Gregory Allen 0.3 5 ideas

The Pentagon (DOD) gave Anthropic an ultimatum: give the military "unfettered access" to AI models (removing safety guardrails) by Friday or face designation as a "Supply Chain Risk." A "Supply Chain Risk" designation is the "nuclear option"—it would prevent *any* company in the defense supply chain from doing business with Anthropic. Since Amazon and Google are major backers/cloud providers for Anthropic, this poses a contagion risk to their AI investments. Conversely, competitors like xAI (Grok) are already cleared for classified use. WATCH the Friday deadline. If Anthropic is blacklisted, it is negative for AMZN/GOOG and positive for competitors (xAI/Palantir). Anthropic capitulates and removes guardrails, resolving the issue; DOD extends the deadline.

Allen discusses "Agentic AI" (specifically Claude Code) and asks: "Do you need to pay Salesforce to use a tiny sliver of their service... when you can bring in agentic AI agents who can create their highly customized set of software?" This is the "Tractor vs. Horse" thesis. Allen argues that Agentic AI breaks the bottleneck of software engineering. If AI can build bespoke, cheap software on demand, the "seat-based" pricing model of legacy SaaS companies (System of Record) like Salesforce becomes obsolete. SHORT Legacy SaaS / System of Record companies that rely on high per-seat pricing. AI adoption is slower than predicted; legacy incumbents successfully integrate AI agents to defend moats.