Ask HN: How is AI-assisted coding going for you professionally?

TL;DR Highlight

Honest takes from working developers on Hacker News about how they actually use AI coding tools day-to-day — what works, what doesn't, and what's overhyped.

Who Should Read

Developers evaluating AI coding tools, engineering managers deciding on tool adoption, and anyone who wants unfiltered practitioner perspectives beyond vendor marketing.

Core Mechanics

The thread collected candid developer experiences with AI coding tools — a useful ground-truth counterpoint to benchmark scores and marketing claims.
Common positive patterns: AI tools excel at boilerplate, documentation, test generation, and syntax lookup — tasks with well-defined patterns and low stakes.
Common frustrations: AI-generated code for novel or complex problems often requires significant rework; the time spent reviewing/fixing can approach the time it would have taken to write it fresh.
Senior developers tend to get more value from AI tools than juniors — they can quickly spot wrong suggestions and have the context to guide the AI effectively.
The 'AI will replace developers' narrative was broadly rejected — but 'AI changes what skills matter' was widely endorsed.
Many developers noted context management as a key skill: knowing what to put in the prompt and when to start a fresh context is as important as the AI's raw capability.

Evidence

The HN thread collected hundreds of developer responses across experience levels, company sizes, and technology stacks.
Recurring pattern: developers working on legacy codebases (old languages, undocumented systems) found AI less useful than those on modern stacks with good documentation.
Several developers noted that AI tools made them more productive at the beginning of projects (greenfield) but less so for maintenance and debugging of existing systems.
A few developers mentioned abandoning AI tools after finding the review overhead exceeded the generation benefit — suggesting the productivity gain isn't universal.

How to Apply

Match AI tool usage to task type: use AI heavily for boilerplate, tests, and docs; use it lightly and verify carefully for core business logic and novel algorithms.
If you find yourself spending more time reviewing AI output than the code would have taken to write, recalibrate — AI tools have an optimal complexity range.
Invest in prompt engineering skills as a first-class engineering capability — writing clear, context-rich prompts is a learnable skill that multiplies AI tool value.
For tech leads: don't measure AI tool success purely by lines of code output or velocity — measure whether understanding and system quality is improving alongside.

Terminology

Greenfield projectA new software project built from scratch, without existing codebase constraints — generally easier for AI tools than legacy maintenance.

Context managementThe practice of deciding what information to include in an LLM prompt to maximize output quality — a key skill for effective AI tool use.

BoilerplateRepetitive standard code (CRUD operations, config files, test scaffolding) that follows well-known patterns — where AI tools deliver the most consistent value.