← back
AI Engineering Without Borders — swyx
Takeaway
Stop treating AI engineering tracks as silos — look for invariants and trends (context, cost, latency) that hold regardless of subspecialty.
Summary
- swyx argues AI engineer 'tracks' (RAG, agents, coden, multimodal) are arbitrary boundaries; AI itself disrespects them — natively multilingual, multimodal, ignores copyright.
- Proposes searching for 'laws' of AI engineering analogous to laws of physics: constants (humans speak 80 wpm read 200 wpm) and contingent facts (30 tok/sec Apple Intelligence as new speed-too-cheap-to-meter baseline).
- Documents trends: improving context windows (60K -> million-token, plus better utilization), and ~99.5% decline in GPT-3-level cost and ~80-90% for GPT-4-level via Llama 3.
- Recaps last-summit's three AI engineer archetypes — AI-enhanced engineer, AI-products engineer, and non-human agentic AI engineer — but says all are blurring.
- Frames Latent Space podcast and AI Engineer conference as deliberate community-building, not just content.
ai-engineeringlawstrends
Original description
One year into the Rise of the AI Engineer, what do we think we know, and are we so sure? Recorded live in San Francisco at the AI Engineer World's Fair. See the full schedule of talks at https://www.ai.engineer/worldsfair/2024/schedule & join us at the AI Engineer World's Fair in 2025! Get your tickets today at https://ai.engineer/2025