AI Agents Eating Software
The AI infrastructure spending spree may slow, but it is not going to stop. So long as LLM performance is improving (albeit at a lesser rate), the spending will continue (especially with the 10YR at 4.23%). On the Agentic AI side, Claude Code (Anthropic) has enjoyed immense success with developers, which makes sense given that Claude Opus 4.5 is the best programming model (my experience). We leveraged Opus 4.5 over the past 4 weeks to:
Onboard our S&P 500 CEO data set,
Add a paywall to our UI and API service,
Build an AI agent that digitally markets CEORater to buyside AI agents,
Introduce our new RevenueCAGRScore metric,
Build a “News & Insights” (subscribe to our mailing list) component to CEORater,
Build an email automation for our “News & Insights” content which sits on top of Google Cloud and costs a fraction (pennies per month) of what Beehiv, Letterhead and Mailchimp cost, while enabling CEORater to maintain the original format of our articles along with our branding.
Now we are seeing entrepreneurs build AI agents such as OpenClaw, which leverages Claude Code for the “brain”, OpenAI Codex for the execution, with multiple other plugins, controlled via messaging apps to perform various automations, all while sitting on a Mac mini. Anthropic, OpenAI and agents like OpenClaw have the potential to be very disruptive to SaaS companies by democratizing the building of both Software applications as well as AI-powered automations that run 24x7, the latter which obviates the need for software.
CEORater “News & Insights”: Founder CEOs: Exceptional Performance, Modest Pay (we built this article with insights uncovered by our AI chat which overlays our CEO data)
