Mark Manger

@karpathy This is of course absolute genius. If you're an academic using Zotero, you don't need raw/ you can just directly access the papers there and build your wiki.

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

@ryancbriggs Hard to know with so many confounders…

@MarkManger okay so you’re probably joking but i genuinely believe that exercise made me more conservative. Testosterone?

Top deadlift is now a comfortable 250 lbs x3. It’s up from 185 when I started a few months ago. It feels to me like almost all of this gain is from technique, confidence, and a better ability to harness my existing strength. I’ve gained no weight since I started lifting.

@jbsteinberg IMHO, this is simply because PoliSci has more demanding standards for data sharing. If economists were held to those standards, they’d probably produce research that would be more likely to replicate than PoliSci. Sociology has no real data sharing req’s at all.

Hot take: econ papers are simply trying to do much harder, much more sophisticated things than poli-sci papers. Likely true even if you restrict attention to reduced-form empirical work, but especially if you include quantitative structural work (IO, macro).

@coach7190 @PaulSpacey Yes but you can get the same with futsal

@MarkManger @PaulSpacey Make the simple pass instead of the complicated one. Don’t dribble the ball a lot pass it before you get covered in defenders. Find the triangle. Hit the back door pass. Crash the board(right in front of the goalie box), and look for the rebound shot.

Top 10 players in the world: - All started soccer by 6
- Almost all played no other *organized* sports 
Your 3-sport balancing act is pure cope. Top players don’t care about your feelings - focus on soccer or stay average.

@RonaiChaker Die Nationalmannschaft kommt eh nicht weit, Nagelsmann wird dann entlassen, und 2030 führt Klopp die Jungs zur Weltmeisterschaft. Schade, ginge auch diesen Sommer schon.

Julian Nagelsmann ist als Bundestrainer in dieser Form ungeeignet und gehört entlassen! Nicht, weil er einmal danebenliegt — sondern weil er ausgerechnet bei Deniz Undav zeigt, wie wenig bei ihm am Ende wirklich Leistung zählt. Undav ist aktuell der torgefährlichste deutsche Bundesliga-Stürmer, stand zuletzt bei 18 Ligatoren und hat gegen Ghana nach seiner Einwechslung auch noch das späte Siegtor gemacht. Reuters berichtete dass bereits nach 20 Minuten Fans in Stuttgart lautstark nach Undav verlangten. Erst in der 60 Minute wurde er eingewechselt und erzielte kurz darauf ein Tor. Und was macht Nagelsmann? Er setzt genau so einen Spieler erst auf die Bank und behandelt ihn weiter wie einen Ergänzungsspieler. Das ist kein Zeichen von Stärke, sondern von Fehlurteil. Ein Bundestrainer, der einen formstarken Torjäger trotz solcher Zahlen klein hält, sendet die Botschaft aus, dass nicht Leistung entscheidet, sondern die Eitelkeit des Trainers und seine vorgefertigte Hierarchie. Reuters berichtete vor dem Ghana-Spiel, dass Nagelsmann rotieren wollte und Undav zwar Einsatzzeit bekommen sollte, die entscheidenden Kaderfragen aber weiter bei ihm lagen. Besonders entlarvend ist dabei der Kontrast zu den Fans. Das Stadion hat längst verstanden, was Nagelsmann offenbar nicht sehen will. Wenn die Ränge nach Undav rufen und der Spieler nach seiner Einwechslung auch noch liefert, dann ist das ein öffentlicher Offenbarungseid für den Trainer. Dann sieht ganz Deutschland, dass die Fans mehr Gespür für Form, Wirkung und Fairness haben als der Mann an der Seitenlinie. Wer so offensichtlich an einem Spieler vorbeicoacht, ist als Bundestrainer fehl am Platz. Die Nationalmannschaft ist kein Ort, für solche Trainer. Sie ist der Ort, an dem die Besten spielen müssen. Und wenn ein Trainer selbst bei einem Stürmer mit 18 Bundesligatoren und einem entscheidenden Länderspieltor noch zögert, dann ist das nicht Vorsicht. Dann ist das mangelnde Eignung. Nagelsmann sollte nicht nur kritisiert werden — er sollte ernsthaft infrage gestellt werden. Denn ein Bundestrainer, der nach diesem Tor auch noch die Frechheit besitzt öffentlich Leistung zu relativieren, Fansignale ignoriert und formstarke Spieler klein hält, beschädigt den sportlichen Wettbewerb im Team. Und wer den Wettbewerb beschädigt, ist für dieses Amt nicht der Richtige. Er will UNDAV nicht einmal bei der WM spielen lassen. Habt ihr sowas schon einmal in der Geschichte des deutschen Fußballs erlebt?

@Alan_Couzens But you’d admit that rugby players are fit, right?

This was one of the weirder things about moving to the U.S. - the way you guys equate "strong" and "fit" (due to your obsession with anaerobic power sports). Personally, I wouldn't look at most NFL linebackers & say... "Man, that guy is *fit*!" In Australia, we'd say that about triathletes or Ironman competitors. But, to each their own.

@pbernier10 @JamesPa19472074 @PaulSpacey Exactly what I meant with “hockey is a faster game.” But that benefit is small if our soccer training is effective at training scanning and quick decisions. Hesitant to say this because many coaches say they’re doing that but neglect that kids have to have technique to execute.

playing soccer is playing soccer and playing hockey is hockey. It’s not because you play another sport that you stop practicing the other one. Played hockey til I was 18 and didn’t stop me for going pro. Transference from other sports is not solely based on playing hands or feet. It goes on physical development and other aspects like scanning, passing movements etc

@CumendaCalcio @MaxBeretta10818 @PaulSpacey Italy has a tradition of outstanding goalkeeper training. I don’t think it has much to do with transfer from another sport because what other sport do kids in Italy play?

@MaxBeretta10818 @PaulSpacey @MarkManger So did Costa Rica then but it’s not a tradition like it is for Italy

@IHateSoccerPod Yes, but I would generally be careful with claims that “research shows” when there’s not even a link provided. I can think of lots of reasons why this might be in the data…

This makes too much sense for a lot of people. 95% of the best players that I’ve trained/coached in 22 years were 1 of 2 things. A second generation soccer player. Or A younger sibling of someone who played the game first.

@pbernier10 @PaulSpacey Interestingly, cognitively, yes - hockey is a much faster game. I see that with many kids - hockey players read the soccer game well at a young age. But inevitably, that advantages erodes as the soccer kids learn, the game gets faster, and specialization benefits kick in.

@MarkManger @PaulSpacey What about hockey to soccer ? 😃

@GuruAnaerobic @bo66ie29 @ikeijeh Objectively much of England is poor, but London is among the wealthiest places on earth. Even there, a middle class person makes 2/3 of a comparable Australian or Canadian. I loved living in London, but I simply couldn’t afford it on a salary from one of the world’s best unis…

I've walked around these areas in Liverpool and similar areas in Manchester. Their astonishingly poor and desolate. Streets where 80% of the houses are boarded up or broken down. No trees, no cars, no life. These places have had no attention or money spent on them, left to rot since the 70's. People moved out, areas decayed. Both Liverpool and Manchester herald their city centres but walk a mile outside, this is what you get. I stayed with a friend in Wallasey - the local shopping centre was like stepping into another world - poor, shabby, cheap, depressed, cheap cafes, the overweight underclass, dishevelled old people, mobility scooters... England is supposedly a rich country, but when you travel around it you see the reality of how poor and depressed many places are.

The same street, 70 years apart. Hale Road, Walton, Liverpool. It looked immaculate in the 1950s compared to today. Why did people stop caring about their communities?

@PaulSpacey That’s probably the real reason: if you don’t start early enough, you simply don’t get enough time with the ball. There’s no “window” - it’s just that you need 20,000+ hours of practice distributed over 20 years or more. That’s how the human body acquires really hard skills.

A kid who starts kicking a ball at four has thousands of hours by age ten. The kid who starts at nine is still a beginner a year later. That gap never closes.

@IHateSoccerPod @jonerfootball What would we do with all the armchair coaches who can’t even juggle a ball ten times themselves?

@DavidAstinWalsh In general, it’s absolutely a disservice to higher education to use computers. Aspiring historians should use quill and ink. It builds character and develops beautiful handwriting. That last skill would be completely lost otherwise.

Being slightly tongue-in-cheek here, but it is absolutely a disservice to whatever is left of the ideals of higher education to baldly admit to using AI to write and saying that you encourage your students to do the same.

People like this should be fired.

@JuergenZimmerer @MattHGoldmann „absolute Kommodifizierung“ - dieses Nonsense-Kategorie verrät den postmarxistischen Historiker. Womit wurden denn Sklavenhändler im arabischen Raum für ihre „Waren“ bezahlt?

Mein Post schlägt Wellen, deshalb doch noch einige Worte der Erläuterung. 1. "muslimischer Sklavenhandel" ist längst zum Dogwhistle islamophober, völkischer Gruppen geworden 2. Wissenschaft verschweigt weder muslimischen noch afrikanischen Beitrag zur Sklaverei 3. Wer über musl. Sklaverei spricht, sollte von christlicher und antiker nicht schweigen. 4. Transatlant. Sklavenhandel unterscheidet sich in wesentl. Facetten von obigen Phänomenen: v.a. die absolute Kommodifizierung, 5. Zusammen mit der Dimension ist es deshalb schon gerechtfertigt, sie von anderen Formen der unfreien Arbeit zu unterscheiden. 6. Nicht das Angebot ist für transatlant. Sklavenhandel verantwortlich, sondern die Nachfrage der Europäer. 7. Kampf gegen "arab. Sklavenhändler" diente schon Koloniallegitimation. Hier zeigen sich interessante Kontinuitäten. 7. Kronzeuge der Rechten ist Egon Flaig. Sein Buch wurde von der Wissenschaft zurückgewiesen. Die Argumente der Wiss. sollte man zur Kenntnis nehmen.

@zpaikin What do you know about the experience of unilingual anglos? You speak around 17 languages or so... 😆

Unilingual anglos do not face meaningful discrimination in government. Francophones outside Quebec have it harder than anglophones in Quebec. The protection of French is existential for French Canadians. If we don’t show French Canadians basic empathy, we will lose Canada.

@CleverN4meHere @Kursk1939 @JJ_McCullough @Unmade333 Heavens, no. In Europe, kids are forced to learn two foreign languages. In Germany, first English and then either French or Latin. No teenager would that voluntarily. And still, nearly everyone was able to hold a conversation after two years. Dunno about the kids who chose Latin.

@MarkManger @Kursk1939 @JJ_McCullough @Unmade333 So you learned well because YOU wanted to? Not everyone is a French language nerd, and thus short of osmosis as JJ alluded too, you won’t speak it well if at all

Everything about official bilingualism in Canada is built on myth. People can’t “just learn” a language they don’t need. Canada isn’t “built on” two equally-sized language groups. Bilingualism isn’t “necessary” to function at an elite level. It’s all just Laurentian folklore.

So annoying sth like this gets published that is pure selection bias. If you want to understand what I mean, look at this graph from our paper: Countries contract and are on a declining path when they get an IMF program, not because of it. Evidence on programs is much more mixed.