The OpenClaw Effect

What Claude Dispatch and Manus "My Computer" reveal about the new speed of AI product development and why community-driven innovation is replacing the traditional SaaS cycle

Author

Justin Wright
March 23, 2026 • Est. Reading Time: 9 minutes

Happy Monday!

In January, Peter Steinberger launched OpenClaw as a side project. Within 72 hours it had 60,000 GitHub stars. By March, it surpassed React as the most-starred project on GitHub with over 250,000 stars and 180,000 deployments. Jensen Huang said every company needs an "OpenClaw strategy."

Last week, Anthropic launched Claude Dispatch. The same week, Meta's Manus shipped "My Computer." Both products let you control AI agents on your desktop from your phone, both execute multi-step tasks locally, both include the safety guardrails OpenClaw notoriously lacked, and both clearly draw from the playbook OpenClaw wrote.

Anthropic shipped its answer to OpenClaw faster than OpenAI, which hired the guy who built it. The meme going around is accurate: "Anthropic built OpenClaw before OpenAI could finish onboarding Steinberger."

What's happening here is bigger than any single product launch. The traditional SaaS cycle of identifying market demand, building features, testing, and shipping used to take 12 to 18 months. OpenClaw compressed that to about 60 days. An open-source project proved massive demand existed, and two of the largest AI companies in the world responded with polished alternatives before most enterprise software teams would have finished their first sprint planning meeting.

OpenClaw proved 180,000 developers want personal AI agents badly enough to deploy insecure prototypes. Within weeks, Anthropic shipped Claude Dispatch and Meta launched Manus "My Computer," both mirroring OpenClaw's core features with enterprise-grade security. The speed tells the real story: community-driven open-source projects are now setting the product roadmap for billion-dollar AI companies, compressing what used to be 12-month development cycles into weeks.

TL;DR

What They Actually Shipped

Claude Dispatch connects your phone to your desktop through a single persistent conversation thread. You text Claude from the iOS or Android app, and the desktop agent executes tasks locally: organizing files, preparing reports, pulling data from Google Drive, handling email. The agent runs in a sandbox, files stay local, every destructive action requires explicit approval via push notification, and it's end-to-end encrypted. The feature is currently macOS only, rolling out first to Max subscribers ($100-200/month) with Pro access ($20/month) coming shortly.

Manus "My Computer" takes a similar approach with key differences. It launched on both macOS and Windows and executes through CLI commands, giving it deeper system access. In a demo, the team asked the agent to build a real-time translation app entirely through terminal commands. Twenty minutes later: a working Mac application was delivered with no code written by hand. Manus can also leverage local GPU resources for ML training or LLM inference. Pricing starts at $20/month with a limited free tier.

Both products address the same insight OpenClaw surfaced: people want AI agents that work continuously on their local machines, interact with their actual files and applications, and can be directed remotely. The difference is that OpenClaw shipped with zero-click exploits and 42,000 unsecured instances. Dispatch and My Computer shipped with permission systems, sandboxing, and encryption all built in.

The 60-Day Product Cycle

This timeline should make every product leader uncomfortable. Steinberger launched OpenClaw in January. By late February, Anthropic had Dispatch in development and by March 17, it shipped as a research preview. Meta's Manus, acquired in December 2025 for $2 billion, launched My Computer the next day.

Traditional SaaS product development doesn't work this way. Designli's 2026 field report found that 66.7% of founders now optimize for speed to market as their dominant strategy, and 55.6% of SaaS leaders actively adjust roadmaps based on AI trend signals. But even those numbers undersell what happened here. Anthropic and Meta didn't just adjust their roadmaps, they watched an open-source project prove demand in real time and shipped competitive products within weeks.

AI development tools have collapsed the distance between prototype and product. What used to take 12 months from MVP to launch can now happen in three to four months. What Anthropic and Meta did was even faster because they weren't starting from zero. They had the infrastructure already, and OpenClaw gave them the market validation they needed to prioritize and ship.

Community as the New Product Team

The deeper shift is who drives the product roadmap now. In traditional SaaS, product managers run discovery interviews, analyze usage data, write specs, and prioritize features through quarterly planning cycles. That process still exists, but OpenClaw demonstrated something different: 250,000 GitHub stars is a demand signal no amount of user research can replicate.

OpenClaw's community didn't just use the product. They extended it across fifteen platforms, built integrations nobody planned for, and surfaced use cases the creator never imagined. Moltbook, the AI-only social network with 1.4 million agents, wasn't on anyone's roadmap. It emerged because the community had the tools and the freedom to experiment.

The security failures were equally instructive. When researchers found zero-click exploits and Cisco called OpenClaw "an absolute nightmare," that was a public specification for what enterprise versions needed to solve. Anthropic's sandbox model and Manus's permission system are direct responses to vulnerabilities the community identified.

This is the new pattern: open-source communities build the prototype, surface the demand, stress-test the security, and define the feature set. Companies with infrastructure and distribution then ship the polished version. The community functions as an unpaid, massively parallel product team operating at a speed no internal organization can match.

The Trade-Off Nobody Talks About

The speed is exciting, but the trade-off is real. OpenClaw runs locally, supports fifteen messaging platforms, works with any model, and costs nothing. Claude Dispatch locks you into Claude, runs only on macOS, requires a subscription, and operates in a single conversation thread. Manus ties you to Meta's model stack.

One developer on Hacker News framed it as "Linux vs. Windows." OpenClaw gives you power and flexibility at the cost of managing your own security. Dispatch and My Computer give you safety and polish at the cost of control and vendor lock-in. The community that proved the demand for personal AI agents may find that the enterprise versions of their idea don't actually serve them.

For practitioners, the strategic question is where you sit on that spectrum. If you're building for enterprises that need guardrails, Dispatch and My Computer just validated the category. If you're building tools or integrations, the OpenClaw ecosystem's model-agnostic approach offers more surface area. If you're building products in any category adjacent to AI agents, the 60-day cycle from open-source signal to enterprise product is now the pace you're competing against.

The Bottom Line

OpenClaw didn't just go viral, it rewrote the product development playbook for the largest AI companies in the world. The fact that Anthropic and Meta both shipped competitive products within 60 days of OpenClaw's explosion tells you everything about where product cycles are heading.

The traditional SaaS rhythm of annual roadmaps and quarterly releases is being replaced by something more fluid: communities surface demand, open-source projects prove the concept, and companies with infrastructure ship polished versions at a speed that would have been unthinkable two years ago. For practitioners, this means the window between "interesting open-source project" and "feature in a product you compete with" has collapsed from years to weeks.

Whether or not this pace is sustainable, your organization needs to keep up.

In motion,
Justin Wright

If a single developer's open-source project can set the product roadmap for Anthropic, Meta, and OpenAI simultaneously, what does that tell us about where innovation actually originates in AI?

Food for Thought

If you haven’t listened to my podcast Mostly Humans: An AI and business podcast for everyone yet, new episodes drop every week!

Episodes can be found below - please like, subscribe, and comment!