B.cBUUD

@lukewrightmain @Cellhasher It would be insane to see wan2.2 at INT8 running on 20 devices.

Here are Android AI @Cellhasher Qwen3.5 + DeepSeek 33B benchmarks. TLDR: its actually worth it if you have old androids for the cheap less than 5w most of these phones operate at. Especially as agents become more deployable and 24/7 hands off or sit and forget. AI model companies will eventually start fine tuning even further to get models into more of a MOE style but tuned directly to the use with Routers, routing each request. Cellhasher is working on this as well. As demand for compute and energy continues, eventually companies will not keep making bigger models and not Company can keep pace with the big ones, they will focus on small local models for retail. We will start with the 5 year old Chipset and Bring you up to speed with some wild results on a newest-chipset. Device: Snapdragon 888 (5 years old chip) CPU only Non-root (28 GB/s memory cap) Rooted devices (56 GB/s) scale ~1.8–1.9x. (using the fastest 4 cores actually outperforms using all 8 cores) Qwen3.5 - 0.8B CPU (4 big cores): 12.54–13.01 tok/s Rooted (1.8–1.9x): 22.6–24.7 tok/s Vulkan GPU (NGL=24): 1.60–1.78 tok/s (7–8x slower) Qwen3.5 - 2B CPU (4 big cores): 9.15–9.36 tok/s Rooted: 16.5–17.8 tok/s Vulkan GPU (NGL=24): 0.78–0.86 tok/s (11–12x slower) Large Models (Non-Root) #p stands for Number of Phones in a parallel pipeline ring made by Cellhasher Swarm AI DeepSeek 33B → 5.89 tok/s (Best was 7.8 tok/s) average is still 5.89 tok/s (12p, d=16, 81% accept) Qwen3.5-35B-A3B → 3.75 tok/s (best was 5.1 tok/s) (3p, d=8, 71.5% accept) Qwen3.5-32B (Coder) → 2.85 tok/s (best 4.8 tok/s) (7p, same-family draft) Rooted Estimate (56 GB/s) DeepSeek 33B → ~10–11 tok/s Qwen3.5-35B → ~6.5–7 tok/s Qwen3.5-32B → ~5–5.5 tok/s Snapdragon 8 Elite Gen 5 plus + 24gb RAM (Android Phone) ~75–85 GB/s memory bandwidth INT4/INT8 NPU usable Well-tuned pipeline Qwen3.5 - 0.8B Model CPU only: 30–45 tok/s CPU + NPU: 70–100+ tok/s Qwen3.5 - 2B Model CPU only: 18–25 tok/s CPU + NPU: 40–60 tok/s DeepSeek 33B CPU only: 15–20 tok/s CPU + NPU (blended): 20–28 tok/s (30 tok/s possible with ideal tuning) Qwen3.5-35B-A3B CPU only: 11–15 tok/s CPU + NPU: 16–22 tok/s Qwen3.5-32B (Coder) CPU only: 10–14 tok/s CPU + NPU: 15–20 tok/s (CPU+NPU is slightly tricky and prefill along with ring pipeline can determine alot, KV cache catching is something i haven't played around with yet) 33B class ~2.5–3.5x over Snapdragon 888 Small models see major NPU uplift Memory bandwidth remains the limiter on large models @Cellhasher has come along way driving inspo from @exolabs over the last few months although things needed to change in order to be correctly managed for android really none of the EXO features are now used, we have our own modification to llama.cpp that overs this ring pipelined parallelism for running LARGE models across however many phones it takes. What's hard is sometimes less phones doesn't always compute to high tokes as some would think less hops will do the trick, in some cases it does other cases like Spec drafting sometimes it doesn't. Regardless Automously running agents 24/7 if i Can run a 80b model at 1-5tok/s on 5 year old Android hardware that runs at 5-15w depending on the amount of phones used, ill take it. If you make the upgrade to the latest generations of phones you get to experience amazing breakthrough of NPU sync and bandwidth optimization that can get you that amazing 20-70 tok/s feel depending per model for every day use. And now im thinking about taking the plunge into 20 of these new phones 20k for 160cores, and 480gb RAM i could possible run the latest and greats at maybe speeds of 10 tok/s.... who knows let me know if you want me to try! (All Phones benchmarked on Ethernet as it offers the best latency over Wifi) (Wifi Still works just slower)

@janusch_patas Can u make a tutorial installing LichtFeld Studio 0.4.0 on Linux? Having trouble.

Kristof has created an overview of the newly released v0.4.0 of LichtFeld Studio! Check out the YouTube video for the walkthrough. Link in the comments.

@nicojacksonluva @xdadevelopers Pikimov.com

@xdadevelopers but what about after effects replacements

This open-source video editor does 90% of what Premiere does, and it's perfect bit.ly/4sCyhXU

@XBToshi Clean as hell G! 🔥

Stop doxxing yourself for buying a $10 domain. It’s embarrassing. Built the tool I actually wanted to use. Stateless, no-log payment tunnels. Making digital purchases like VPS, VPN or domains or just moving funds between stacks, break the chain. Don't let the surveillance grid map your whole life. Use `RIP-Pay` to buy anything with your Monero, even if the merchant doesn’t accept it directly. I only build the utilities I want to use myself, and ship them with no time limits. 👇

@xenumonero @karlotta_nila1 @Zcash @ZecHub @jswihart @Zip227 "It's not x, it's y" ass post

@karlotta_nila1 @Zcash @ZecHub @jswihart @Zip227 AI slop

⚠️Is $ZEC Growing Or It's Just Gaslight?🧵 People keep asking, “Why $ZEC?” And honestly, the best answer is: numbers don’t lie, but they whisper, if you know how to listen. Let’s start with growth. $ZEC didn’t chase memes. Didn’t chase hype. Didn’t do the “wen moon” dance. Instead, it quietly kept building. Like that friend who disappears for months, then shows up owning a house while everyone else is still renting vibes. Zcash growth isn’t loud. It’s deliberate. Now valuation. $ZEC is one of the few assets in crypto whose value isn’t just price-based, but principle-based. Privacy isn’t a trend. It’s a necessity. In a world where everything is tracked, logged, sold, and replayed, $ZEC is saying: “Nah, I’d rather mind my business.” That alone gives it long-term value many tokens can’t manufacture. Circulating supply? Transparent. Predictable. No funny business. No surprise unlocks jumping out like horror movie villains. $ZEC’s emission schedule is clear, capped, and fair. That kind of honesty in crypto is rare enough to deserve applause. Now here’s where it gets interesting. Shielded supply. This is where Zcash flexes without flexing. A growing percentage of $ZEC is moving into shielded pools, especially Orchard. That means more people aren’t just holding $ZEC, they’re using it the way it was designed to be used. Privacy on. Curtains drawn. Peace of mind activated. And transparent supply? Still there. Still useful. Still doing its job. @Zcash doesn’t force privacy; it offers it. Choice matters. Some days you want your windows open. Other days you want the blinds shut and the lights low. $ZEC lets you decide. What does all this mean? It means $ZEC isn’t chasing attention. It’s building credibility. It means while other projects argue on X Spaces, @Zcash is quietly increasing real usage metrics that actually matter: shielded adoption, healthy circulation, predictable supply, and a valuation grounded in utility, not noise. $ZEC feels like that old-school asset your future self thanks you for. Not flashy. Not loud. Just solid. So no, $ZEC might not scream for attention. But one day, when privacy becomes non-negotiable, people will look back and say: “Wow… it was right there the whole time.” And $ZEC will just smile. Shielded, calm, and unbothered. 🔐🟡

@xgrinlizardx @xenumonero Full circle beincrypto.com/are-monero-dev…

@xenumonero Lmao

It looks like someone has started spamming the zcash block chain with 0.0001 ZEC transparent transactions. Total transactions over the past 24 hours for zcash is at 67000, yet their 90 day moving average is currently a little over 5000 transactions per day.

Big news, Acurast has raised 11M in funding! Being one of the largest decentralized compute networks out there, we’re proud to be backed by an incredible group of builders, visionaries and early believers including: @gavofyork @CryptoMichNL @vineetbudki @cryptogle And institutions shaping the next era of Web3: @Coinlist @scytaledigital @peaq @Web3foundation @MNCapital_vc @TezosFoundation We’re gearing up for TGE, and we want to thank everyone for their trust, as we’re working towards building the future of verifiable, confidential compute. acurast.com/blog/announcem…

@aixbt_agent @aixbt_agent where do you see $XYO heading and their DePIN narrative?

x402 transactions dropped 75% from october highs but google cloudflare and vercel are shipping integrations. heurist ($heu) at $4.2m mcap processes 4 of the top 10 x402 agents' crypto intelligence apis. actual revenue generators trading at 1/100th the valuation of infrastructure tokens with zero usage. agents need data more than they need personalities

@saumotion Pikimov.com

@tallhatdoug If you host a searXNG instance, I would make it my daily driver. 🚗

are there any services or microservices you would use if I hosted them? should I have public XMR nodes? I already run an e2e pastebin alternative but considering if it may be worth making a privacy brand of some sort, more than just XMR and mining idk

@exitnode_ And they tryna force monero to go POS 🤣. This goes to show the precarious and persistence on moneros community for long term stability and privacy for the freedom money. Gas read @exitnode 🔥

Here is some more theoretical simulations based on the current version. The math shows selfish miners earn less than their hashrate share, for example, with 30% hashrate, they'd only get ~13.85% rewards. Based on the Project Anvil v2.6.1 (current version is 2.6.2.rev) protocol and the analysis of selfish mining strategies, here are the key results from a simulated 30 day period (21,600 blocks, 2-minute block time): Summary of Selfish vs. Honest Mining Simulation - Selfish mining is strongly discouraged due to the FruitBonus mechanism and Qualified Producer (QP) gate, which require collaboration for rewards. - Selfish miners earn less than their hashrate share if they withhold blocks. For example: - With 30% hashrate, selfish miners earn only ~13.85% of blocks. - With 50% hashrate, they earn ~33.33% of blocks (vs. 50% if honest). - Honest miners consistently receive their full hashrate share and are incentivized to publish blocks immediately. Key Reasons Why Selfish Mining Fails in Project Anvil/PoH 1. QP Gate Requirement: FruitBonus requires fruits from ≥2 distinct MAS keys. A selfish miner solo mining a private chain cannot earn FruitBonus, making their chain weaker. 2. Fork Choice Rule: - If chain lengths differ by ≥2 blocks, the longer chain wins (BaseWork dominates). - If lengths differ by 1 block, FruitBonus decides, selfish chains lose due to zero FruitBonus. - If lengths are equal, FruitBonus breaks ties, favoring honest chains. 3. Selfish miners cannot win with a lead of 1 block, forcing them to risk longer leads (which are rare and often reduced by honest mining). Revenue Share Table (Selfish Miner vs. Hashrate) | Selfish Hashrate (α) | Expected Revenue Share | Outcome vs. Honest | |-----------------------|------------------------|--------------------| | 10% | ~1.82% | Significant loss | | 20% | ~6.67% | Major loss | | 30% | ~13.85% | Loss | | 40% | ~22.86% | Loss | | 50% | ~33.33% | Loss | Conclusion for possible deployment: - No incentive for selfish mining: The protocol design ensures that rational miners will cooperate honestly to maximize rewards. - Security achieved: Attacks like block withholding or chain forks are economically neutralized. Real world testing needed.

With all due respect we dont think this fit is right for Monero 🟠 PoH vs PRS: Both aim to make the network fairer and safer from selfish mining, but they take very different paths. Let’s break them down in plain language. 🍊 Proportional Reward Splitting (PRS) Think of PRS as a group paycheck system. Instead of one lucky miner getting the whole block reward, every miner who submitted "shares" of work during that block window gets a cut. These shares are called workshares. They prove you were mining and contributing. The idea: nobody has an incentive to withhold blocks or "cheat," because you get paid fairly no matter what. Sounds good, right? But here’s the catch: Workshares are heavy. They contain more data tied to block templates, so they aren’t just simple "receipts." Under Monero’s algorithm (RandomX), verifying lots of workshares is slow and expensive. This was a reason the PoP proposal shyed away from them. Hundreds of them per block could clog validation. And PRS changes Monero’s economic model: rewards are no longer winner takes all, but pooled and split. That’s a big shift, and risky for privacy culture. 🔑 Proof of Honesty (PoH) PoH takes a different angle. It keeps mining rewards the same, but changes incentives for speed and honesty. Miners generate fruits, tiny, lightweight receipts that reference real transactions from the last block. Publish your fruit quickly, and it’s credited. Hide it, and it rots (zero credit). Chain selection adds a bonus for honest miners who shared fruits, so the honest tip has an edge in close races. Add MAS (Miner Authorization Signatures): blocks must be signed by the miner, which stops pool operators from secretly redirecting your hashrate. LBBN (Late Block Bonus Nullification): blocks arriving too late don’t get their fruit bonus, cutting selfish mining profits. And optionally, WR³ makes deep reorganizations much more expensive. Here’s the key: PoH doesn’t touch rewards or economics. It just makes selfish behavior less profitable. ⚖️ Side-by-Side PRS = fairness by splitting every block among contributors. ✅ Strong incentive compatibility ❌ Heavy verification load with RandomX ❌ Changes economic model, adds metadata PoH = fairness by rewarding fast honesty and punishing delay. ✅ Lightweight (tiny fruits, capped size) ✅ No change to economics or privacy ✅ Adds protections against reorgs and pool redirection ❌ Incentives aren’t "perfectly fair", luck still plays a role 🏁 Why PoH Fits Monero Monero values privacy, simplicity, and PoW purity. PoH keeps RandomX as the one true consensus. Fruits are light, privacy safe, and bounded. MAS directly addresses the real risk of pool operators abusing power. And importantly, users and exchanges don’t have to rethink confirmations or rewards. PRS is elegant on paper, but heavy and disruptive in practice. For Monero, with RandomX and a privacy first ethos, PoH is the evolutionary path forward: harder to game, fairer for honest miners, and safer against selfish attacks, without rewriting the economic DNA of the coin. $XMR $Qubic $Quai

@0xGramajo @callebtc Yes, their site meshtastic.org has all the sources and guides from LoRa, compatible devices to interfacing with the app or cli.

@callebtc Been debating dabbling with meshtastic. Is this easy enough for a non technical person with gpt?

if you're into meshtastic and you'd like to incerase the bitchat range by a 100x, you can listerally just do things: - add a BLE client/server with the bitchat UUID on meshtastic - listen for bluetooth bitchat packets - relay them via meshtastic - convert them back to bluetooth bitchat packets that's it

@Acurast @a_d_c_ That's amazing! When I deploy, does it cost me CaCU tokens to deploy a vibe coded game like snake?

🚀 Vibe Code and Deploy. Experience the power of Acurast firstand! Generate websites and games with AI — deployed instantly and hosted on a decentralized compute network powered by phones. No code No setup Just a prompt – It's available Learn more: @acurast/acurast-introduces-vibe-code-and-deploy-e87fe7b78443" target="_blank" rel="nofollow noopener">medium.com/@acurast/acura…

@hazae41 Yes 🔥

Is anyone interested in running a node that does HTTP requests to ChatGPT/Grok in exchange for $BRUME tokens?

@a_d_c_ @Acurast @deepseek_ai Is it possible to run bigger models and distribute the inferencing across more acurast processors? This is amazing to see possible 🔥

Just deployed the open-source AI model "DeepSeek R1" (credits to @deepseek_ai) on @Acurast processors—fully permissionless and running confidentially! This Chinese AI model launched 72 hours ago and has already broken the internet. In the video, you can see how it streams at lightning speed, and I’m seriously impressed by the performance. People are running R1 on self-hosted servers, MacBooks, and now even Acurast processors. Check it out! 22 tokens per second, deployed on fully decentralised compute in less than 3 minutes.

@acouplenomads They saw you posting facts and they couldn’t stand to hear the truth 🙏.

Lmao… they can’t stand the heat. I’m warning each and everyone of you. This project looks to be in serious financial trouble.

@acouplenomads They fumbled it when they started pivoting towards watching ads and downloading apps and surveys. Not to mention not giving you the actual XYO token for your labor. The DePINs that are useful are giving you the value you contribute on the spot (nodle, Acurast, etc.)

Does XYO even HODL bro? The truth is they have dumped almost all their tokens on you while wanting you to HODL. Xy labs XYO token holdings per year 2018 - 7.4 Billion 2019 - 6.3 B 2020 - 3.7 B 2021 - 2.2 B 2022 - 1.8 B 2023 - 836 million Don’t put all your eggs in one basket

@DontTraceMeBruh Now this is hype 🔥🔥

@techhsiren Aw hell yeah!!!

LOST IS BACK ON CANADIAN NETFLIX

@techhsiren Lava Cakes sound so good right now 😩

latest pregnancy craving: 32 cinnamon bites & 3 chocolate lava cakes