| | | | Neo Kim, Kent Beck, and Paweł Huryn posted new notes | |  | |  | | Neo Kim restacked |
| | | | Most software engineers think AI coding makes them faster. They’re wrong… AI coding increases output. But it also increases issues… — According to CodeRabbit report, AI co-authored pull requests created 1.7x more issues than human-only pull requests. Yet AI-generated code often looks correct at a glance. Why? — Because AI optimizes for surface-level correctness, not deep project context. So AI coding is neither “good” nor “bad”. It ‘amplifies’ patterns (including the wrong ones). — Here’s what engineers should focus on with AI coding (according to CodeRabbit report): 1 Logic and correctness ↳ Logic and correctness issues were 75% more common in AI co-authored PRs. ↳ Algorithm and business logic errors appeared 2.25x more often. ↳ Error and exception-handling gaps were 2x higher. ↳ Null-pointer risks, misconfigurations, dependency ordering, and concurrency mistakes all showed large increases. —— 2 Code quality and maintainability ↳ Readability issues were over 3x higher in AI PRs. ↳ Formatting problems appeared 2.66x more often. ↳ Naming… | | Read More %26strokeWidth%3D2) |
| | | | |
|
| |  | | | | I saw reports that telling Claude Code “code like Kent Beck” improves… | | Read More |
| | | | |
|
| |  | | | | Stop telling AI what to do. Tell it how you’ll know it worked and why it matters. One meta prompt instead of memorizing task-specific ones: SYSTEM CONTEXT - We are working on [larger initiative]. - Your output will be used by [who] to decide [what], under [constraints like time, incentives, uncertainty]. OBJECTIVE (Intent) - The outcome I care about is [end state]. - I’m open to how we get there. WHY THIS OBJECTIVE MATTERS - It matters because [reason]. - If this reason is weak or misaligned, say so. SUCCESS SIGNALS (Not metrics, signals) - We would believe this worked if [observable behavior or decision changes] happened. - If success is hard to observe, explain why… | | Read More |
| | | | |
|
|
| | |
|
|
