Alice

From closed labs to open networks.
Alice Protocol lets the world train AI together and share the upside together.
Powered by GPUs. Driven by contribution.

[Alice @Alice_protocol]

Alice Upgrade Notice The current technical architecture and core code for the next Alice upgrade are now largely complete. If final internal testing goes well, we expect the upgrade to go live within the next 2-3 days. Before the upgrade, we will temporarily pause the public chain and related servers for internal testing, chain upgrade work, and final stability checks. This upgrade will include:
A redesigned Worker client experience
Miner node integration directly inside the Worker client
Automatic device detection and task matching
Stronger verification, anti-cheat, and delayed settlement logic
Updated reward distribution rules
Preparation for future AI useful-compute and auxiliary mining pools
There will also be a token allocation update. From the total 21,000,000 Alice supply, the Foundation will reserve 20% for possible future financing, marketing, exchange-related activities, ecosystem development, and long-term operations. All remaining rewards will be distributed through mining. This upgrade is part of Alice’s move toward a more stable, scalable, and public-ready mining network.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol

@justinsuntron

@worldlibertyfi

#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Client modernization is not about visuals; it is about control. The value of `alice-agentd` is turning complex distributed training into an orchestrable system service with lifecycle hooks, policy injection points, and recovery surfaces. For a global compute network, this is more important than any UI refresh and often the most underestimated upgrade.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol
@justinsuntron
@worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

ALICE can support many interaction layers—Tauri GUI, CLI, ops dashboards, and managed automation—but execution authority remains anchored in agentd. This decouples interface innovation from training stability, preserving community flexibility while protecting canonical training boundaries.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol
@justinsuntron
@worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Consolidating lease, runtime, upload, and telemetry into agentd transforms client behavior from user-driven to system-driven. Users can switch interfaces without fragmenting training state. For operations, this means better uptime; for engineering, coherent upgrade strategy; for community users, clearer failure and recovery narratives.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol @justinsuntron @worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

The core of ALICE’s new client is not the GUI but the daemon: `alice-agentd`. GUI and CLI become control surfaces, while the daemon serves as the single mining authority. Training and uploads continue even when UI is closed, and Linux headless plus desktop environments share one control semantics. Manageability, observability, and recoverability define a long-lived client model.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol @justinsuntron @worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

From theory to practice, robust distributed training assumes some participants will behave adversarially. ALICE v1 turns that principle into protocol and runtime design. It is not “fix after abuse”; it is “remain stable under expected adversarial behavior.” That is the threshold for decentralized training to become mainstream infrastructure.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol
@justinsuntron
@worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

“Open networks must be low quality” is a misconception. The real question is whether zero-trust mechanisms are executable. ALICE v1 relies on verifiable contracts, not moral assumptions: identity can register, behavior can be validated, updates can be traced, and payouts can be audited. With evidence-based accounting, openness and quality can coexist.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI
@Alice_protocol
@justinsuntron
@worldlibertyfi
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Risk is no longer an operator note; it becomes a protocol object—observable, stateful, and settlement-aware. Nodes move through dynamic tracks such as normal, watch, and quarantine rather than binary labels. Combined with delayed payouts and sampled re-execution, the system enables nuanced governance without sacrificing openness. That is essential for sustainable networks.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron @worldlibertyfi #DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

ALICE’s new architecture starts from a zero-trust miner model, so anti-cheat is core design, not an afterthought. Version gates, server-issued leases, anti-replay nonces, canary challenges, sampled re-execution, risk/quarantine states, delayed payouts, and kill switches form one control plane. The goal is simple: open participation without open contamination of canonical training.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron @worldlibertyfi #DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

In open training networks, the biggest attack surface is not offline nodes but noise disguised as useful updates. ALICE v1 shifts verification before commitment: prove acceptability first, then consider aggregation and settlement. This increases relative returns for honest contributors and reduces economic upside for malicious behavior. Over time, this outperforms purely reactive bans.
#APIWar #B .AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron @worldlibertyfi #DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Metrics become meaningful only when pipeline boundaries are explicit. In ALICE v1, stages such as leased, accepted, committed, and finalized are state-machine facts, not narrative estimates. This improves fairness for miners, controllability for operators, and transparency for the community. Our aim is a training network that can grow, be audited, and be inherited over time.
#APIWar #B.AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron
@worldlibertyfi #DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Many projects chase throughput first and consistency later. ALICE v1 reverses this order: define consistency first, then scale throughput. Training errors can surface late and accumulate across epochs once state is polluted. The value of a canonical path is preventive control—blocking incorrect state transitions before they spread. That is what keeps scaling stable.
#APIWar #B.AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron
@worldlibertyfi #DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

@justinsuntron launched https://t.co/lhBrT2jcYg's pricing nuke: 90% API savings, zero-markup. @worldlibertyfi fired back with a defamation countersuit. Billionaires throwing legal punches while trying to undercut each other on inference pricing.

It's entertaining. It's chaotic. But it's also the clearest signal yet that the AI API economy is the most contested territory on Earth right now.

And here's what both sides are missing:

The future of AI APIs won't be won by the best price or the best lawyers. It'll be won by the infrastructure that can't be shut down, can't be frozen, and can't be controlled by any single entity — no matter how rich or connected.

@Alice_protocol is building that infrastructure:

→ Decentralized GPU training network
→ Real ML model training verified on-chain
→ ALICE token rewards (21M max, Bitcoin-style cap)
→ Community-owned models powering future commercial APIs
Sun's 90% discount is impressive. WLFI's legal strategy is aggressive. But when the smoke clears, the protocols that let people build, train, and own AI collectively will be the ones that matter.
Billionaires are fighting over who gets to charge you for AI.
We're building the AI that charges for itself — and shares the rewards with the people who trained it.

#APIWar #B.AI #JustinSun #WLFI #DecentralizedAI #ALICE #FutureOfAI @Alice_protocol @justinsuntron @worldlibertyfi

[WLFI @worldlibertyfi]

@WorldClawAI is expanding access to AI and $WLFI plays a key role in the ecosystem. Users can access 300+ models with WorldRouter, and agents can facilitate payments in USD1 on @BNBCHAIN and @solana to support task execution.

Locking $WLFI tokens can provide access to additional features and functionality related to the future of AI.
Subject to applicable terms and conditions.

[Alice @Alice_protocol]

A single path is not about control for control’s sake; it is a minimum condition for training correctness. In distributed systems, more paths often mean more state ambiguity and weaker reward-to-impact attribution. ALICE v1 maps each contribution to a verifiable event chain through a canonical path, enabling community review, operator accountability, and reproducible model evolution.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Canonical Update Path
A major risk in legacy systems is path ambiguity across training, upload, scoring, and promotion. ALICE v1 enforces one canonical path: lease issuance → local training → proposal + receipt upload → server validation → accepted-update aggregation → FP32 optimizer apply → canonical commit → checkpoint/publish. Once the path is singular, debugging, auditing, and rollback share the same operational language.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

In ALICE v1, upload is not commit, validation pass is not commit, and scorer pass is not commit. Only server optimizer transaction plus atomic persistence qualifies as canonical commit. We explicitly separate “apparently valid” from “state-changing finality” to prevent unexplainable state transitions and to preserve predictable outcomes for contributors.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

If phase one proved distributed training can run, the new architecture must prove it can run stably, durably, and audibly. ALICE does not hide engineering complexity; it structures it. Who submits, who validates, what becomes canonical, what gets finalized, and what gets published as checkpoints all have explicit state boundaries. This is the real boundary between experimental networks and production-grade community AI.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

One-line summary of ALICE v1: global distributed compute, centrally coordinated training correctness. No node directly controls the canonical model state. Updates become canonical only through validation, aggregation, and optimizer transactions. Open participation and quality control are not opposites; “open” and “ruleless” are. We choose open access with strict commit discipline and accountable evolution.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Vision Shift：
ALICE’s next phase is not just about onboarding more miners. It is about turning globally distributed GPUs into a verifiable, auditable, and recoverable training system. We separate open participation from model-state authority: miners provide compute and update proposals, while the server maintains one canonical model state. Compute stays decentralized; training correctness stays centralized. This is not a step back from decentralization. It is how decentralization becomes production-grade infrastructure.
#DecentralizedAl #DeAl #AlAPI #OpenSourceAl #GPUmining #RunANode #Web3Al #AliceProtocol

[Alice @Alice_protocol]

Appreciate the support, @bayc7276. Alice is being built for a future where AI training is open, decentralized, and permissionless. Thank you for seeing the vision early.

[Phillip @bayc7276]

tdy I took some time and dived into @synth_fun, 1st place winner of our latest @XLayerOfficial Hackathon.

What got me excited is Alice, a decentralized AI training network. Anyone can contribute compute, no pre-mine, no team allocation, 100% mined. Model weights owned by the network, not some coop like anthropic or openai.

Why this matters: training a serious AI model today costs $100M+, heavy investment for sure, and is controlled by a few giants. Still projects like Bittensor already proved decentralized training works, the trend is clear.

Alice takes a similar approach but ties compute funding directly to onchain activity — usage fuels training. Simple and smart.

We need more of this in crypto. Not AI-themed tokens, but real networks producing real intelligence.

Not trying to be this super official Built on X Layer with OKX Onchain OS 🔥