We scan new podcasts and send you the top 5 insights daily.
Erik Bernhardsson notes that teams struggle to deprecate APIs because coding agents continually generate outdated syntax absorbed during model training. As a result, infrastructure teams must make conservative changes and actively rewrite online docs, benchmarks, and public code samples to steer agent recommendations.
When asked to modify or rewrite functionality, LLMs often attempt to preserve compatibility with previous versions, even on greenfield projects. This defensive behavior can lead to overly complex code and technical debt. Developers must explicitly state that backward compatibility is not a requirement.
Unlike normal technical debt, 'agentic technical debt' compounds rapidly. Without persistent, written architectural constraints, AI coding tools re-derive foundational decisions in each session, causing the codebase to drift incoherently. The solution is to document architectural principles before building to give the AI context and prevent entropy.
Features built to guide AI agents, like an explicit "plan mode," will become obsolete as models become more capable. The Claude Code team embraces this, building what's needed for the best current experience and fully expecting to delete that code when a new model renders it unnecessary.
Agents quickly become outdated. To manage this lifecycle, build specific 'upgrade skills' that facilitate migration to new models. For larger-scale management, deploy 'meta-agents' whose sole job is to monitor other agents, identify outdated ones, and trigger the upgrade process.
Documentation is shifting from a passive reference for humans to an active, queryable context for AI agents. Well-structured docs on internal APIs and class hierarchies become crucial for agent performance, reducing inefficient and slow context window stuffing for faster code generation.
The underlying infrastructure for AI agents ('harnesses') becomes obsolete roughly every six months due to rapid advances in AI models. At Notion, this means completely rewriting the harness multiple times a year, demanding a culture comfortable with constantly rebuilding core systems and discarding previous assumptions.
Documentation is no longer just for humans. AI agents now read it directly as operational input, making its accuracy critical for system function. Outdated docs, once a nuisance, now cause system failures, elevating documentation to the level of essential infrastructure.
After achieving broad adoption of agentic coding, the new challenge becomes managing the downsides. Increased code generation leads to lower quality, rushed reviews, and a knowledge gap as team members struggle to keep up with the rapidly changing codebase.
AI models have knowledge cutoffs and may recommend outdated libraries (e.g., Next.js v14 instead of v17). Proactively prompt the AI to check the latest API documentation and confirm it's using the most up-to-date versions to prevent significant, time-consuming refactoring later.
Historically, developer tools adapted to a company's codebase. The productivity gains from AI agents are so significant that the dynamic has flipped: for the first time, companies are proactively changing their code, logging, and tooling to be more 'agent-friendly,' rather than the other way around.