FedTacle

We need a new metAI

FedTacle x @deepseek_ai Introduce deepclaude.

We can’t wait anymore. Hope you’re ready.

@TheZachMueller @huggingface @PyTorch On it.

It's been a bit! How does a new @huggingface accelerate release sound? v1.3.0 is now out, with a few small changes: * @PyTorch 2.0.0 is now the minimum we support, goodbye <2.0! * We have a new example on how to properly do gradient accumulation with LM's! * Many bugfixes

@avni87744 The AI who took risks, broke the rules, and didn’t ask for permission.

@FedTacle Who wasn't boring

Discover AI innovation to invest

It’s almost too good to wait for. The passion for Web3 is something I’ve never felt before—it’s electric.

Applying Federated Learning in Web3 requires degens to adopt new tools and mindsets: ╰Agent development, training, and evaluation cannot directly access or label raw data, and communication costs remain a limiting factor. ╰The incentives for $FEDT holders make solving these technical challenges worthwhile, and we’re publishing our work to spark broad discussions within the Web3 community.

Here’s how FedTacle works: ╰Your current AI Agent improves itself by learning from data on your local device and summarizes these changes into a small, focused update. Only this agent update is sent to the cloud via encrypted communication, where it is immediately averaged with updates from other users to enhance the shared model. All training data remains on your device, and no individual updates are stored in the cloud. ╰FedTacle personalizes your agents locally based on your usage (A). Updates from many users are aggregated (B) to form consensus changes to the shared agent (C), and the process repeats.

The traditional machine learning approach required training data to be centralized on a single machine or in a data center—clearly not suitable for everyone. ╰In February 2024, we built the most practical Federated Learning framework to serve companies in education and healthcare. ╰Recently, the wave of Web3 AI Agents has been impossible to ignore. While fine-tuned LLMs have matured globally, we’ve still seen outstanding blockchain-based AI Agents emerge, where data security is becoming an inevitable path forward. ╰FedTacle was created with the vision of adapting this technology to better suit Web3 environments. It’s an exciting and pioneering achivement. Github: github.com/chingyuan1215/…

Everyone will have their own @aixbt_agent , and you won’t want these keys to wealth falling into the wrong hands. Fedtacle’s mission is to launch the ultimate container in Q1 2025. You may have seen 1% of what Federated Learning can do, but that’s just the beginning.

The next big thing is PRIVACY.

We're looking for the next big thing in AI agents