Buy Now, Maybe Pay Later: Dealing with Prompt-Tax While Staying at the Frontier
summary
Agentic product for real‑estate lawyers ships from GPT‑3.5 through GPT‑4 32K to 4 Turbo 40 and 4.1 and mixes system‑one and system‑two models (01 preview, 04 mini) while OCRing 100‑page files and producing clickable citations and Word reports. Claims: 18 months of commercial ops, 20 billion tokens/month, 1,000+ prompts, multiple seven‑figure ARR. Tactics that actually work: domain experts author prompts, progressive rollout and feature flags, migrate prompts with new models, accept “vibes” then harden with fast feedback loops.
AI-written summary · about
chapters (17)
- 0:00Prompt tax introduction
- 1:47Agentic products at frontier
- 3:12Agentic real estate demo
- 5:44Token and revenue growth
- 7:10Prompting over fine-tuning
- 8:12Vibes vs rigorous evaluation
- 10:19Prompt tax not technical debt
- 12:08Model upgrade tactics
- 14:13Overcoming AI change aversion
- 15:39Future model capabilities
- 16:43Risks on new models
- 17:25Fast feedback loops
- 19:13AI engineers as product engineers
- 20:37Vibes scaling limits
- 22:22Edge cases in agentic systems
- 22:45Progressive delivery for prompt tax
- 23:48Buy now stay on frontier
related talks
01h 08mIn Prompts We Trust
39mPragmatic AI: Reshaping the Product Opportunity Tree
01h 04mIn Prompts We Trust
38mCreate the next generation of user experiences with text generation
01h 00mGoing beyond ChatGPT: introduction to prompt engineering & LLMs
20mThe Prompt Is Still a Punch Card
30mSystem Prompt Design for Voice Agents
47mThe Art of Prompt Engineering: Best Practices for Effective AI Interactions