Loading Runway...
Loading Runway...
What AI is doing to the work — with evidence on every claim.
For Software Engineers: The Throughput Divide Deepens
The market is crystallizing around two competing pressures: enterprises are aggressively deploying AI coding agents (Codex, Claude Opus 4.8, Cursor) to compress requirements analysis from weeks to hours—Endava and Cisco are shipping agentic workflows at scale—while simultaneously discovering that wholesale AI adoption without engineering rigor backfires (ClickUp's 22% workforce cut signals premature over-automation).
Claude Opus 4.8's 3X cheaper fast mode and OpenAI Codex's enterprise traction at scale mean the productivity multiplier is real, but engineers who can architect *when* to apply agents versus when to ship traditionally will outcompete those betting on replacement.
Meanwhile, DeepSeek's permanent 75% price cut on V4 Pro is eroding the token-cost moat that justified Western LLM vendor margins—watch for downstream pricing pressure on Copilot and Claude API seats within Q1.
What matters now
What it means: Boston Children's Hospital deployed OpenAI technology to help diagnose more than 40 rare disease cases, demonstrating AI's ability to improve diagnostic accuracy and reduce clinician burden. This concrete deployment shows AI can augment and accelerate rare disease diagnosis workflows that typically require expert physician interpretation.
What it means: Merck deployed AI agents that reduced drug discovery cycles by one-third and accelerated compliant marketing material production by 80%, enabled by strong underlying infrastructure. These concrete, measurable improvements in specialized workflows demonstrate significant capability gains in both research and compliance-heavy functions.
What it means: Cognition's $1B funding round at $26B valuation reflects massive investor confidence in Devin, an AI agent that automates programming tasks for engineers. This direct automation of core engineering workflows signals imminent competitive pressure on software development roles, particularly for routine and junior-level coding.
What it means: OpenAI, Thrive, and Crete deployed a self-improving tax agent using Codex that automates tax filing, improves accuracy, and accelerates workflows. Tax preparers and accountants face direct automation of core filing and compliance functions, with the agent's self-improvement capability suggesting ongoing capability expansion.
What it means: DeepSWE benchmark reveals GPT-5.5 significantly outperforms competing models on autonomous code generation tasks, breaking the previous performance plateau. This concrete capability leap demonstrates measurable progress in AI-assisted and autonomous software development, directly threatening junior and mid-level developer roles.
What it means: AI is being deployed to automate debt collection calls, replacing human collectors in what is described as an extremely undesirable role. This represents direct workforce displacement in collections as AI systems handle customer interactions and payment processing autonomously.
What it means: ClickUp cut 22% of its workforce citing replacement by AI agents, according to commentary on overly aggressive AI adoption without understanding actual job scope. This is direct evidence of workforce reduction attributed to AI agents replacing specific operational roles.
On the horizon
AI capabilities closing on their next threshold. When one crosses, what AI can do against the tasks it touches steps up — ranked by how soon.
→ AI matches senior engineers
→ AI provides structured feedback
→ AI conducts simple negotiations
→ AI matches expert analysts
→ AI matches human office workers
→ AI generates brand-consistent design systems
Signals are public. Your Brief is private.
Free · 4 min · Task Exposure Map + 30-day Defense Plan, personalised to your role.