WhisperHash (@whisperhash) profile picture

Nostr and Bitcoin and hardware tinkerer. #SovEng SEC-08 npub1t6m495kynns7xjmgz8yt6krrw4r8tac0mvr3f25f9r35ke72ap3su4qycp

Joined December 2025
Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)Photo by WhisperHash (@whisperhash)
I can watch this all day.
How CPU coolers are made
1
24
I'm discussing with another open-source Bitcoin dev and the right approach feels like running Kimi K3 or later and steering it for unrestrained security audits on our projects 24/7 until the end of time. Token war. Thank God for the open-source AI models.
37
46
4
624
19,659
WhisperHash retweeted
How to go viral on your first post on Primal. 1. Introduce yourself. 2. Tag who brought you here. 3. Use the hashtag #introductions 4. Add a photo of you or your interests. Introduce yourself & join the web of trust. WARNING: random people will send you bitcoin.
4
19
3
55
6,708
WhisperHash retweeted
Goooood morning! First block for the hosted instance but #5 overall. Hash on!
The first @Public_Pool_BTC on the hosted version has been found! Congratulation: mempool.space/block/00000000…
7
21
4
163
8,299
WhisperHash retweeted
Public Pool had a 9.5-hour outage on March 29. We were curious what the publicly available data showed, so we pulled the numbers and dug in. The data tells an interesting story about failover behavior, geographic distribution, and a wallet address configuration issue that affects hundreds of Bitaxe owners. No hot takes. Just data: atlaspool.io/resources/artic…
4
5
21
904
It’s fine - should be a positive for code consistency…
🚨 BREAKING: Researchers at UW Allen School and Stanford just ran the largest study ever on AI creative diversity. 70+ AI models were given the same open-ended questions. They all gave the same answers. They asked over 70 different LLMs the exact same open-ended questions. "Write a poem about time." "Suggest startup ideas." "Give me life advice." Questions where there is no single right answer. Questions where 10 different humans would give you 10 completely different responses. Instead, 70+ models from every major AI company converged on almost identical outputs. Different architectures. Different training data. Different companies. Same ideas. Same structures. Same metaphors. They named this phenomenon the "Artificial Hivemind." And the paper won the NeurIPS 2025 Best Paper Award, which is the highest recognition in AI research, handed to a small number of papers out of thousands of submissions. This is not a blog post or a hot take. This is award-winning, peer-reviewed science confirming something massive is broken. The team built a dataset called Infinity-Chat with 26,000 real-world, open-ended queries and over 31,000 human preference annotations. Not toy benchmarks. Not math problems. Real questions people actually ask chatbots every single day, organized into 6 categories and 17 subcategories covering creative writing, brainstorming, speculative scenarios, and more. They ran all of these across 70+ open and closed-source models and measured the diversity of what came back. Two findings hit hard. First, intra-model repetition. Ask the same model the same open-ended question five times and you get almost the same answer five times. The "creativity" you think you're getting is the same output wearing a slightly different outfit. You ask ChatGPT, Claude, or Gemini to write you a poem about time and you keep getting the same river metaphor, the same hourglass imagery, the same reflection on mortality. Over and over. The model isn't thinking. It's defaulting to whatever scored highest during alignment training. Second, and this is the one that should really alarm you, inter-model homogeneity. Ask GPT, Claude, Gemini, DeepSeek, Qwen, Llama, and dozens of other models the same creative question, and they all converge on strikingly similar responses. These are models built by completely different companies with different architectures and different training pipelines. They should be producing wildly different outputs. They're not. 70+ models all thinking inside the same invisible box, producing the same safe, consensus-approved content that blends together into one indistinguishable voice. So why is this happening? The researchers point directly at RLHF and current alignment techniques. The process we use to make AI "helpful and harmless" is also making it generic and boring. When every model gets trained to optimize for human preference scores, and those preference datasets converge on a narrow definition of what "good" looks like, every model learns to produce the same safe, agreeable output. The weird answers get penalized. The original takes get shaved off. The genuinely creative responses get killed during training because they didn't match what the average annotator rated highly. And it gets even worse. The study found that reward models and LLM-as-judge systems are actively miscalibrated when evaluating diverse outputs. When a response is genuinely different from the mainstream but still high quality, these automated systems rate it LOWER. The very tools we built to evaluate AI quality are punishing originality and rewarding sameness. Think about what this means if you use AI for brainstorming, content creation, business strategy, or literally any task where you need multiple perspectives. You're getting the illusion of diversity, not the real thing. You ask for 10 startup ideas and you get 10 variations of the same 3 ideas the model learned were "safe" during training. You ask for creative writing and you get the same therapeutic, perfectly balanced, utterly forgettable tone that every other model gives. The researchers flagged direct implications for AI in science, medicine, education, and decision support, all domains where diverse reasoning is not a nice-to-have but a requirement. Correlated errors across models means if one AI gets something wrong, they might ALL get it wrong the same way. Shared blind spots at massive scale. And the long-term risk is even scarier. If billions of people interact with AI systems that all think identically, and those interactions shape how people write, brainstorm, and make decisions every day, we risk a slow, invisible homogenization of human thought itself. Not because AI replaced creativity. Because it quietly narrowed what we were exposed to until we all started thinking the same way too. Here's what you can actually do about it right now: → Stop accepting first-draft AI output as creative or diverse. If you need 10 ideas, generate 30 and throw away the obvious ones → Use temperature and sampling parameters aggressively to push models out of their comfort zone → Cross-reference multiple models AND multiple prompting strategies, because same model with different prompts often beats different models with the same prompt → Add constraints that force novelty like "give me ideas that a traditional investor would hate" instead of "give me creative ideas" → Use structured prompting techniques like Verbalized Sampling to force the model to explore low-probability outputs instead of defaulting to consensus → Layer your own taste and judgment on top of everything AI gives you. The model gets you raw material. Your weirdness and experience make it original This paper puts hard data behind something a lot of us have been feeling for a while. AI is getting more capable and more homogeneous at the same time. The models are smarter, but they're all smart in the exact same way. The Artificial Hivemind is not a bug in one model. It's a systemic feature of how the entire industry builds, aligns, and evaluates language models right now. The fix requires rethinking alignment itself, moving toward what the researchers call "pluralistic alignment" where models get rewarded for producing diverse distributions of valid answers instead of collapsing to a single consensus mode. Until that happens, your best defense is awareness and better prompting.
30
WhisperHash retweeted
We've got a big job! If you'd like to support this critical mission, make your tax deductable donation at 256foundation.org 256 Foundation is 100% pass though to developers working on open source Bitcoin mining.
I am stoked to announce I've joined the board of the @256FOUNDATION ! We are going to dismantle the proprietary Bitcoin mining empire -- and rebuild decentralized open source so Bitcoin remains freedom money.
2
7
37
5,502
Integrated a particulate matter (PM) sensor - Sensirion SPS30 - into my envbox build. This should hopefully give me some insight into the dust buildup / particulates in the air around where my miners are running. Ran out of available GPIOs on the original ESP32 “CYD” board, so spun this up as a separate “dustbox“ unit for now. Will be getting a different ESP32 board with more available GPIOs next most likely - to bring this all back onto a single MCU board, and reduce the overall footprint.
35
Running Hydra Pool :) On an Umbrel, on mainnet. Much delight. Smooth sailing - great instructions on hydrapool.org. This is running :latest via docker compose. Also built locally from Rust source earlier - breezed right through it 👌 github.com/256foundation/hyd… Awesome job guys, amazing! Thank you @256Foundation @jungly @econoalchemist et al
23
WhisperHash retweeted
Open source hashboards, controlled by an open source control board, running open source firmware, mining to an open source mining pool. This is the @256FOUNDATION
17
59
5
323
15,846
WhisperHash retweeted
Bitcoin Miners: LFG, today is the day! Support open source mining and point your hashrate to the @256FOUNDATION telehash. dash.256f.org
TODAY TELEHASH #3 dash.266f.org Miners big and small, point us some hash and let’s hit another block…or 2! 🫡
3
13
38
4,322
Integrated a sound pressure level sensor into my envbox / sensor array build. This now will allow me to cap the max dB(A) / noise level generated by the miners as I auto-tune / overclock them - in the same way you’d normally cap max ASIC and VR temperatures, and max power.
22
Love the idea of a Telehash fundraiser - what a beautiful and easy and native way to participate! @256foundation - love what you guys are doing supporting open-source mining stack development and education, good luck winning a block this year! ⛏️
33
Very very pleased to see some very tasteful, plain awesome-looking designs, emerge in the community
Sometimes, the “best” Bitcoin miner isn’t the fastest miner. If you want max hashrate, buy a Bitmain S23. If you want max (tolerable hashrate) at home, look at the NerdOctAxe. However, If you want beauty, keep reading. “Best” is subjective. With home mining (and anything I’ll be looking at every day), I’ll choose beautiful design over outright hashrate every time. If it doesn’t exist, I’ll design it and make it. Shown here: the BitForge Nano in a bespoke all-metal case with a glass viewing window — designed by me from the ground up — plus full-colour, controllable internal LEDs. And yes: it has the same trademark “share submission” LED pulse you may've seen on my BitHalo LED Board. ✨ The OG Bitforge Nano is on sale now, available from myself or DTV Electronics (gotta show them some love!) This beautiful metal edition (name TBC - open to suggestions) is going live in due course (few weeks based on production parts incoming) The Bitforge Nano is a direct competitor to the Bitaxe Gamma GT: both run 2× BM1370 ASICs, but with very different aesthetics. Both are open source, both are maintained by the same Bitaxe team and both rock. However, only one will look kick-ass on your desk. The BitForge Nano was created by the lead Bitaxe software dev' @wantclue, with the mining hardware developed by his awesome sidekick @kliA90_ — both of whom I’m lucky to call friends. The heatsink and airflow paths and relevant calc's were also done by me. Without @dtvelectronics pushing the BitForge Nano into production we may not have seen a finished product. What do you think? It needs a name too. Help!
1
34
Show your miner a little love today and de-dust the heatsink / fan 🧹 This guy went from laboring at 100% fan -> ~60%, all else being equal. 

The fan acoustics were starting to sound peculiar, which is what drew my attention. Hopefully no lasting damage to the fan, but I’m not so sure. Wish there was a better way to get alerted earlier.
27
Fanless means: • Less dust • Zero distractions • Increased reliability
22
22
6
1,120
63,010
😅
“The U.S. dollar isn’t even backed by anything.” The U.S. dollar:
1
70
WhisperHash retweeted
Love seeing what the community builds with #InfluxDB + Grafana. A look at real-time Bitcoin miner fleet metrics. 🔍
This just keeps getting more and more delightful - fresh iteration of my Grafana dashboard plotting Bitcoin miner fleet metrics. Miners pushing data into @InfluxDB -> @grafana, all running on an @umbrel 👌
2
3
467