Anthropic Engineer's Four Ambitions Prompting Guide Gains Attention

Felix Rieseberg of the Anthropic Claude team published a long piece, "How to Be More Ambitious: A Tutorial for Humans and Models." @petesena called it the best tutorial they've read on "being more ambitious with AI" and broke it down layer by layer across multiple tweets. Rieseberg's core observation: model capability growth is uneven—as Terence Tao put it, "astonishing in some directions, inadequate in others"; some people are already flying, while most pay $200 for a subscription only to ask the model trivial questions like "the capital of France."

Confirmed

• The guide splits ambition into four levels: Speed, Altitude, Depth, and Range; @petesena expanded on the last three
• Altitude: have the model solve the goal behind the goal. Rieseberg gives the example of once asking Claude to fix crashes one by one, versus now simply asking it to keep the app from crashing
• Depth: demand the model work to the standard of a top domain expert—if a website needs to load in 5 seconds, specify 200 milliseconds; have it use the tools experts actually use (databases, statistical models)—pretty tables don't count; never let the model grade itself—bring in an external judge to verify results
• Range: turn task nouns into verbs to widen the model's perspective, e.g., reframing a quarterly review as "get a busy person to believe in a ... within 30 minutes"

Why It Matters

• The guide targets a widespread phenomenon today: a huge gap between model capability ceilings and how users actually use them
• The four-level framework offers actionable prompt-rewriting methods, with concrete examples covering goal hierarchy, professional standards, and task phrasing—ready to apply directly to daily workflows

2026-10-26 ~ 2026-10-26 · 5 related posts

• [source] Anthropic engineer's '4 types of ambition' guide: stop asking a rocket for France's capital — petesena · 2026-10-26
• Ask for the goal behind the goal: Anthropic engineer's 'Altitude' prompting rule — petesena · 2026-10-26
• Prompt like a top expert: set extreme bars, real tools, and outside judges — petesena · 2026-10-26
• [source] Claude team member's tutorial: four types of ambition for using AI models well — petesena · 2026-10-26
• Turn nouns into verbs: the 'Range' trick to unlock cross-domain AI output — petesena · 2026-10-26