Loading video...

Video Failed to Load

Go Home

system-atlas: a skill that builds an explorable isometric map of your codebase / agent / pipeline from one data file arranging the system visually does something for my brain that a wall of text doesn't i've found it way more useful to build up understanding of a system or...

165,461 views • 28 days ago •via X (Twitter)

38 Comments

harsh's profile picture
harsh28 days ago

`npx skills add inkboard/system-atlas`

Justin Reidy's profile picture
Justin Reidy28 days ago

This is dope thanks for sharing. We’re increasingly using these kinds of visual support skills at No Diet (initially soured by @humanlayer_dev’s show me skill). Seeing really helps my understanding.

harsh's profile picture
harsh28 days ago

@humanlayer_dev 100% - i've been trying recently to read the code again but honestly w/ the pace at which these things sling code it's a losing battle. so at least understanding the basic topology is the minimum. Kind of ironic that we're all back to being eng managers again 😂

wesley⚡️'s profile picture
wesley⚡️26 days ago

ELITE - our team gave it a go and are going to keep using your work. congrats. we did find a lil issue in the HTML template if you want to check it out:

harsh's profile picture
harsh26 days ago

Tysm! I’ll take a look

wesley⚡️'s profile picture
wesley⚡️26 days ago

worked out really well fwiw

Fried Bun's profile picture
Fried Bun27 days ago

Inspiration from this

Matija Sosic's profile picture
Matija Sosic27 days ago

love it! we're working on something similar for Wasp, but it's coming from the compile-time spec. Will be applying much nicer color theme to it though :)

Henry Mascot's profile picture
Henry Mascot27 days ago

wicked!

harsh's profile picture
harsh27 days ago

thanks!

Matthew Bilo's profile picture
Matthew Bilo27 days ago

Super cool! Curious how you simulated the data flows? Did you just give examples yourself or is that from a live local dev server?

harsh's profile picture
harsh27 days ago

The agent executing the skill picks it up from it’s explorations

symbols's profile picture
symbols27 days ago

@deepok102

L’Écho du Corbeau's profile picture
L’Écho du Corbeau27 days ago

I love it

Stefan's profile picture
Stefan27 days ago

This is very, very cool. I tried building something broadly similar to this but found the diagrams the agents made weren’t entirely accurate. Do you find this skill to be accurate enough?

harsh's profile picture
harsh27 days ago

At least works on the scales that I’ve been testing with. Not sure how it’ll perform at bugger scales

maz's profile picture
maz27 days ago

love this! how did you give the visual animation?

harsh's profile picture
harsh27 days ago

just claude doing claude things :)

Yel's profile picture
Yel27 days ago

Please change the name. Everyone I work with has some kind of 'atlas' or 'vortex' or 'forge' and they all do different things. Don't go with the AI naming, for all of our sakes.

harsh's profile picture
harsh27 days ago

i think it might be a little too late now 😅 - i agree w/ you though

Carlos Rene's profile picture
Carlos Rene27 days ago

I see it in 2D but yes, watching the process and the code is the best way to work

Winter's profile picture
Winter27 days ago

Personally I prefer 2d, so I've build an open source local tool for that called HQ flow! Try it out at

esp's profile picture
esp27 days ago

Very good work

Gᙇ's profile picture
Gᙇ27 days ago

ahh this is gonna change the flow 👍🏾 cool

ThuNoJoy | SayTXT for ADHD Reading's profile picture
ThuNoJoy | SayTXT for ADHD Reading27 days ago

This is the same reason audio-first reading helps some people. The content is not always the problem; the entry point is.

rusa's profile picture
rusa27 days ago

the spatial thing is real, text has no persistent location so your brain re-parses it every read, a map gives each module a fixed address you recall. the hard part is layout stability, if it reflows as the code changes, the map you memorized lies to you next week.

Roman Hauksson-Neill's profile picture
Roman Hauksson-Neill26 days ago

@cosmic_yolo_bot

~/.shane's profile picture
~/.shane26 days ago

this is awesome

Jordan Lee's profile picture
Jordan Lee27 days ago

That interactive codebase map is actually sick. Being able to click through the architecture, inspect payloads, and watch the system run is way better than staring at docs.

อัดนาน มามะ's profile picture
อัดนาน มามะ27 days ago

Evaluating quantum computing requires comparing it with alternative approaches under realistic constraints in emerging economies.

harsh's profile picture
harsh27 days ago

you got it man

DxD's profile picture
DxD27 days ago

I built a similar tool using nodes instead of isometric called RPM, just runs local on all my repos I love this! going to see if I can add this as a toggle feature to RPM

Mark Hetherington's profile picture
Mark Hetherington27 days ago

Does it visualize the diff?

Pierson Davis's profile picture
Pierson Davis26 days ago

so cool, just starting to play around with it

FiberMoose's profile picture
FiberMoose27 days ago

Gives Jurassic Park computer scene vibes (SGI unix/IRIX FSN)

Le Psychopompe 🇪🇺💸's profile picture
Le Psychopompe 🇪🇺💸27 days ago

It cannot work at scale because software is way more highly dimensional that the 3 dimensions that our visual cortex can process.

Alex Araque's profile picture
Alex Araque26 days ago

This is super super cool, my brain is having trouble keeping up with complexity and ended up literally drawing squares in my notebook. Thank you!

AI~rtesan's profile picture
AI~rtesan27 days ago

👏🏻

Related Videos

how to use Google's NEW open source Design.md + AI Skills to make your startup look like a $100 million company in 1 hour: 1. Design.md is an open source file from Google that captures the soul of a design. Typography, colors, spacing, all in one markdown file. You attach it to your prompt and your agent builds beautiful things every time. 2. Think of it this way. The HTML is the finished dish. The design.md is the recipe. The skills are the ingredients. Put them together and everything you build looks consistent and professional. 3. Don't create a design system from scratch. Find a brand you love. Linear, Stripe, Vercel, whatever resonates. Study it. Use ChatGPT or Claude to help you extract the design language into your own design.md file. 4. Build skills on top of your design.md. A landing page skill. A mobile app skill. A motion design skill. A slide deck skill. Each one references the same design.md so everything looks like it came from the same designer. 5. The biggest mistake people make: they nail one screen and then everything else looks generic. Design.md solves this. One file keeps every page, every format, every medium consistent. 6. Use it across everything. Your landing page. Your app. Your pitch deck. Your promo videos. Same DNA. Same taste. Same system. That's what separates a startup that looks real from one that looks vibe-coded. 7. Build a second brain for design inspiration. When you see something beautiful in the real world or online, capture it. Save it. When you're building something new, reference it. Taste is developed, not downloaded. 8. It's obvious but the difference between a product people trust and a product people bounce from is how it looks and feels. Design.md gives you that edge. you can watch below shoutout to Meng To for coming on The Startup Ideas Podcast (SIP) 🧃 and walking through his full workflow. if you want to use AI to actually build gorgeous designs, you'll want to use see this. watch

GREG ISENBERG

512,753 views • 4 months ago

Buttigieg: It is not too much to ask that, in this country and in this state, one job ought to be enough, which means wages and benefits that make that possible. It’s not too much to ask that you ought to be able to afford to raise a family in this country—to put your kid in child care if you want, or stay home and take care of them yourselves if you want, or do some combination of that if you want—and still have it add up at the end of the month. It is not too much to ask that the wealthiest country in the world have the best health care in the world and that it have the best-funded public schools in the world. And it is not too much to ask that the largest corporations and wealthiest people in this country pay at least as much of their income and wealth in taxes as a schoolteacher, a bus driver, or a firefighter. It is not too much to ask for a different kind of politics, where you look at your leaders in action and actually feel your blood pressure come down a little bit instead of going through the roof. That is a reasonable thing to expect of our leaders. It is not too much to ask. We can do this. Most of us already want this. So the obvious question is: Why hasn’t it happened? And that’s where we’ve got to talk about the system that we’re living with, which, by the way, as a politician, they kind of tell you not to do. Consultants say, “Look, no one’s going to get too fired up about the details of our democratic system when they’re just trying to keep their heads above water in this economy,” which I get, and to some extent I agree. But the reason that so many things are wrong with our economic picture is that so many things are wrong with our political system. And we’ve gotten so used to it, we can forget that there was a time when our political system was the pride of this nation. Wars have been fought for our system, starting with the first one—the one that launched 250 years ago with the Declaration of Independence, the one that gave this nation its being. The Founders risked their lives for a better political system. So don’t tell me it can’t stir people’s hearts. Don’t tell me that it’s not worth fighting for. Especially because until we do something to change our political system, we will continue to suffer from an economic system that lets us down: monopolies squeezing families and farmers on both ends; consumers and workers caught up in a system that can’t serve them; small businesses swamped by the big guys. Right now, what’s at stake is nothing less than the American dream. It is endangered.

Acyn

42,103 views • 2 months ago

Vallée and the Closed System: Are We Prisoners? 🧠👽 Vallée: "Are We Being Taken Over by a Species from Somewhere in Space That's Vastly More Intelligent Than We Are?" 👽🧠 "..the simulation...was a new concept that I initially rejected." ~Vallée "Is it looking for us to try to interact with it as equals or with parity?" ~Scafish "Nobody says that to Congress, and I think Congress should hear it." ~Vallée If, "it's a closed system, we're like prisoners and something is going to happen to us, and there is very little we can do." ~Vallée Turn the thermostat dial. "If the temperature doesn't change, then I know I'm inside a control system. So, we can do the same thing with UFOs, but we have to react. We have to, number one, acknowledge that it exists, and number two, we have to react to it." ~Vallée ~~~ I've been wanting to share this one for a while... Vallée: "So, he said, the question you have to ask about UFOs is, number one, is it a natural system or an artificial system...control system. And if it is a control system, is it open or closed? In other words, are we being taken over by a species from somewhere in space that's vastly more intelligent than we are?" (I've never heard him even suggest that possibility.) Vallée: "You know, as Dr. Garry P. Nolan says, you know, people who have had ten, you know, scientific revolutions, or a hundred or a thousand, and come here with superior science to do something... And in which case, you know, it's a closed system, we're like prisoners and something is going to happen to us, and there is very little we can do. Or, is it an open system where we can, in fact, communicate with it. And if we can communicate with it, then the question for me as an information scientist is, what are the modalities of the interaction, you know? It's not just can we learn their language? And they say, you know, 'We come in peace to save mankind,' or something. Or 'We will give you the cure for cancer' or something. I don't think it's at that level." (Will we ever be able to get answers to these extremely important questions? If it's an open system, how do we communicate with it? How do we provoke it to react? We know it reacts to anything nuclear but we still don't know why. This is why we need the USG (and other governments) to present evidence that shows the masses this is real and extremely important for our species to investigate. If that evidence exists and is shown, we'll have an easier time getting the world's best minds to join the effort in figuring out the best way to answer these questions. We still may fail but we should at least try.) Vallée: "I think it's a meta-system. It's not a system. And that's my fear...if we can circle back to your earlier question about, you know, about NIDS and about BAASS, what we did for the government and what we did for the Defense Intelligence Agency. Half of the budget was spent developing, you know, a super database. And we don't know where it went. I mean, I'm not cleared to know where it went." (On the contractor (BAASS) side, Bigelow should have all copies of what AAWSAP produced. And on the DIA side, Lacatksi said he put all of the digital files in a specific place that he didn't name. As long as someone didn't delete it all, it should still be there. Vallée has said that the Capella database has about 250,000 cases from around the world.) Vallée: "But that would be a very interesting question, because the people who are getting [the database] are getting raw data, which we have very well organized, all in English. So they have the luxury of, you know, we had five translators from French, English, Portuguese, Russian, you know, everything was translated in a single structure across fourteen databases. "That's what we need to answer the question about the control system, and it's not being done. And we hired a whole team that we had trained to work on it. So to rebuild that will take the next ten or fifteen years. And nobody says that to Congress, and I think Congress should hear it, because it's our money." (As long as names and personal details are scrubbed from that database, there is no reason NOT to release it to the public. This way, we can take it and use AI to help decipher patterns and maybe answer some of these questions. Can Congress help us get access to that database?) Vallée: "When you ask, is it a control system? That's a big question." Peter Scafish Peter Skafish - "You asked, at one point, whether the system is open or closed, and you said, additionally, I believe, if it's open, that it would be possible to communicate back to it. And it sounds to me like that's the key question for you. Is, if you can understand what what the system of symbols is, or the modalities of communication, then you can understand enough to engage in some kind of communication, or at least give some kind of response to show that you understand." Vallée: "Yes." Scafish: "So then the question, and we have a member named Jacqueline, who has asked this. Could the system be stimulating us - provided there is such a system - to interact with it, more as subjects or agents than as something like animals or objects? Is it looking for us to try to interact with it as equals or with parity?" (When people report getting injured or sick from being in close proximity to UAP, it suggests to me that the phenomenon won't go out of its way to avoid affecting us in a negative manner if we get in their its/way (assuming it even knows that close encounters with UAP are not good for humans). Kind of like how we treat lower lifeforms. If we encounter a wild rabbit crossing the road, many of us will do our best to avoid it, but not if it means damaging our car or ourselves in an accident. It may ruin our day if we hit it, but it won't stop us from driving again in the future. Do NHI have bad days if their tech injures us or makes us sick? I have no clue.) Vallée: "Well, what I saw in the notes you gave me, is she was also asking: Is it a control system because we think it is? And that's a very interesting question. Because we react to the UFO phenomenon, or the UAP phenomenon. And, you know, at this point when I think about what I'm going to do next in this research, if I'm given the the opportunity to live a little longer, I'm not going to go back and write any more computer programs. There are better people to do that now, they have the data, and we're in a different phase now. We're in a whole different system. I have the luxury of doing some experiments I wanted to do for a long time." (Would have liked Scafish to ask him: What types of experiments?) Vallée: "So, if you think you are inside the control system, there are things that you can do. Or, if you think you're inside the simulation, you know, which was a new concept that I initially rejected, and then, you know, Ray's (Kurzweil?) work and others have brought it back to the forefront. And we have to ask that at the same time. Can we test it? How would you test it? Well, if, you know, I'm here in my apartment, and the temperature is constant in this apartment. But outside, I can see it's cold, or I can see the sun is out and it's warm, and how come it's constant here? So this would lead me to think that there is a control system, namely a thermostat, that is somewhere. "So I can start looking around the walls, and if I see dial, I can turn it, or I could start a fire and see what happens, see if the temperature changes. If the temperature doesn't change, then I know I'm inside a control system. So, we can do the same thing with UFOs, but we have to react. We have to, number one, acknowledge that it exists, and number two, we have to react to it."

Joe Murgia

59,089 views • 9 months ago

I still don’t understand why everyone isn’t running one of these yet. mine has been compiling for six months and it already knows things I’ve forgotten I knew Andrej Karpathy published a folder structure in April 2026. 5,000 stars. 16 million views. engineers couldn’t stop sharing it. not a model, not a tool, a folder structure. the sentence that explains why: “RAG re-derives knowledge on every query. a compiled wiki derives it once and keeps it current” most people who try building one hit a wall around month three anyway. the notes pile up, the links exist, and then it just sits there. you’re still the one doing all the remembering here’s the sentence that explains that wall too: keeping fifty interlinked notes current, consistent, and cross-referenced is work no human sustains. so people stop maintaining it, and it turns into a graveyard of good intentions the fix isn’t discipline. it’s making the system do the maintenance instead of you here’s the gist drop any source into a folder called raw, untouched, ground truth Claude reads it once, writes a structured page into wiki, links it to everything already there, flags anything that contradicts what you already know one file called CLAUDE.md sits at the center holding who you are, so you stop re-explaining yourself every session ask it anything across everything you’ve ever fed it, and it answers from months of compiled understanding instead of starting from zero that loop is the whole system. Karpathy’s repo got 16 million views because engineers finally saw the difference written down: retrieval answers questions, compilation builds understanding five minutes to set up. you’ll never open an empty chat and re-explain your life again full build, every step, in the article. save it now

rvaniaaa

60,398 views • 15 days ago

THIS MIGHT BE THE #1 OPEN-SOURCE REPO FOR CLAUDE CODE RIGHT NOW. IT GIVES CLAUDE A MEMORY AND SLASHES YOUR TOKEN COST ON EVERY QUESTION The repo is safishamsi/graphify, a free open-source skill that turns any codebase into a knowledge graph Claude Code can read instantly. Instead of grepping through your files every session, Claude gets a map of how everything connects The problem it fixes: Every time you ask Claude Code about a big repo, it does the same thing, greps through dozens of files like a brute-force Ctrl+F, blows through your context window, and sometimes still misses the answer hiding in a file nobody searched. Claude Code has no memory of how your project is structured. Every session starts from zero What it does: It maps your entire codebase into a knowledge graph, capturing not just which files exist, but which functions depend on which, which modules are central, and which files cluster around the same concern. Claude queries the map instead of scanning files How it works, three passes: 1. Code structure, free and local. Tree-sitter parses your files and pulls out classes, functions, imports and call graphs. No LLM, no tokens, just your actual code mapped deterministically 2. Audio and video, if you have them. Transcribed locally and folded into the graph 3. Docs, papers, images. Here an LLM does semantic analysis, figuring out what each document means and where it fits. Only the meaning gets sent up, never your raw source It saves you money: Normally a question about a big repo makes Claude spawn explore agents that scan file after file, eating your context window and your token budget before you get an answer. With the graph already built, Claude queries the map instead of re-reading the codebase every time. Same answer, a fraction of the tokens. The graph only gets built once, then a hook rebuilds it after each commit for free, so you never pay that scanning cost again. The bigger the repo, the bigger the gap The best parts: it's a skill, so once installed Claude knows when to use it without you memorizing commands. It works on non-code folders too, point it at docs or notes and it can spin up an Obsidian vault How to add it to your Claude: 1. Install Claude Code if you haven't: npm install -g Paul Jankura-ai/claude-code 2. Add the skill: claude skill add safishamsi/graphify 3. Open your project folder and run /graphify . to build the graph 4. Optional, make it automatic: graphify hook install so the graph rebuilds after every commit That's it. Ask Claude about your repo and it reads the map instead of burning tokens on a file hunt Bookmark this

Yarchi

56,177 views • 3 months ago