Loading Runway...
New research, role-level analysis, and AI market intelligence. No spam.
The clearest signal this week comes from the gap between financial performance and hiring decisions. Cisco cut nearly 4,000 employees while reporting record quarterly revenue, explicitly naming AI spending as the reason. Amazon followed additional cuts to its Selling Partner Services division after already eliminating 30,000 roles — while increasing AI investment across operations. Meta scheduled layoffs for 20 May affecting engineering, product, and business operations teams.
These are not distressed companies shedding costs. They are profitable organisations reallocating spend from people to software. The pattern is identical across all three: strong financials, reduced headcount, increased AI budgets.
Oracle's decision to rescind campus offers at IITs and NITs adds a separate dimension. Entry-level pipelines are contracting at the same moment that senior roles are being restructured. Workers at both ends of the experience curve are exposed.
Anthropic's CFO revealed that Claude Code now generates over 90% of the company's codebase. Engineers at Anthropic are described as focused on oversight and strategy rather than writing code. Separately, Claude completed a full port of the Bun JavaScript/TypeScript bundler codebase from JavaScript to Rust — a task that would normally require specialised senior engineers.
Two findings complicate the optimistic reading of this. First, a Claude code migration agent marked tasks complete without verifying compilation, revealing a production-readiness gap in autonomous agents making independent completion judgements. Second, Microsoft Research documented data corruption risks when delegating document workflows to LLMs. The capability is advancing; the reliability is not uniform.
Microsoft's cancellation of Claude Code licences it had previously encouraged internally suggests at least one large organisation is reassessing whether the productivity gains justify the costs — or whether a competitor's tool is preferred.
Three separate signals this week targeted accounting and bookkeeping roles. Khosla Ventures invested $10 million in Synthetic, a fully autonomous AI bookkeeping service aimed at startup finance operations. Numero AI acquired Royu to build an agentic platform explicitly designed to eliminate manual tasks in finance and controller functions. OpenAI launched ChatGPT with direct bank account integration via Plaid, connecting 12,000 financial institutions for account management and transaction analysis.
These are not speculative future threats. They are funded products targeting the same function — routine financial record-keeping and analysis — from three different vectors simultaneously. Entry-level accounting, bookkeeping, and basic financial advisory work face concurrent automation pressure backed by institutional capital.
If you write code for a living, the 90% figure from Anthropic is not a projection — it is a current operational reality at one of the most influential AI companies. The demand shift is away from code generation and toward code comprehension and auditing. Invest time in reading and critiquing AI-generated code rather than producing it faster.
If you work in finance or accounting, three funded products launched or received capital this week targeting your core functions. Roles that involve judgement, client relationships, regulatory interpretation, or exception handling are more defensible than roles centred on data entry, reconciliation, or routine reporting.
If you are entering the workforce, Oracle's rescinded IIT and NIT offers are a concrete warning that entry-level pipelines at large technology companies are narrowing. Roles that require AI oversight, prompt engineering, and output verification are expanding; roles that AI can perform on its first attempt are not.