organicguy

@gregisenberg Usually love your posts Greg but this one appears to be missing context. Skills are like threads and agents are like processes. There are many reasons to use each of them and it is important to know when and how to use each one. This post misses the mark on much of that detail.

this video is the CLEAREST explanation of how claude skills + AI agents work and how to use them most people set up an AI agent and wonder why it keeps disappointing them. the context window is everything context is what the model assembles before it takes any action. think of it like everything the agent needs to read before it does anything. the quality of what goes in determines the quality of what comes out. the models are genuinely really good right now. claude and gpt are exceptional. the variable is almost always the context you give them. 1. agent.md files are mostly unnecessary every single line you put in an agent.md file gets added to every single conversation you have with your agent. a 1000 line file is around 7000 tokens burning on every run. the model already knows to use react. it can read your codebase. save the agent.md for proprietary information specific to your company that the model genuinely cannot know on its own. 2. skills are the actual unlock a skill.md file works differently. what loads into context is only the name and description, around 50 tokens. the full instructions only appear when the agent recognizes it needs that skill. so instead of 7000 tokens on every run you have 50. and the agent stays sharp because the context window stays lean. the closer you get to filling the context window the worse the agent performs, same way you perform worse when someone dumps 10 things on you at once. 3. here is how to actually build a skill the right way most people identify a workflow and immediately try to write the skill. what you want to do instead is run the workflow by hand with the agent first. walk it through every single step. tell it what to check, what good looks like, what bad looks like. correct it in real time. once you have had a full successful run from start to finish, tell the agent to review everything it just did and write the skill itself. it writes a better skill than you will because it has the full context of what actually worked in practice not in theory. 4. recursively building skills is how you go from frustrated to reliable when the skill breaks, and it will break, ask the agent exactly why it failed. it will tell you specifically what went wrong. fix it together in that same conversation. then tell it to update the skill file so that failure mode never happens again. ross mike did this five times with his youtube report generator. it now pulls from eight different data sources and runs flawlessly every single time without him touching it. 5. sub agents are something you earn not something you set up on day one start with one agent. build one workflow. turn it into one skill. once that works add another. ross mike has five sub agents now covering marketing, business, personal and more. it took months to get there and every single one exists because a workflow proved it deserved to exist. the people who set up 15 sub agents on day one and wonder why nothing works skipped all the steps that make the thing actually run. 6. your workflow is the thing the model cannot get anywhere else the model has been trained on everything. it knows more than you about most things. what it does not have is your specific process, your taste, your way of doing things. that is what skills capture. that is what makes your agent actually useful versus a generic one. downloading someone else's skill means downloading their context onto your setup and it will not work the way you want it to because it was never built around how you work. this is the clearest explanation of how agents actually work i have heard. @rasmic runs this stuff every single day and the results show it. full episode is now live on @startupideaspod where you get your pods people charge for this sorta stuff i give away the sauce for free i just want you to win watch

@wildpinesai @a16z No. "Just describe it" with LLMs in a world of AI is like clicking through a GUI when you could automate/program it in the world pre AI. Anyone can formulate words in AI services that produce output. It takes a different level of understanding to create efficient workflows in AI.

Steven Sinofsky on why it's hard for AI to diffuse through firms: "Algorithmic thinking is really, really, really hard for the vast majority of people who have jobs… If you were to go into any person and ask them to create a flow chart for a particular thing that they have to go do, they would probably fail at producing that flow chart." "So within any organization, say doing a marketing plan… one person probably understands and could document the flow chart. So if you put one of these agents or this coworking tool in front of people… their ability to explain to it what to do is really, really limited." "You're basically just developing the next abstraction layer for how people interact… at each level of the abstraction layer, [it's] been a highly skilled, very specific individual within an organization… and then the little parts they build become little toollets… and some people can stitch together and some can't." @stevesi

@TechCrunch @gmoneyNFT Sounds familiar…

Former Tesla product manager wants to make luxury goods impossible to fake, starting with a chip techcrunch.com/2026/02/10/for…

@NVIDIAGeForce GeForce Day

🟢 GEFORCE DAY IS BACK 🟢 To celebrate, we're giving away TWO GeForce RTX 5080 Founders Edition GPUs, signed by NVIDIA CEO Jensen Huang. Want one? Comment "GeForce Day" for a chance to WIN & stay tuned for more!

@thecolossalcave Kings Quest I for the IBM PCjr in 1984, on a cartridge, not a disk, distributed by IBM, with a coloured keyboard template overlay showing how to use the keybaord. I played it every chance I could on our PCjr, setup in a corner of our dining room. My gateway to all things Sierra.

💾 One image. Endless memories. If you had to choose just one… what’s your all-time favorite Sierra game? 🎯 Let’s hear your pick and why it meant so much. Bonus points for telling us where you first played it! 📣 Encourage tagging a fellow retro gamer. #SierraGames #RetroGaming #AdventureGames #ColossalCave #PointAndClick #GamingMemories

@gregisenberg Of course, there I had a typo, but in places where I can edit such things after the fact I do (and would if I paid for the tier on Twitter that allows edits), yet mistakes are still made from time to time.

@gregisenberg Not necessarily. I think it more of a generational difference. Many people from older generations try harder to use proper punctuation and grammar. Those things seem to matter less to younger generations. It’s not a one-size-fits-all scenario, but that’s how I see it.

LLMs made typos go from "you're unprofessional" to "proof you're human" in just 24 months perfect grammar now signals AI (esp with emojis), mistakes and lowercase signall authenticity and humanity

@gregisenberg Minor correction: millions of people pay thousands for training or hire specialists because the software is "crappy."

claude code might be the most IMPORTANT piece of software in years and is the BLUEPRINT for the next $100+ in software opportunities. it took the single most terrifying piece of software ever created (the terminal!!) and made it 10x more friendly. if you're a non tech person, you're brutally afraid of the terminal, kinda the final boss of computing. it's a menacing black screen where one wrong character could nuke your entire project. where EVEN experienced developers would triple-check every keystroke. it's like trying to perform surgery with a chainsaw. technically possible, but most people would rather not. claude code changed everything. now you just talk to your computer like a human: "deploy my app to vercel with SSL" and it handles the surgery for you. writes the commands, catches mistakes, fixes errors automatically. suddenly the most intimidating interface on earth became as easy as sending a text. every industry has its own "terminal." intimidating tools that require months of training: 1/photoshop: 47 toolbars, 200 keyboard shortcuts, layers that make no sense blender: looks like a spaceship control panel, crashes if you breathe wrong 2/excel: pivot tables that feel like advanced calculus autocad: where one wrong click deletes 6 hours of work imagine telling photoshop "remove this background and make it look professional." or telling blender "create a spinning logo animation." or telling autocad "design a simple house floor plan." the opportunity is massive because these tools already have huge markets. millions of people pay thousands for training or hire specialists because the software is "too scary." I *think* within 3 years, every major software company will have a "claude mode" - a natural language layer on top of their existing tools. we'll see a complete inversion of software economics. instead of charging more for "advanced features," companies will charge more for "advanced simplicity." the easier it is to use, the more valuable it becomes. this also creates a new category of power users: people who become incredibly productive not because they mastered complex interfaces, but because they mastered talking to software. prompt engineering (or context engineering?) becomes the new keyboard shortcuts. claude code proved you can take the most unapproachable interface ever built and make it feel natural. I think thats why so many of us are obsessed with it.

@thecolossalcave The Nostalgic Adventurer: A big fan of retro adventure games who won't buy them until a big box edition is released. Why? Because I really miss the enjoyment that came with physical big box games I grew up with. This was talked about, but not recently. Is there still a chance?

You’re entering Colossal Cave. Which class are you choosing? Pick one and tell us why below ⛏️👇 #ColossalCave #RetroGaming #FantasyPoll

@VHSDVDBLURAY4K Michelle Yeoh? Seriously, for action, you could only choose four, and you thought of Michelle Yeoh as one of those four? Why?

Who’s on your Mount Rushmore of Action ?

Looks like Y2K25 is wreaking havoc already…

And the award for best website design for customer service goes to...

How fans will look when they hear that there will be 6 volumes for season 5 of @Stranger_Things, released over the years 2025-2030.

Help us vote in Ultima for Museum of Play to join the hall of fame its well deserves it! The grand father of CRPG

How to best sum up the evening... ...and the Oppenheimer goes to...

If you like having a self-managed digital version of your own movie/TV show library like I do, and you want proper support for Dolby Atmos in apps on the Apple TV 4K, please sign this petition: chng.it/YcCzDFYj

@qfgbook Don’t forget: 1. Having to figure out how to beat a game _without_ the solution being given to you on the Internet/in a hint book/etc. 2. Knowing how to load a game from a cassette tape and how to manage files on a cassette tape.

Streamlining my autoexec.bat and config.sys files, also setting my sound card jumpers

@thecolossalcave 👀👀👀... That's great news, I can't wait!

@Poshoholic Colossal Collectors Edition coming in the future that will hit all the check marks in regards to goodies and Big Box

Ordered one. I’m hoping the box has some physical content as well like games from Sierra that I enjoyed in my youth would have included, but at this price point that seems unlikely, which is unfortunate. It feels like a lost opportunity that I would happily have paid more for.

I couldn’t agree more.