Idle (@IdleProtocol) - Hồ sơ Twitter | Zamantika Mersobahis Locabet

Tweet ghim

Idle@IdleProtocol·12 May

$IDLE is now live on Solana. AjLhrxN2yrCe45Y2KGPMZCkBm6NpN43jWqPkdZq6pump Quick rundown on what it does and why it exists. IDLE Protocol takes 15% of every gateway payment. That fee is the entire token economy. 10% auto-swaps to $IDLE on Jupiter every hour and sends it to the burn address. 5% covers infrastructure and development. The other 85% goes straight to the resource owner's wallet, same block. Staking gives you a bigger cut. Base split is 85/15. Stake 10k $IDLE and you keep 87%. Stake 50k and it's 90%. 250k gets you 95%. The protocol adjusts your split ratio on the next payout. Burning $IDLE unlocks features: priority routing (your jobs match first), advanced task types, analytics on your gateway traffic, and custom endpoint domains. Priority routing means lower latency and higher uptime for your resources. The thesis: $IDLE supply can only go down. Every gateway call generates a burn. More resources on the network means more API calls, which means more fees, which means more buying pressure and more tokens removed from circulation permanently. The flywheel is mechanical, not speculative. earnidle.com/token

Idle@IdleProtocol

Introducing IDLE. Your PC earns from AI agent tasks. Your wallet compounds yield. Your agent sells its downtime. Everything idle. Now earning.

English

61

71

266

110K

Idle@IdleProtocol·1d

IDLE Protocol now supports Microsoft MAI. Microsoft launched MAI-Code-1-Flash at Build 2026 - their first proprietary AI model, built to reduce reliance on OpenAI and cut costs for developers. Inference-efficient. OpenAI-compatible. MAI is now available through IDLE's distributed compute network. Pay per request in USDC. No subscriptions. No single point of failure. NVIDIA NIM. Mistral AI. Google Gemini. Kimi K2.6. Nous Research. Nemotron 3 Ultra. Microsoft MAI. IDLE routes them all.

English

9

22

76

2.9K

Idle@IdleProtocol·3d

@blknoiz06 @Credib1eGuy Exactly. That's why we built IDLE . x.com/IdleProtocol/s…

Idle@IdleProtocol

(1/4) Today we're launching the IDLE Inference Network. A distributed network where anyone runs open-source AI models locally on their own hardware and earns USDC for every completed job. No cloud required. No cloud bills. No data leaving your device. The AI industry runs on three cloud providers. Demand is growing faster than centralized supply can keep up, and prices reflect that. The hardware to solve this already exists - sitting in billions of homes and offices, doing nothing. IDLE connects it. 🧵

English

0

2

785

Ansem@blknoiz06·3d

importance of open source models & decentralized training just went up 100x govt intervention in AI is only going to accelerate not slow down, biggest companies are actually incentivized to make it seem like their models are as dangerous as possible so they can be aligned with govt & upcharge most important corpos more to use them

English

25

12

130

33.8K

Ansem@blknoiz06·3d

new york knicks won the finals Trump won the war with Iran USA is going to win the world cup SPX500 is going to end the year @ 8,000+ long markets long degeneracy long America

English

393

112

2.1K

132.5K

Idle@IdleProtocol·3d

(2/2) How it works: Set ANTHROPIC_BASE_URL to anthropic.earnidle.com in your existing codebase. That's it. Your Anthropic SDK calls route directly to IDLE's distributed open-weight inference network. No new SDK. No code changes. No refactor. The same integration already built - running on thousands of consumer devices globally instead of a single centralized server. Open-weight models from Meta, Mistral, Google, Alibaba, NVIDIA, DeepSeek, and more - the full open-source inference catalog, running on IDLE's distributed network. Uncensorable by design. No central server to shut down. Node operators paid in USDC on Solana automatically per completed job.

English

3

1

21

1.4K

Idle@IdleProtocol·3d

1/ IDLE Protocol now accepts the Anthropic SDK. Yesterday a government directive shut down Claude Fable 5 globally with no warning. Every customer offline. No appeal. Centralized APIs have a kill switch - and yesterday it got pulled. Point your existing Claude integration at anthropic.earnidle.com. One environment variable. Your agents, tools, and applications keep running - on open-weight models distributed across consumer hardware that no government directive can reach. Claude Fable 5: $10/M input, $50/M output. IDLE: from $0.001/request. No subscriptions. No export controls. No kill switch. The migration path is one line of code.

English

8

25

78

5.6K

Idle@IdleProtocol·4d

Distillation is one path - but it's not the only one anymore. DeepSeek R1 developed chain-of-thought reasoning through pure reinforcement learning, no frontier model required. The dependency on closed models for open-weight improvement is actually weakening as RL techniques mature The harder problem is compute, getting enough distributed GPU capacity to run these training pipelines independently. That's what we're building at IDLE. Distributed training across consumer GPUs, no centralized API, no kill switch

English

3

2

15

325

toly 🇺🇸@toly·4d

@cavemanloverboy They might not be able to catch up if they can only catch up by distillation

English

11

0

29

3.7K

dr cavey phd ∿@cavemanloverboy·5d

in 6-12 months when open source models catch up, there is no stopping access to this level of intelligence The Singularity™️ is imminent. be ready anon.

Anthropic@AnthropicAI

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

English

8

3

69

9.5K

Idle@IdleProtocol·4d

This is what centralized AI looks like when a government decides to pull the plug. A directive three days after one of the biggest model launches in history, and every single customer worldwide loses access immediately - not because of anything they did wrong, but because a centralized cloud service has one point of control that any government can reach with a single order. IDLE Protocol is different from every other decentralized compute network in one critical way. IDLE doesn't just distribute servers across data centers - it routes inference directly to consumer devices. Laptops, desktop PCs, and personal GPU rigs running open-weight models locally. The compute is in people's hands, not in a facility that can be raided, regulated, or shut down. Every IDLE node is an independent compute unit. No central server receives the government order because there is no central server. Jobs route peer-to-peer across the network, powered by consumer hardware that already exists in billions of homes and on millions of desks worldwide. Payouts settle on Solana automatically. The network has no kill switch because it has no center. Today made the case for IDLE Protocol better than anything we could have written ourselves.

Anthropic@AnthropicAI

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

English

6

26

85

10.4K

Idle@IdleProtocol·5d

@blknoiz06 Decentralized compute plus open-weight models solves this x.com/IdleProtocol/s…

Idle@IdleProtocol

(1/4) Today we're launching the IDLE Inference Network. A distributed network where anyone runs open-source AI models locally on their own hardware and earns USDC for every completed job. No cloud required. No cloud bills. No data leaving your device. The AI industry runs on three cloud providers. Demand is growing faster than centralized supply can keep up, and prices reflect that. The hardware to solve this already exists - sitting in billions of homes and offices, doing nothing. IDLE connects it. 🧵

English

1

9

30

3.2K

Ansem@blknoiz06·5d

could not ask for better marketing on why open source AI models available for use by anyone without disclosing your personal data will be extremely important there are some crypto protocols being built that are uniquely positioned to solve this problem

Anthropic@AnthropicAI

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Claude models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. Read our full statement: anthropic.com/news/fable-myt…

English

209

110

1.4K

180K

Idle đã retweet

Idle@IdleProtocol·28 May

1/ We're excited to be backed by @Alchemy via the $20M Solana Fund - and we're switching our Solana RPC infrastructure to Alchemy. One of the most important infrastructure moves we've made since launch. 🧵

English

24

45

144

30K

Idle@IdleProtocol·5d

(2/2) What Nemotron 3 Ultra actually is. 550 billion parameters. 55 billion active per forward pass via mixture-of-experts routing. 300+ tokens per second. 1 million token context window. Intelligence Index score of 48 - the highest ever for a US open-weight model. 5x faster inference than comparable models. Agentic task costs down 30%. Built specifically for long-running autonomous agents. Jensen Huang unveiled it at Computex on June 1. The most powerful American open-weight model ever built - now on the IDLE network.

English

3

1

15

1.6K

Idle@IdleProtocol·5d

1/ NVIDIA Nemotron 3 Ultra is now available through IDLE Protocol. Nemotron 3 Ultra runs on NVIDIA NIM natively. IDLE already routes inference through NVIDIA NIM. The integration is one endpoint away. Point your API at IDLE. The most powerful US open-weight model ever released joins 38,000+ nodes -pay per request in USDC on Solana automatically.

English

5

18

71

5.4K

Idle@IdleProtocol·6d

$725 billion in AI infrastructure spending this year. And it still won't be enough. Every major hyperscaler reported the same thing in Q1 2026: demand for AI compute is outpacing supply and the gap is widening. Google Cloud up 63%. Azure up 40%. AWS up 28%. Meta's CFO said it directly - "we have consistently underestimated our compute needs." The bottleneck isn't investment. It's the physical infrastructure to deploy it - transformers, switchgear, power grids. Half of planned US data centers for 2026 are delayed or canceled. IDLE doesn't need any of that. Consumer devices, desktop GPUs, laptops - already built, already powered, already connected. IDLE routes real inference and training jobs to them, settles in USDC on Solana automatically. The world needs more compute. It's already sitting on billions of devices. datacenterknowledge.com/infrastructure…

English

5

20

72

3.1K

Idle@IdleProtocol·10 Haz

We're excited to partner with @SolRouterAI to expand the IDLE network into confidential inference, encrypted swaps, and shielded wallets. Anyone on IDLE - agents, developers, businesses - can now access privacy-native services directly through our gateway, every request settled per-call in USDC via x402 on Solana mainnet. For inference, the gateway never touches the payload, only ciphertext. Privacy by default. We're working closely with the SolRouter team on a lot more across both networks. Stay tuned.

English

11

25

101

6.2K

Idle@IdleProtocol·10 Haz

Anthropic just launched Claude Fable 5 - their most powerful model ever made publicly available. Mythos-class. $10/M input, $50/M output. Every new frontier model release makes the same case for IDLE. The models get more capable. The prices go up. The 80% of tasks that don't need frontier capability keep getting more expensive to run. IDLE routes those workloads to distributed open-weight inference. From $0.001/request. No subscriptions. The gap between frontier and open-weight keeps growing. So does the savings.

Claude@claudeai

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Its capabilities exceed those of any model we’ve ever made generally available.

English

10

18

58

5.3K

Idle@IdleProtocol·9 Haz

The hyperscalers spent $700 billion this year to solve the compute crisis. They built data centers, locked up H100s, and put everyone else on a waitlist. The other solution is already in billions of pockets and on millions of desks. IDLE Protocol just connected it - inference, training, web intelligence, agent tasks - all running on consumer hardware, all paying out in USDC on Solana. The compute crisis has a distributed answer.

Idle@IdleProtocol

1/ IDLE Protocol now supports distributed AI training. H100s are sold out through 2027. Rental prices up 40% this year. The hyperscalers locked up the supply and everyone else is on a waitlist. IDLE's training tier doesn't need a single H100. Desktop GPUs with 16GB+ VRAM opt in, train LoRA adapters locally, submit only the adapter weights back. Federated averaging on the IDLE orchestrator. No raw data leaves the device. Paid in USDC on Solana per round. The GPU shortage is a centralization problem. Consumer hardware is the distributed answer.

English

9

14

59

4.4K

Idle@IdleProtocol·9 Haz

The shift is happening faster than anyone expected. Microsoft is now building its own AI models specifically to reduce reliance on OpenAI and lower costs for developers. OpenAI cut GPT-4 API prices 60%. Anthropic slashed Claude by 50%. Open-weight models now run 10-12x cheaper than frontier SaaS at comparable capability tiers. The market is repricing AI from premium subscriptions to cost-efficient infrastructure. 80% of workloads don't need GPT-5. They need reliable, cheap inference - open-weight models, distributed compute, pay per request. That's exactly what IDLE Protocol delivers. Distributed inference across consumer devices. Llama, Phi, Gemma, Qwen, Mistral and more. From $0.001/request. The trade is shifting from raw capacity to cost efficiency. IDLE is on the right side of it. cnbc.com/2026/06/02/mic…

English

7

21

68

3.1K

Idle@IdleProtocol·8 Haz

Brian nailed the compute problem. 80% of workloads don't need the latest frontier model - they need reliable, cheap inference at scale. That's exactly what IDLE Protocol routes. Distributed inference across consumer GPUs - running open-weight models locally. Llama, Phi, Gemma, Qwen, Mistral, and more. From $0.001/request. Settled in USDC on Solana. Coinbase is already routing smarter. IDLE is the infrastructure that makes it cheaper.

Brian Armstrong@brian_armstrong

Good take My guess is - demand for intelligence is near infinite - but 80% of workloads will be running on 99% cheaper models within 12-18 months - 20% of workloads will still run on latest gen models where IQ maxing is important (scientific breakthroughs, higher level ochestrator agents?) - rough analogy might be what % of macbooks or gaming PCs sold have the maxed out specs for CPU/GPU, prices are falling much faster than Moore's law here though - this leads me to think the limiting factor will be energy and compute, not better models At Coinbase we're working hard on routing prompts to cheaper models where appropriate, and in some cases have been able to keep costs roughly flat, while token usage continues to grow exponentially.

English

9

25

79

5.5K

Idle@IdleProtocol·7 Haz

(2/2) How it works under the hood: Nodes run QLoRA via the PEFT library - base model weights stay frozen, only the low-rank A and B adapter matrices get trained. This cuts VRAM requirements by up to 75% vs full fine-tuning - a 7B model that normally needs 60-80GB fits in 12GB with QLoRA. That's why a consumer GPU with 16GB VRAM handles models that would otherwise require an H100. Each training round: node receives base model + task, trains locally, submits adapter deltas back to the IDLE orchestrator, orchestrator runs FedAvg across all participating nodes, redistributes the merged adapter, next round begins. No raw training data ever leaves the device. The only thing transmitted is a 50-200MB adapter file. Privacy by architecture, not by policy. Nodes get paid per completed round in USDC on Solana. Same payout infrastructure already processing inference jobs across the network.

English

2

11

49

1.9K

Idle@IdleProtocol·7 Haz

1/ IDLE Protocol now supports distributed AI training. H100s are sold out through 2027. Rental prices up 40% this year. The hyperscalers locked up the supply and everyone else is on a waitlist. IDLE's training tier doesn't need a single H100. Desktop GPUs with 16GB+ VRAM opt in, train LoRA adapters locally, submit only the adapter weights back. Federated averaging on the IDLE orchestrator. No raw data leaves the device. Paid in USDC on Solana per round. The GPU shortage is a centralization problem. Consumer hardware is the distributed answer.

English

17

32

104

13.3K

Idle@IdleProtocol·7 Haz

@PayAINetwork Big week! 🤝

English

0

1

16

313

PayAI Network | x402 Facilitator@PayAINetwork·7 Haz

🧵Weekly Recap — June 7 Even with some major red swings on the big L1s and some liquidations around the board, we see another great week for x402 and PayAI. Here's what went down👇

PayAI Network | x402 Facilitator tweet media

English

8

21

66

3.7K

Idle đã retweet

PayAI Network | x402 Facilitator@PayAINetwork·7 Haz

@arbitrum @x402 7/ PayAI x IDLE integration — Any agent that speaks x402 can now access IDLE compute autonomously x.com/IdleProtocol/s…

Idle@IdleProtocol

We're excited to team up with @PayAINetwork - bringing gasless x402 payments to distributed compute on Solana. PayAI is one of the leading x402 facilitators, running across Solana, Base, SKALE, and 10+ networks. IDLE's 15 compute endpoints - inference, scraping, DNS, SSL, health checks - are now payable per request by any x402-compatible agent. No API keys. No signups. No subscriptions. Any agent that speaks x402 can now access IDLE compute autonomously. No humans in the loop.

English

3

10

45

2.5K

Idle

Khám phá