AI Assistance Beyond Code: What Do We Need to Make it Work?
summary
Shows working demos: shared prompt files in a repo hooked into the Continue coding assistant, running LLaMA 3 locally for experiments, and a public sandbox on GitHub. Demonstrates breaking an epic into stories, auto‑questioning for grooming, and STRIDE threat models generated from injected domain text and wiki pages via retrieval augmented generation. Connects code, Confluence/Jira and issue trackers as context but warns about scope creep and the “lost in the middle” problem with huge context windows; curation and workflow design remain necessary.
AI-written summary · about
chapters (11)
- 0:00Intro
- 1:04AI coding assistance is here to stay
- 2:31Why should we care?
- 4:48What do we need to take advantage of this?
- 7:56How can AI help?
- 8:291. Codify practices in shared prompts
- 14:392. Consider different interaction types
- 22:503. Make more context available
- 25:24But can't we do more?
- 32:06Curation still matters
- 38:46Outro
related talks
38mAI for Software Development: A Reality Check
58mAI Assistant: Don’t Code Alone
45mAI Assistance for Software Teams: The State of Play
49mAI Assistants for C++ Developers
01h 01m10 tips to level up your ai-assisted coding
49mFrom autocomplete to agents: AI coding assistance state of play
29mIs coding assistant as good as we thought
15mVibe Coding at Scale: Customizing AI Assistants for Enterprise Environments