We built OpenClaw Ultron to replace 20 people at our company | E2246

Key Concepts

• Local AI Deployment: The core strategy of OpenClaw Ultron and ExoLabs focuses on running Large Language Models (LLMs) locally on commodity hardware (Apple Silicon) for data sovereignty, control, and avoiding vendor lock-in.
• AI-Driven Productivity Enhancement: The overarching goal is to leverage AI to significantly increase individual and team productivity, freeing up human workers for higher-level tasks.
• The Shifting AI Landscape: The “model layer” of AI is maturing, with open-source models closing the gap with closed-source alternatives, shifting focus to hardware and efficient deployment.
• Emerging Business Models: Companies like ExoLabs are pioneering new business models around local AI deployment, offering both open-source and enterprise subscription options.
• Hardware as a Key Differentiator: Apple Silicon, particularly Mac Studios, is currently favored for its cost-effectiveness and performance, with RDMA technology enabling scalable deployments.

OpenClaw Ultron: Replicating Human Work with AI

The project, OpenClaw Ultron, launched by Launch, aims to replicate the work of 20 employees – encompassing an estimated 100-200 individual skills – using a single AI agent, termed a “replicant.” This isn’t about job replacement, but about automating “chores” to allow employees to focus on strategic tasks within Launch’s venture capital and podcast production businesses. OpenClaw builds upon previous iterations, Maltbot and Clawbot, and emphasizes a modular approach to AI capabilities through the development of individual “skills.” A custom dashboard, developed by Oliver Cororsan, provides visibility into the system’s operations, addressing the “black box” nature of the AI.

The Rise of Local LLMs & Hardware Considerations

A central theme is the shift towards running Large Language Models (LLMs) locally, driven by concerns over data sovereignty and control. Jason Calacanis articulated the concern about entrusting sensitive data to profit-seeking companies. While open-source models are improving, the focus is on optimizing the hardware backend technology (HBT) and stateful workflows. Currently, the most cost-effective solution is considered to be two Mac Studios with 1TB storage (approximately $20,000 total), leveraging Apple Silicon’s performance and stable pricing. Sufficient RAM (“hot in memory”) is prioritized over storage capacity. Apple’s recent implementation of RDMA (Remote Direct Memory Access) via Thunderbolt 5 is a key advancement, enabling low-latency memory sharing and effectively creating a larger, unified GPU. Scaling options include scaling out (multiple instances) and scaling up (connecting devices), with current RDMA limits around four devices, but supporting ad-hoc mesh networking for larger deployments. ExoLabs currently manages an HPC cluster of over 100 Mac Minis.

ExoLabs: A Business Model for Local AI

ExoLabs offers software to facilitate running open-source LLMs on local hardware. They operate a dual model: an open-source core for “prosumers” and an enterprise offering with support, compliance features, and a licensed subscription. Subscription costs start at $2,000 per year for a single Mac Mini and scale based on the number of nodes. The company, a team of seven engineers based in London, has recently secured venture funding (details undisclosed). They’ve seen deployments ranging from individual users to clusters exceeding 100 Macs. They are focused on metrics like MRR (Monthly Recurring Revenue), CAC (Customer Acquisition Cost), and LTV (Lifetime Value).

Skill Development & Automation Examples

The OpenClaw project utilizes “cron jobs” to automate tasks. Examples of developed skills include:

• Guest Booking: Automating podcast guest sourcing, research, and calendar management (initial testing revealed issues with email subject lines).
• Attendance Tracking: Automating daily check-in/check-out processes.
• Self-Optimization: An AI skill analyzing the OpenClaw setup and identifying issues like time zone bugs.

Skill development follows an iterative process: defining a task, prompting the AI to create a skill, testing, and refining the prompt based on feedback. Dashboard creation was achieved by prompting OpenClaw to recreate a template based on a screenshot.

Next Visit AI Pitch & Industry Review

Ryan Enelli, CTO and co-founder of Next Visit AI, pitched an AI-powered scribe and documentation platform for behavioral health clinicians. Next Visit AI has 311 users, 68 paying customers, $9,000 MRR, a CAC of $189, and an LTV of $1,700, targeting a $2 billion TAM with a 5% capture goal. The speakers also reviewed the HBO show "Industry," praising its realistic portrayal of finance and complex characters.

Conclusion

The presented information highlights a significant shift in the AI landscape. The focus is moving beyond simply accessing powerful models to controlling the infrastructure and maximizing productivity through local deployment. Apple Silicon is emerging as a key hardware platform, and companies like ExoLabs are pioneering new business models to make local AI accessible. The overarching theme is leveraging AI not as a replacement for human workers, but as an extension of their capabilities, enabling them to focus on higher-value tasks and achieve significant productivity gains. The success of projects like OpenClaw Ultron and the emergence of companies like ExoLabs suggest a future where AI is deeply integrated into workflows, empowering individuals and organizations to operate more efficiently and effectively.