Anthropic engineering lead: most people's AI ambitions are too small — aim higher, deeper, wider
AI寒武纪 · wechat · 2026-10-19
Felix, Anthropic's engineering lead for Claude.ai, Cowork, and Claude Code Desktop, argues most people use frontier models for trivial tasks, and the gap comes down to ambition. He outlines three leaps:
- Height: push tasks up abstraction levels — instead of fixing one crash, demand "zero crashes across all apps" with crash-report access for autonomous reproduction and verification; turn quarterly reports into a redesigned real-time early-warning system.
- Depth: force expert-grade tools (write code against databases, not Markdown tables); set seemingly unreasonable targets (200ms load times); use raw data to destroy folk assumptions; build user-behavior simulators to find blind spots; always use external executable environments for objective scoring.
- Breadth: convert nouns to verbs (a deck is really decision support under uncertainty), and mobilize multiple roles' actual research methods — analysts, journalists, even short-sellers — not shallow roleplay.
He shares a reusable prompt template: ask the model to list what top experts' raw data, tools, and objective validation standards would be, then rewrite your requirements against them.
Related event: Anthropic Engineer: Aim Absurdly High to Unlock AI Model Limits(2 posts)→
More from Apps
- Consumer agents are a new way to use computers — day 1 of wave 1, argues investor — illscience · 2026-10-23
- Messi asks ChatGPT for a chess opening — it picks the 'Goat Horns' — OpenAI · 2026-10-23
- Have AI Build Your Full Hotel Bill Before Booking — 45% of Marrakesh Rates Hide On-Site Fees — Calm_Cartographer324 · 2026-10-23
- Canva unveils Craft, its first in-house image model that plans before generating — The Verge AI · 2026-10-23
- Decisions API hits GA as agents automate cloud video editing via feedback loops — dkundel · 2026-10-23
- Anthropic's Alex Albert: Claude Duet could matter more to small businesses than Square — alexalbert__ · 2026-10-23