I just applied for the @playcambria Genesis whitelist. Score: 219 (45 + 174 Cambria) · Degen 7 · Chad 57 lobby.cambria.gg/genesis?cod…

came to build my dream PC, ended up in a cemetery👻

changed to 333 creator spots suddenly the hardest part of building a pc is the creative writing degree. still, the NFT boosting in-game earnings is the bit i want to see tested. does it make upgrading more interesting, or just make the numbers bigger?

fomies has turned the whitelist queue into part of the project. the clerk turns up late because he went shopping. management refuses to explain anything. access cards arrive alongside a suspicious shipment of pickles. the whole thing feels like an office where nobody can get fired because nobody knows who hired them. what works for me is the consistency. on the @FomiesNFT website, you enter the department and ring a bell to speak to the clerk. on x, the same office keeps producing excuses, announcements and paperwork. even waiting for approval fits the story. that gives the community something specific to work with. you can invent a job, file a complaint or write an incident report, and it still feels like part of the same world. my proposed position: assistant to the assistant printer technician. qualifications: - can identify a paper jam - cannot fix a paper jam - will escalate the paper jam to management for anyone following the mint, the team says more applicants are receiving access cards each day. it’s a free mint, but an application still needs approval. i’m curious how they carry the department forward after mint. employee files tied to characters, ridiculous internal assignments, office rivalries - there’s plenty they could do with it. for now, someone needs to explain why the pickle department has a bigger budget than hr.

the funniest trait in @FedProtocol is a money printer that produces ious. a machine built to print money, handing you a promise instead. fairly strong financial commentary for a 32x32 sprite. the collection has 2,222 printers on robinhood chain. free mint plus gas, one per wallet. each combines an era, a condition and an output. that means the same basic machine can tell very different stories: an old press printing hundreds, an overheating industrial model, or something that has simply run out of ink. there’s also a detail worth understanding before chasing a rare one. the site lists these fee-share weights: - common: 1x - rare: 1.5x - legendary: 2x a printer uses the highest tier among its three traits. the weights don’t add together. so one legendary trait puts it in the listed 2x tier. three legendary traits don’t mean 6x. those are relative weights, not a promised return. i still want to see the source of the fees and the distribution mechanics explained properly. the art already has a clear joke. the next thing to judge is how the economics hold up. if you’re applying, the site’s flow has five steps: wallet, x, follow, apply, share. check that you’ve completed the share step too. personally, i’m attached to the iou printer. finally, a piece of financial infrastructure willing to admit it’ll get back to you.

Applied for a printer on the Federal Reserve. Application #7620. Dual mandate: maximum brrr, stable memes. moneyprinter.lol/r/M28M21N7

I've just joined Early Access for @ComputersRh. I'll be playing 24 hours before the public launch. My ref link= computersrh.xyz/ref=nemesisv…

the courier gets an AI assistant, the restaurant gets an AI dashboard, and you get to text a bot "the usual." at the end of this whole chain, a real person is still folding the burrito. but soon this will be replaced too lol.

Edge AI is having its GPT-3 moment, and the door is finally open to beginners You can now describe a circuit board in one sentence and get back real KiCad files. Whether that board works is a different question, and it's the whole game. Plenty of engineers are already building with this. Microcontrollers now ship with neural accelerators, and platforms like Edge Impulse have tens of thousands of developers. What's changing is who can join them: for a beginner, the path from "I have an idea" to "it runs on a board" is much shorter than it used to be. It reminds me of how language models opened up around GPT-3. Not better than experts, but reachable by far more people. (That's an analogy, not a benchmark.) WHY NOW The hardware caught up. ST's STM32N6 is a microcontroller with a built-in neural accelerator. ST quotes up to 600 GOPS and 4.2 MB of on-chip RAM, plus a camera interface. That's the class of chip that can run real vision models without a Linux board. The software caught up too. Edge Impulse reported 118,185 projects from 50,953 developers as of October 2022, per its own paper. And on the board side, heypcb turns a plain-language description into a schematic, layout and fab files as KiCad files you keep. Its terms reportedly warn that AI designs can be wrong or unsafe, so the review is on you. Drafting got cheap. Verifying didn't. That's where the skill moved. WHAT TO ACTUALLY LEARN (IN THIS ORDER) - C, just the core. Pointers, structs, arrays, how memory works. You don't need C++ templates or a kernel. Python only enough to train a small model with NumPy and Keras. - One cheap board. An ESP32, RP2040 or Arduino Nano 33 BLE Sense. Read a sensor, blink an LED, print over serial. Skip RTOS and Linux for now. - The three ideas that make TinyML tiny. Quantization (int8 instead of float32), the memory budget (flash and RAM, not "GPU"), and signal windows (slicing sensor or audio data into chunks and turning them into features). A lot of beginner problems trace back to one of these. - Ship the classic projects. The book TinyML by Pete Warden and Daniel Situnayake walks through three: a speech recognizer, a person-detecting camera and a gesture "magic wand". The publisher says no prior ML or microcontroller experience is needed. Pair it with Edge Impulse for a no-code start and Shawn Hymel's Coursera course. To go deeper into on-device deployment, HarvardX's Deploying TinyML on edX teaches TensorFlow Lite for Microcontrollers. - PID control. One of the oldest tricks in embedded, and still everywhere: read a sensor, compare it to the target, correct the output. Thermostats, motor speed, drone stabilization. Build a temperature or motor-speed loop and tune the three gains by hand. It teaches you timing, sensors and actuators in a way blinking an LED never will, and it's a good reminder that not every problem needs a neural network. Phil's Lab has videos on control systems. - Read a datasheet and a schematic. Power, decoupling caps, USB, crystals. Phil's Lab on YouTube shows KiCad and STM32 board design from schematic to layout, and covers real-time DSP in C, which is the "features" skill from the TinyML ideas above. Open KiCad next to the video. - Then use AI as your board drafter. Describe your device to heypcb, or fork one of the 8,800+ open boards the company says are in Circuit World. Go through the result net by net until you can defend every connection. Run ERC and DRC. Order the board. - Level up to vision. Get an NPU-class board (the STM32N6 has an ST toolchain and a public model zoo) and run a real detection model. That's your own GPT-3 moment demo. WHAT TO SKIP AT FIRST FPGAs, custom silicon, writing your own drivers from scratch, and anything that promises "mastery of embedded" before you've shipped one board. One finished project beats ten courses. The barrier to entry dropped. The responsibility didn't.

DUKZ is now online.🔥 Whitelist applications are officially open. The flock is forming. 🦆 Request access → dukz.fun

The trade you regret most isn't the one that lost. It's the one you took when you already knew better. You saw the setup wasn't clean. You knew the size was too big. You felt yourself chasing. And you clicked anyway. Losing trades are part of the game. You planned them. You sized for them. You accepted them before the entry. The trade that hurts is different. It's the one that broke your own rules before it broke your account. The one that told you something about yourself you didn't want to know. Bad losses fade. Broken discipline compounds. That's the trade to review. Not the one that went wrong. The one where you did.

Yesterday we spoke at the Nado Builders AMA in the Discord. Walked the community through THORNado - where the product stands today, why we built it around Nado, and what's coming next. Thanks to the Nado team for the space. And to everyone who showed up, asked questions, and shared feedback. This is the part of building that doesn't fit on a changelog but shapes the direction of the product more than most releases do. More soon.

Trading is a solo sport with team problems⚡️ You sit at your screen alone. You make the click alone. You take the loss alone. Nobody enters the position with you, and nobody holds it for you. That's the sport. But the problems? The problems are shared. Every trader has: Taken a revenge trade after a stop-out. Sized too big because "this one I'm sure about." Sold the bottom, bought back higher. Stared at a chart for six hours and taken nothing. Watched a setup they saw coming happen without them. Convinced themselves a losing thesis was still valid. You think you're the only one because nobody talks about it. Everyone posts wins. Nobody posts the trade they took at 3am because they were bored and lost half a percent for no reason. The isolation is the illusion. The mistakes are the reality, and the reality is universal. That's why we built THORNado's community the way we did. Not as a shill channel, not as an alpha group, not as a leaderboard. As a room where traders can be honest about the parts of the sport that don't fit on a P&L screenshot. The trade is yours. Nobody can help you enter or exit - that's still solo work. But the loneliness in between is a design flaw, not a feature. Fix that, and you trade sharper.

More @inkonchain integrations coming soon. We're working right now on Strategy Labs - see ya in our trading automatisation.

Also important update coming in 2 days.. Excited. See ya guys.