immunosuppressive factor, activates immune suppression via the 5-HT2A receptor” – https://pubmed.ncbi.nlm.nih.gov/17085585/“Molecular basis for cis-urocanic acid as a 5-HT2A receptor agonist” – https://www.sciencedirect.com/science/article/pii/S0960894X0...This is absolutely fascinating. I think people take me as a crackpot when I point to these papers. These are just simple and solid papers from molecular biology :)
  7. Nov 2020
    1. Micro is a platform for cloud native development. It addresses the key requirements for building services in the cloud. Micro leverages the microservices architecture pattern and provides a set of services which act as the building blocks of a platform. Micro deals with the complexity of distributed systems and provides simpler programmable abstractions to build on.

      What they are doing is like AWS lambda - BUT - they abstract away the details of things like Auth, pub/sub, config, networks AND the DB (which is a KV store).

      So you simply write your services letting the platform take care of the particulars of adjacencies.

  8. Jun 2020
  9. Apr 2020
    1. ake Wikis as an example. Most of them have two different modes: The reading mode. The editing mode. The reading mode is the default. But most of the time you should create, edit and re-edit the content.

      Most of the time you should create edit and re-edit

    1. There are two methods to work with a Zettelkasten: Indirectional Directional Indirectional work means that you just feed your Zettelkasten. It is hungry and demands for notes a.k.a. Zettels. The more you feed it, the more complex it becomes, and the more it will actually talk to you. For additional information read: Create a Zettelkasten for your Notes to Improve Thinking and Writing Directional Work means that you are working on a text, a book or an article, and feeding your Zettelkasten with notes directly and intentionally related to this book.

      Great insight into how to use the zettle

    1. The Zettelkasten as a septic tank - don’t put in only filtered notes. Delay examining and deciding - and a question of speed.(Zettel 9/8a2)

      This is like org-capture

    2. Later, he wrote every distinct point on a different Zettel and decided individually for every thought where to put it.

      And this is where he entered each point into the zettle.

    3. He created brief excerpts for a book and a list of thoughts with references to the pages in the book.

      Not yet in the zettle. Just notes while reading the book