20-year-old interviews 12 OpenAI, Anthropic and Google researchers for their 10 go-to prompts
JafarNajafov · x · 2026-10-23
A 20-year-old interviewed 12 AI researchers from OpenAI, Anthropic, and Google and distilled the 10 prompts they all actually use to ship frontier products.
Highlights:
- Show Your Work: force step-by-step reasoning before the final answer.
- Adversarial Interrogation: make the model argue against its own answer with the 3 strongest counterarguments.
- Constraint Forcing: hard limits on length, mandatory citations, no hedging.
- Format Lock: valid JSON with fixed keys like analysis and confidencescore.
- Expertise Assignment: a specific persona with years, banned mistakes, and habits.
- Thinking Budget: 500 words of visible thinking including dead ends.
- Comparison Protocol: table comparing options across speed/accuracy/cost/complexity/maintenance.
- Uncertainty Quantification: 0-100 confidence per claim, below 70 flagged speculative.
- Edge Case Hunter: name 5 adversarial inputs that would break the approach.
More from Apps
- Jack Dorsey's open-source AI agent "Company OS" repo hits 14.4K GitHub stars — JarnoDuursma · 2026-10-23
- Chain of Verification: the self-correction prompt frontier researchers rely on — JafarNajafov · 2026-10-23
- A 20-year-old interviewed 12 frontier AI researchers — here are the 10 prompts they all use — JafarNajafov · 2026-10-23
- Canva launches Craft, its first in-house image generation model — The Decoder · 2026-10-23
- Google Play now installs private self-built apps as AI Studio turns prompts into Android apps — aitrendz_xyz · 2026-10-23
- Muse agent handles real online shopping via its VM browser, human just taps pay — FuSheng_0306 · 2026-10-23