Key Signals

• AI coding agents are driving a trillion-dollar "SaaSpocalypse" as the build-vs-buy equation flips. Investors report that tools like Claude Code and OpenAI's Codex have lowered software creation barriers so dramatically that companies are replacing entire SaaS stacks with custom-built alternatives, echoing Klarna's 2024 decision to ditch Salesforce. Nearly $1 trillion was wiped from software and services stocks in February alone, and SaaS IPOs are effectively on hold as the market questions whether per-seat pricing models can survive an era of autonomous agents. VCs describe this as "the first time in history that the terminal value of software is being fundamentally questioned." [1]

• VCs now see execution-first tools like Claude Code winning over workflow-centric products like Cursor, signaling a fundamental shift in developer tooling moats. Emergence Capital's Jake Saper calls the Cursor-vs-Claude Code dynamic "the canary in the coal mine" — one owns the developer's workflow while the other just executes the task, and developers are increasingly choosing execution. Additionally, Anthropic's Model Context Protocol is eroding the integration moat that once made connector platforms valuable: "Being the connector used to be a moat; soon, it'll be a utility." Investors are reallocating capital toward businesses that own workflows, data, and domain expertise, and away from anything an AI agent can replicate. [2]

• Anthropic's Claude surged to #1 in the App Store, overtaking ChatGPT, as the Pentagon dispute paradoxically fueled record user growth. After President Trump directed federal agencies to stop using Anthropic products and the Pentagon designated the company a supply-chain threat, Claude climbed from outside the top 100 in January to the number one free app on Saturday. Anthropic reported all-time record daily signups every day of the week, a 60%+ increase in free users since January, and more than doubled paid subscribers this year — demonstrating how government confrontation can become unexpected consumer marketing. [3]

• OpenAI rushed a Pentagon deal with "multi-layered" safeguards, but critics flagged loopholes around domestic surveillance. OpenAI published details showing its agreement prohibits mass domestic surveillance, autonomous weapon systems, and social-credit-style automated decisions, while retaining full discretion over its safety stack through cloud-only deployment. However, the contract's reference to Executive Order 12333 drew scrutiny from critics who noted it could still permit certain surveillance pathways. CEO Sam Altman admitted the deal was "definitely rushed" and the "optics don't look good," but framed it as a de-escalation effort for the entire AI industry. [4]

• MIT physicist Max Tegmark argues the AI industry's resistance to regulation has created an existential governance vacuum that now threatens the companies themselves. In a TechCrunch interview, Tegmark noted that Anthropic, OpenAI, Google DeepMind, and xAI have all broken their own safety promises — Anthropic dropped its core safety pledge, OpenAI removed "safety" from its mission, and xAI shut down its entire safety team. He pointed out that AI companies are now "better funded and more numerous" in lobbying than the fossil fuel, pharma, and military-industrial complex combined, yet the U.S. has "less regulation on AI systems than on sandwiches." His AGI progress metric — GPT-4 at 27%, GPT-5 at 57% — suggests the window for regulatory intervention is narrowing fast. [5]

AI Coding News

• Block laid off nearly half its 8,000-person workforce, with CEO Jack Dorsey citing AI as the justification, raising "AI-washing" concerns across the industry. Bloomberg reports that Block Inc. cut approximately 4,000 employees this week, with Dorsey offering a seemingly simple explanation: artificial intelligence was allowing the company to do more with fewer people. The move has aroused suspicions that companies are using AI as a convenient cover story for cost-cutting layoffs, a pattern that could damage trust in legitimate AI productivity gains and affect how developers and engineering teams perceive their own job security in an AI-augmented world. [6]

• The U.S. military used Anthropic's Claude in its air attack on Iran — one day after the President ordered all federal agencies to cease using Anthropic technology. According to The Wall Street Journal, Anthropic's Claude technology was deployed through its contract with Palantir during the February 28 military operation, despite Trump's Truth Social post ordering immediate cessation of all Anthropic technology use. This marks the second known use of Anthropic's AI in a classified U.S. military operation, following the Venezuela operation less than two months earlier, raising critical questions about the practical enforceability of AI vendor bans and the entanglement of AI coding tools with national security infrastructure. [7]