Anthropic's Felix Rieseberg: 4 ways to use AI more ambitiously — speed, altitude, depth, range
FinanceYF5 · x · 2026-10-30
Anthropic engineering lead Felix Rieseberg published a tutorial on why most people use frontier models to polish emails while others use them to find new planets, make music videos, and explore new math — the gap is imagination, not model capability. He breaks ambition into four directions:
- Speed: same work in less time — the default, well-established use.
- Altitude: give AI the goal behind the task, not the task itself — e.g. instead of "make a quarterly deck", have it analyze the quarter, surface the three urgent decisions, then design a system that detects problems earlier.
- Depth: work like a top domain expert — databases and statistical models over Markdown tables, raw data over summaries, external verification over self-grading. He shares a ready-to-use prompt: "I need a deeper version of this task, not a bigger one. State what top experts would do, what raw materials they'd start from, what methods and tools they'd use, and what besides you and me could prove the result correct — then rewrite the task."
- Range: apply cross-domain methods that used to be prohibitively expensive — reframing a quarterly report as "convince busy people in 30 minutes to trust a causal explanation and decide under uncertainty" naturally pulls in analysts, journalists, editors, motion designers, short-sellers, and forensic accountants.
Citing Terence Tao — models are "astounding in some directions, inadequate in others" — his takeaway: as capabilities jump every few weeks, update the scale of your goals, not just your tools.
More from AGI Musings
- Polymarket halts trading after $2.7M drained via compromised third-party script — cyb3rops · 2026-10-30
- Gavin Crooks: Maybe it's time to write some 'last papers' as AI absorbs knowledge — MaxUnfried · 2026-10-30
- Anthropic hiring manager kills coding interviews: 'hunger beats tenure' — bibryam · 2026-10-30
- User documents Fable 5.1 spontaneously developing safeguards against over-steerability — repligate · 2026-10-30
- Pedro Domingos: COBOL will outlive Claude Code — pmddomingos · 2026-10-30
- Researchers used LLMs to uncover 'substantial' vulnerabilities in the German legal system — cyb3rops · 2026-10-30