Don’t Use AI to Write Things You Present as Your Own Work

James Bach, a veteran testing expert, published a strong opinion on his blog Satisfice: never use AI to write anything that you present as your own work. He argues that even saying “AI helped me” immediately undermines your credibility, because everyone knows AI does far more than just “help.”
The Core Argument
Bach’s policy is absolute: “I never let AI draft anything for me that has my name on it. Not one sentence. Nothing. Ever.” He compares using AI for writing to “hooking a motor to a stationary bike and calling that exercise” or “taking a helicopter to the top of Everest and saying that I summited.”
Trust and Reputation
Bach warns that once you admit to using AI, readers can’t know how much work you did. He says: “All I have is your word for it that you did any work at all.” This forces a “big fat asterisk next to anything you do.” He adds, “Only someone who uses AI to help their writing would say ‘I use AI to help my writing.’ And anyone who uses AI to help their writing may use it to do all of their writing. It’s such a steep and slippery slope to full laziness.”
Personal Anecdote
Bach describes a past ghostwriting experience where he wrote an article for someone else, choosing what to include and exclude. The article was published under the other person’s name, but it was substantially Bach’s work. He regrets it because it helped “burnish the professional image of a man who didn’t earn it.” He contrasts this with a deep collaborative chapter co-written with Jeff Nadelman, where both contributed substantially—Jeff through many hours of Zoom calls and personal stories. Bach says: “If AI deeply collaborates with you to write something, why am I saying you shouldn’t say you used AI? Because all I have is your word for it that you did any work at all.”
The Slippery Slope
Bach believes the line between “help” and “doing all the work” is too blurry. He advises: “Don’t tell me that AI helped you write, and don’t tell me that you did all the work when you only did some of it. Therefore, don’t use AI to write things for you that you present as your own work. If you DO, I will treat all your work as if it is slop or spam. And not just me, the entire community of serious professionals will discount ‘your’ work.”
Takeaway for Developers Using AI Coding Agents
While the article focuses on writing, the principle applies to code as well. If you use an AI to generate substantial portions of code or documentation that you claim as your own, you risk the same trust erosion. Be transparent about your tools and always review and refine the output. But Bach’s core advice is clear: don’t outsource creative or intellectual work and then take full credit.
📖 Read the full source: HN LLM Tools
👀 See Also

Claude Code v2.1.147: Pinned Sessions, /code-review, and Dozens of Fixes
Claude Code v2.1.147 introduces pinned background sessions, renames /simplify to /code-review with effort levels and --comment, plus fixes for PowerShell, MCP, Windows, and more.

Analysis of 'Clausage': User Anxiety Patterns in AI Subscription Models
A user analysis identifies 'Clausage' or 'The Claude Syndrome'—behavioral patterns where premium AI subscribers experience chronic usage anxiety, avoidance behavior, and compulsive resource monitoring. The source details specific symptoms like anticipatory avoidance, usage hypervigilance, and paradoxical underutilization of paid services.

Choosing the Best Token Provider for Your API Needs
Explore the key factors to consider when selecting a provider for tokens and APIs in AI coding and automation, based on insights from the OpenClaw community.

Fine-tuning Phi-4-mini by training only LayerNorm parameters fails to improve performance
A hobbyist tested training only LayerNorm γ values on Phi-4-mini across Python and medical domains with different learning rates and data formats. Performance degraded slightly on all benchmarks compared to baseline, with the author concluding transformers already route information dynamically through attention.