H-FARM EVENTS ON AI

ClawCon Venice — The Personal AI Festival

ClawCon is the gathering for the Internet's most curious personal AI community and it's coming to the H-FARM Campus. What started as a niche OpenClaw meetup quickly turned into a global movement: builders, researchers, designers, and curious humans meeting IRL to test what happens when AI becomes open and social-first.

ClawCon Venice is Italy's first ClawCon, hosted right here on the H-FARM Campus. Save the date: Thursday 4 June, 18:00–21:00. Here's what you'll find:

• A global community of OpenClaw builders, researchers, and designers meeting IRL, plus live demos of open, social-first AI projects.

• Unfiltered, hands-on conversations on where personal and agentic AI is heading next, in the OpenClaw spirit: curious, open, and refreshingly social.

Whether you build, research, design, or are simply AI-curious: there's a seat for you.

🧮 AlphaProof Nexus solves 9 open Erdős problems 🧮
DeepMind's system verified 9 Erdős solutions at speed, outpacing OpenAI's single result.

Google DeepMind's AlphaProof Nexus has solved 9 open Erdős mathematics problems — including two that remained unsolved for 56 years — along with 44 open OEIS conjectures. The system pairs AI models with the Lean proof assistant, which verifies each logical step to eliminate hallucinations. Some solutions reportedly cost only a few hundred dollars to generate. The announcement landed just one day after OpenAI claimed its own single Erdős result, sharpening the rivalry between the two labs. This signals a capability jump in formal mathematical reasoning, moving AI beyond benchmarks into genuine open research problems spanning combinatorics and graph theory.

🛡️ Anthropic readies Mythos 1 for Claude Code 🛡️
Mythos 1 rollout signals Anthropic's dual-track strategy for enterprise security and coding.

Anthropic appears close to a broader release of Mythos 1, a model already spotted on Google Cloud and AWS through vulnerability discovery programs. The model is being positioned specifically for Claude Code and Claude Security use cases, aimed at protecting organizations from emerging threats. A general rollout seems imminent, while Claude Opus 4.8 is also rumored to be in development. This dual-track approach — pairing frontier reasoning models with specialized security tools — could differentiate Anthropic in the competitive enterprise market. If confirmed, the expansion adds meaningful competitive pressure against OpenAI and Google in enterprise AI security.

⚠️ OpenAI's $445K safety role targets self-improving AI ⚠️
OpenAI's new hire addresses regulation concerns as recursive self-improvement grows more plausible.

OpenAI is hiring for a new safety position with compensation up to $445,000, focused on preparing for AI systems capable of self-improvement. The role involves studying risks including AI-generated cyberattacks, hidden model behavior, and measuring how much technical work AI can autonomously automate. The hire comes as leaders like Sam Altman and Demis Hassabis predict autonomous AI researchers could emerge within the next few years. The high salary reflects the urgency OpenAI places on safety infrastructure — a concrete investment addressing the tension between racing toward more capable systems and ensuring adequate safeguards against recursive self-improvement.

INTERESTING TO KNOW

🖥️ Antigravity 2.0 builds an OS with 96 agents 🖥️

Google's Antigravity 2.0 showcases a speed milestone in multi-agent development: 96 AI agents collaborated to build a functioning operating system from scratch in just 12 hours, spending under $1,000 in token costs. The resulting OS can even run the classic game Doom. The project demonstrates how multi-agent architectures are enabling complex system-level engineering at a fraction of traditional costs, signaling a capability jump for agentic software development.

💰 DeepSeek cuts V4 Pro pricing by 75% permanently 💰

DeepSeek has made its 75% pricing discount on V4 Pro permanent, dropping costs to $0.435 per million input tokens and $0.87 per million output tokens. Originally set to expire this month, the cut now significantly undercuts OpenAI's GPT-5, Anthropic's Claude Opus 4.7, and Google's Gemini 3.5 Flash for enterprise reasoning workloads. This intensifies pricing pressure across Western AI labs and may accelerate enterprise adoption of Chinese AI models for cost-sensitive applications.