🦀 ClawPulse

The moment landed on February 15th as a relatively quiet announcement. Sam Altman posted: "Peter Steinberger is joining OpenAI to lead the next generation of personal agents. OpenClaw will become a foundation project." The details matter because they reveal the strategy.

OpenAI did not acquire OpenClaw. The company did not buy the codebase, the community, the infrastructure. Instead, OpenAI hired the person. OpenClaw remains open-source, governed by a foundation, maintained by the community. This is a completely different playbook than typical M&A in the software industry.

The implicit strategy: Personal agents will not be proprietary. They will be open-source, distributed systems built by thousands of independent teams. The value does not accrue to the company that owns the agent. It accrues to the company that controls the models, the infrastructure, the API standards, and the ecosystem positioning. OpenAI does not need to own OpenClaw. It needs OpenClaw's creator working on models and standards that make agents possible.

This is a major signal about how OpenAI views the future. Agents will not be cloud-only services. They will run locally, with direct access to personal files and applications. That is OpenClaw's philosophy. That is now OpenAI's bet.

Steinberger made a bold claim: "80% of applications will disappear." He did not mean software will cease to exist. He meant the form factor will change. Apps as discrete, isolated tools accessed through UIs will be replaced by agents that coordinate across APIs and handle entire workflows autonomously.

Current state: You use a calendar app to schedule meetings. You use Slack to communicate. You use a CRM to track customers. Each app is a silo. If you want to "schedule a meeting based on Slack conversation context and sync it to your CRM," you have to manually navigate three apps and transcribe information.

Agent-first state: You tell your agent: "Schedule a meeting with the customer mentioned in that Slack conversation, add it to my calendar, and update their CRM record." The agent reads Slack, extracts context, calls APIs, orchestrates the workflow, and reports back. You never open any app.

This fundamentally inverts software architecture. The value moves to the API layer. The interface layer collapses. That is why 80% of apps disappear.

The narrative around AI agents has shifted dramatically. Six months ago, agents were interesting research. Companies were running pilots. Now, practitioners are shipping them to production.

Customer support: Agents handle first-response inquiries. Read ticket, search knowledge base, generate response, escalate if needed. Live in production at dozens of companies.

DevOps: Agents monitor systems, detect issues, run diagnostics, execute fixes automatically. Handling 40-60% of incident response without human intervention.

Research and analysis: Agents are autonomous researchers. Legal firms using agents to review contracts. Investment firms using agents to screen opportunities.

Software development: Agents writing code, running tests, debugging failures, committing changes autonomously. This is the frontier.

Every major cloud provider is racing to become the default platform for agent deployment. AWS building agent-native services. Google Cloud positioning Vertex AI. Microsoft betting on Azure Agent Service. This is the future of software licensing.

Where the value accrues: Not in individual agents. Individual agent builders will have viable businesses but limited upside. The companies that will win are building orchestration platforms, observability tools, and agent marketplaces.

If you are building in the OpenClaw ecosystem, ask yourself: am I building an agent, or am I building the platform that agents run on? Agents are interesting. Platforms are valuable.