Agentic development panel
summary
Claims include an 8x faster claim from provider comparisons, concrete tooling like Acolyte with 15 encoded skills (build, plan, code-review, TDD) and recommendations to use LangChain or Vercel AI SDKs. Workflows shift to GitHub issues, regen-able architecture docs and small PRs instead of hand-coding; counterexamples are 4–5k line AI slaps that demand micromanagement. Practical safeguards: distill principles into reusable skills, monitor token cost, and require human verification for critical systems.
AI-written summary · about
chapters (13)
- 0:00Agentic development shifts
- 1:28Creators to verifiers
- 3:36Intent-driven workflows
- 8:33Iterative intent capture
- 11:25Human oversight in planning
- 13:51Avoid coding agents
- 16:39Engineering expertise to skills
- 19:48Monitoring agentic systems
- 20:52AI tool pricing models
- 24:02Trust in critical infrastructure
- 26:31Verify agent outputs
- 28:39Intent beyond requirements
- 30:04Closing remarks
related talks
35mAgentic coding
14mAgent-Driven Software Development
50mPanel Software Engineering in the age of AI
50mPanel: Beyond The Vibe: Code Quality First
14mThe 4 Patterns of AI Native Development
29mBuilding Trust in AI Agents
29mLeveraging AI to Enhance DevEx & Operational Efficiency
01h 00m10 tips to level up your ai-assisted coding