Rethinking Software and Software Engineering in the Era of Foundation Models...
summary
Claims ten concrete pain points labeled 10 SE4FMware challenges, cites operational costs where a regression test suite can add $3–$5 per CI checkin, and shows that demos reach ~80% quickly while each remaining percent grows harder; demonstrates extreme prompt sensitivity with a punctuation tweak dropping accuracy prompt fragility (0.8 → 0.036). Describes tooling gaps (LangChain emitting Python that blocks low‑level optimization), versioning/licensing issues for data and agents, and an implemented platform called FM Arts GitHub 2.0 with studios for prompts, agents, orchestration, knowledge and operations.
AI-written summary · about
chapters (12)
- 0:00Foundation model apps intro
- 2:28Challenges in FM apps
- 3:11Codeware to AIware
- 7:03FMware lifecycle
- 10:53Scaling FMware challenges
- 13:45Data shift challenges
- 15:30Prompt engineering complexity
- 18:22Industry collaboration
- 20:06Performance engineering and operational constraints
- 23:41FMware platforms
- 25:29FMware standardization
- 27:11Future directions
related talks
01h 00mFrom Copilots to Agentic SE Teammates: Engineering the SE 3.0 Era with AI-Native...
24mSubstrate engineering
31mBuilding High-Quality & Trustworthy FMware
26mMastering Large Language Models: Prompt Engineering for Precision & Performance
27mThe Many Ends of Programming
43mAdvanced Prompt Engineering and Beyond
39mFrom Demo to Production: Why Agentic AI Systems Fail - and How to Fix Them
19mOn Engineering AI Systems that Endure The Bitter Lesson