H-FARM PROGRAMS ON AI

AI Super Power

The Professional Master to bring AI into your department

AI Super Power is the Professional Master by H-FARM Business School for managers who want AI inside their workflows, not just in a browser tab.

In 85 hours you design no-code automations and AI agents, with a faculty that includes Diego Pizzocaro and Alessandro Scandaletti of H-FARM AI.

Starts on 27 October 2026 on the H-FARM Campus in Roncade. Part-time and taught in Italian, with in-person weekends and online evening sessions, for managers, department heads and entrepreneurs ready to lead AI adoption. Last Early Bird seats available.

• Hands-on workshops on real business briefs: data analysis, autonomous agents, multi-agent orchestration, AI Act compliance and security.

• You leave with an AI agent trained on your company’s data, an adoption roadmap for your team and an H-FARM Business School Open Badge.

🔒 ChatGPT adds locked chats for private conversations 🔒
OpenAI rolls out a new privacy feature that hides sensitive conversations from history and search.

OpenAI has shipped a new feature in ChatGPT that lets users lock individual chats, effectively hiding them from regular chat history, pinned conversations, and the platform's search function. Locked chats remain accessible only through deliberate user action, keeping sensitive or personal conversations separate from everyday use. The update addresses a long-standing user request for better privacy controls — a signal that OpenAI is prioritizing trust as ChatGPT handles increasingly personal tasks. This matters because it shifts ChatGPT closer to a secure personal assistant, which could accelerate adoption in professional and enterprise settings where data sensitivity is a concern.

🎙️ ElevenLabs ships Eleven Realtime voice agent models 🎙️
New speech-to-speech models target sub-second latency for real-time voice agents.

ElevenLabs released Eleven Realtime, a new class of speech-to-speech models purpose-built for ultra-low-latency voice agent applications — from digital assistants and AI tutors to interactive gaming characters. Instead of stitching together separate transcription, reasoning, and synthesis components, ElevenLabs tunes speech recognition, turn-taking, the LLM, and voice end-to-end, reaching a time to first audio of 570–620 ms while preserving natural intonation and expressiveness. The models are available in ElevenAgents and through a Realtime API compatible with OpenAI's Realtime API: developers can switch by changing the endpoint and model name. The release creates direct competitive pressure on OpenAI and Google's voice capabilities, particularly for developers building always-on conversational products where every millisecond of latency impacts user experience. For enterprises deploying voice AI at scale — call centers, healthcare triage, customer support — faster and more natural-sounding models could meaningfully shift vendor choices.

🧬 Claude designs scorpion antivenom on a $200K budget 🧬
A non-scientific firm used Claude to replace traditional drug discovery pipelines at remarkable speed.

Citrini Research, a financial media firm with no in-house scientists, used Anthropic's Claude to design 211 potential scorpion antivenom protein binders over three months on a $200K budget. Of those, 17 designs passed binding tests in laboratory validation — though none have yet been tested to neutralize actual venom. Anthropic published accompanying research on the project. The experiment represents a significant capability jump for AI-assisted drug design: it demonstrates that non-experts can leverage frontier LLMs to generate viable therapeutic candidates, potentially reshaping early-stage pharmaceutical research access and pricing for neglected diseases.

INTERESTING TO KNOW

⚙️ TorchTPU brings PyTorch natively to Google TPUs ⚙️

Google and Meta have jointly launched TorchTPU, an open-source PyTorch backend that lets developers run and migrate models directly on Google's TPU accelerators. The project supports both training and inference through optimized eager execution and torch.compile integration, eliminating much of the friction that previously kept PyTorch-native teams locked into GPU workflows. This is a competitive pressure play: by making TPUs a first-class PyTorch target, Google lowers the switching cost for the massive PyTorch developer community currently reliant on NVIDIA GPUs. For AI teams, it means more hardware options and potentially lower cloud compute pricing as TPU access becomes seamless.

💻 Apple's October 28 event targets touchscreen MacBook Pro 💻

The rollout of Apple's first touchscreen MacBook Pro is expected at an October 28 event taglined "Pro and tell." Bloomberg reports the headliner will feature an OLED display, Dynamic Island, and the new M5 Pro/Max chips. The move marks Apple's long-rumored convergence of iPad and Mac input paradigms, a shift that could reshape how professionals interact with macOS and creative software.