+2 -2

Artificial Intelligence Models.md

reviewed

··· 1 1 # Large Language Models 2 2 3 3 - - LLM build internal [[Knowledge Graphs]] in their network layers. 3 3 + - LLMs build internal [[Knowledge Graphs]] in their network layers. 4 4 - LLM models shine in the kinds of situations where "good enough is good enough". 5 5 - Classic ML system where humans are designing how the information is organized (feature engineering, linking, graph building) scale poorly ([the bitter lesson](http://www.incompleteideas.net/IncIdeas/BitterLesson.html)). LLMs are able to learn how to organize the information from the data itself. 6 6 - [LLMs may not yet have human-level depth, but they already have vastly superhuman breadth](https://news.ycombinator.com/item?id=42625851). ··· 58 58 - Make the model ask you more questions to refine the ideas. 59 59 - Take advantage of the fact that [redoing work is extremely cheap](https://crawshaw.io/blog/programming-with-llms). 60 60 - If you want to force some "reasoning", ask something like "[is that a good suggestion?](https://news.ycombinator.com/item?id=42894688)" or "propose a variety of suggestions for the problem at hand and their trade-offs". 61 61 - - Add relevant context to the prompt. Context can be external docs, a small pesudocode code example, etc. Adding lots of context can confuse the model, so be careful! 61 61 + - Add relevant context to the prompt. Context can be external docs, a small pseudocode code example, etc. Adding lots of context can confuse the model, so be careful! 62 62 - [Teach the agents to use tools](https://www.anthropic.com/engineering/claude-code-best-practices). 63 63 - Be aware of the "cache" (e.g: never edit files manually during a session) 64 64

+1 -1

Asynchronous Communications.md

reviewed

··· 7 7 - You can get into "Deep [[Focus]]" session without context switching that allows for better [[productivity]]. 8 8 - You can work whenever, since you are not dependent on anyone immediately. 9 9 - You'll have written records of everything. 10 10 - - [Async communication](http://web.archive.org/web/20241006094947/https://protocol.almanac.io/docs/async-work-ezPny9x7Q50QISL4UIUhB3PoURV0lgxP) takes more time but it enable better thinking. Learn to[[Asking Questions|ask better questions]] and [[Writing|write requests]]. 10 10 + - [Async communication](http://web.archive.org/web/20241006094947/https://protocol.almanac.io/docs/async-work-ezPny9x7Q50QISL4UIUhB3PoURV0lgxP) takes more time but it enables better thinking. Learn to [[Asking Questions|ask better questions]] and [[Writing|write requests]]. 11 11 - The 4 components of a [great asynchronous message](https://protocol-labs.gitbook.io/launchpad-curriculum/launchpad-learning-resources/protocol-labs-network/os-stewardship#sync-comms): 12 12 1. Enough information to cover all follow-up questions. 13 13 2. A deadline. When do you need a response by? How urgent is it? Which task is being blocked right now?

+1 -1

Communication.md

reviewed

··· 15 15 - Without going overboard, use a tasteful amount of graphic design (e.g: bolding one key sentence). 16 16 - Break up a giant nuanced block into sections. 17 17 - If something is critical, make it visual. 18 18 - - If you want an answer, you have to[[Asking Questions|ask a question]]. People typically have a lot to say, but they'll volunteer little. 18 18 + - If you want an answer, you have to [[Asking Questions|ask a question]]. People typically have a lot to say, but they'll volunteer little. 19 19 20 20 ## Resources 21 21

+1 -1

Company Knowledge Management.md

reviewed

··· 36 36 - [Decisions (and rationale) must be documented in a durable location. At GitHub they used to say everything should have a URL](https://haacked.com/archive/2020/04/07/introducing-aboard-beta/). That gives the company a **Decision Log**. 37 37 - Important documents like Roadmaps should be easy to discover and people should be able to comment on and have discussions around them. That promotes keeping it up to date. 38 38 - [[Design Docs]] should have an explicit place, also acting as an historical log. 39 39 - - [[Documentation]] should have a common entry point to increase discovery. Each team can have it's own [[processes]] on top. 39 39 + - [[Documentation]] should have a common entry point to increase discovery. Each team can have its own [[processes]] on top. 40 40 - At a company level, each team documents differently. To make the most of that, grow the knowledge organically and locally, not top down. Each team should have its own permissionless and open way of working and should be made public to the rest of the company. That way they can evolve the system to fit their needs. Then, there is a standard protocol to share information between teams. 41 41 - The protocol serves as the team communications API. An abstraction over the inner works of the team that is common to all the other teams. 42 42 - E.g: Each team having a homepage [README](https://tom.preston-werner.com/2010/08/23/readme-driven-development) with links to their slack, ticket system and processes.

+2 -2

Coordination.md

reviewed

··· 33 33 - No [[processes]] requires trusting other people. More trust means better coordination without processes. [Trust is the currency of interactions](https://youtu.be/-vbPXbm8eTw). 34 34 - When a process becomes the proxy for the result you want, you stop looking at outcomes and just make sure you're doing the process right. 35 35 - Every process will slow you down, and some will make you better. 36 36 - - If we imagine human society as it's own organism. We need processes and other coordination tools to make it remove the hand from the fire when it starts to burn. 37 37 - - The hand doesn't know what to do, but relies information to the brain, that makes the appropriate changes. 36 36 + - If we imagine human society as its own organism. We need processes and other coordination tools to make it remove the hand from the fire when it starts to burn. 37 37 + - The hand doesn't know what to do, but relays information to the brain, that makes the appropriate changes. 38 38 - Something similar could be achieved at a society level, where pain triggers processes that make it stop. 39 39 - Only a few bits of information are possible to reliably convey to a large number of people. [The larger the group, the smaller the message needs to be](https://www.lesswrong.com/posts/4ZvJab25tDebB8FGE/you-get-about-five-words). 40 40 - The requirements to govern a commons without tragedy:

+1 -1

Habits.md

reviewed

··· 2 2 3 3 [We, humans, are basically habit machines](https://twitter.com/JamesClear/status/1059504529111158784). We form habits. We run on those habits all day long. Your outcomes are a lagging measure of your habits. How you spend your days is how you spend your life. Every action you take is a vote for the type of person you wish to become. 4 4 5 5 - Habits can be great because they help us get tasks done efficiently without having to spend willpower (a limited resource) on them all the time. The purpose of a habit is to remove that action from self-negotiation. You no longer expend energy deciding whether to do it ([ego depletion](https://en.wikipedia.org/wiki/Ego_depletion)). [Having an habit collapses hundreds of future decisions into one, and gives you focus](https://twitter.com/kepano/status/1698388043278012621). Habits will keep the frontal cortex free to solve other problems. Make a deliberate choice about what needs consistency and what doesn't. 5 5 + Habits can be great because they help us get tasks done efficiently without having to spend willpower (a limited resource) on them all the time. The purpose of a habit is to remove that action from self-negotiation. You no longer expend energy deciding whether to do it ([ego depletion](https://en.wikipedia.org/wiki/Ego_depletion)). [Having a habit collapses hundreds of future decisions into one, and gives you focus](https://twitter.com/kepano/status/1698388043278012621). Habits will keep the frontal cortex free to solve other problems. Make a deliberate choice about what needs consistency and what doesn't. 6 6 7 7 [Make part of your identity to be an agent](https://www.neelnanda.io/blog/become-a-person-who-actually-does-things) that does things. E.g: noticing the small problems, and fixing them. The most effective way to change your habits is to focus not on what you want to achieve, but on who you wish to become. Focus on [[The Four Laws of Behavior Change]]. 8 8

+2 -2

Ideas.md

reviewed

··· 20 20 21 21 ### Hyper Mega Awesome Game 22 22 23 23 - - Modular Approach. The idea is to some kind of state (main character, currency, ...) and multiple shards to play. Each shard could implement a different rule set or genre. 23 23 + - Modular Approach. The idea is to have some kind of state (main character, currency, ...) and multiple shards to play. Each shard could implement a different rule set or genre. 24 24 - [[Modularity]] could also be implemented in the graphic side. You can choose the graphics pack you like just like another cosmetic similar to Rimworld or Dwarf Fortress. 25 25 - Player Driven Economy. Everything is made by players and traded for real life currency. The developers only get a fee for each trade. This makes the game fully F2P but also supports the developers. 26 26 - Companion Apps. Some tasks like trading or [[Planning]] could be done from a mobile device. ··· 96 96 - Similar to Numerai, participants send submissions and stake some amount of money. 97 97 - The best submissions are selected and the money is distributed among the participants depending on their stake. 98 98 - For each task, rewards are given per row/prediction, not per model. This encourages a [[Plurality]] of models to be used that specialize in different tasks. 99 99 - - Similar approach could be done to incentivize the creation of new datasets. An entity holds out some "groud truth" dataset. Participants can submit datasets and stake some amount of money. The closest dataset to the ground truth is selected. There might be dragons (generating fake data that follows the distribution)! 99 99 + - Similar approach could be done to incentivize the creation of new datasets. An entity holds out some "ground truth" dataset. Participants can submit datasets and stake some amount of money. The closest dataset to the ground truth is selected. There might be dragons (generating fake data that follows the distribution)! 100 100 - Pluggable Identity with ENS, DID, ... 101 101 - Infrastructure for [infofinance](https://vitalik.eth.limo/general/2024/11/09/infofinance.html).

+3 -3

Learning.md

reviewed

··· 2 2 3 3 - Learning is a process of [[Conceptual Compression]]. 4 4 - Understand the base and build from it. First Principles Method. Seek the big picture understanding - focus on concepts not details. 5 5 - - Read about topics you care, observe the world around you and keep a beginner's mind (*shoshin*). 5 5 + - Read about topics you care about, observe the world around you and keep a beginner's mind (*shoshin*). 6 6 - Find a mentor if you can and ask them questions. 7 7 - - Don't be afraid to[[Asking Questions|ask a question]] that may sound stupid because 99% of the time everyone else is thinking of the same question and is too embarrassed to ask it. 7 7 + - Don't be afraid to [[Asking Questions|ask a question]] that may sound stupid because 99% of the time everyone else is thinking of the same question and is too embarrassed to ask it. 8 8 - You can apply [The Feynman Technique](https://www.farnamstreetblog.com/2012/04/learn-anything-faster-with-the-feynman-technique/): 9 9 1. Identify the subject. 10 10 2. Teach it to a child. Use simple vocabulary and make it short. Keep questions and answers simple. ··· 15 15 - There are two categories of learning. [Mix them both to learn faster](https://www.joshwcomeau.com/blog/how-to-learn-stuff-quickly/). 16 16 1. **Guided:** Reading a tutorial, taking a course, watching a YouTube video. Anything where you're following a guide. 17 17 2. **Unguided:** Creating your own projects from scratch, extending a tutorial, looking things up in the docs. Anything where you aren't following a guide. 18 18 - - Learn to[[Asking Questions|ask better questions]]. Distill what you know to figure out what part you're missing or which link is missing. 18 18 + - Learn to [[Asking Questions|ask better questions]]. Distill what you know to figure out what part you're missing or which link is missing. 19 19 - Learn by [[Writing]]: 20 20 1. Pick a topic. 21 21 2. Read and/or discuss with others (a bit).

+1 -1

Plurality.md

reviewed

··· 10 10 - [[Governance]] mechanisms should count uncorrelated signals additively but correlated signals with diminishing returns. 11 11 - [[Organizations|Organizations]] should allow different degrees of membership, not just true-or-false. 12 12 - Local currencies and property rights can coexist with global mechanisms for cooperation. 13 13 - - We should understand the world through a patchwork combination of models, and not try to stretch any single model to beyond its natural applicability 13 13 + - We should understand the world through a patchwork combination of models, and not try to stretch any single model beyond its natural applicability 14 14 - There is a set of principled mathematical techniques by which you can design social, political and economic mechanisms that treat not just individuals, but also connections between individuals as a first-class object. 15 15 - We should take connections between individuals really seriously, and work to expand and strengthen healthy connections. 16 16 - Plurality technologies include:

+1 -1

Productivity.md

reviewed

··· 13 13 - [[Journaling|Keep a log]] of what happened each day. You can also add what you've learned! 14 14 - Create [[checklist]] for repetitive processes. For example, a [[checklist]] detailing all the task to do before ending the day. 15 15 - "Where is the good knife?" If you're looking for your good X, it means you have bad Xs. Throw those out. 16 16 - - Your mind should be flexible, but your processes should be repeatables and predictable. 16 16 + - Your mind should be flexible, but your processes should be repeatable and predictable.

+2 -2

Teamwork.md

reviewed

··· 76 76 - [Some experiments won't work](https://www.lesswrong.com/posts/97LgacucCxmyjYiNT/the-archipelago-model-of-community-standards). But oftentimes it _feels_ like it wont work when in fact you just haven't stuck with it long enough for it to bear fruit. This is hard enough for _solo_ experiments. For group experiments, where not just one but _many_ people must all try a thing at once and _get good at it_, all it takes is a little defection to spiral into a mass exodus. 77 77 - The group with the most power determine the system that reflect and reinforce their own way of thinking. Aim for inclusion. _Diversity is being invited to the party. Inclusion is being asked to dance and help organizing the party_. 78 78 - [Brainstorm for questions first (explore). Then find the answers (exploit).](http://web.archive.org/web/20240522210302/https://getpocket.com/explore/item/better-brainstorming) 79 79 - - Strive for constructive conflict. Get people to[[Asking Questions|ask questions]]. Engage in passionate, unfiltered debate about what you need to do to succeed. 79 79 + - Strive for constructive conflict. Get people to [[Asking Questions|ask questions]]. Engage in passionate, unfiltered debate about what you need to do to succeed. 80 80 - Encourage to fail. Failing is good if the team [[Learning|learns]] from it! 81 81 - Encourage effectiveness. Find ways to free up your time. 82 82 - Communication is a central part of working in teams. Prefer [[Asynchronous Communications]], use common [[Communication|communications]] techniques and be friendly. Trust and efficient communication has a big impact on team effectiveness. ··· 124 124 - Keep a [private work log](https://youtu.be/HiF83i1OLOM?list=PLYXaKIsOZBsu3h2SSKEovRn7rGy7wkUAV). It'll make it easier for everyone to advocate what you did. 125 125 - [Don't sabotage the team](https://erikbern.com/2023/12/13/simple-sabotage-for-software)! 126 126 - [Nobody gets credit for fixing problems that never happened](https://news.ycombinator.com/item?id=39472693). People get credit for shipping things. Figure out how to reward and recognize people for preventing problems. 127 127 - - The same practices that make great [[Artificial Intelligence Models]] [promts](https://platform.openai.com/docs/guides/prompt-engineering) also make [great practices with humans](https://x.com/tayloramurphy/status/1849269205155123568): 127 127 + - The same practices that make great [[Artificial Intelligence Models]] [prompts](https://platform.openai.com/docs/guides/prompt-engineering) also make [great practices with humans](https://x.com/tayloramurphy/status/1849269205155123568): 128 128 - Give clear instructions. 129 129 - Share relevant background info. 130 130 - Break big problems into chunks.