Rich Hickey’s AI-as-Automation Argument Faces Pushback on Hacker News
A Hacker News discussion challenges Rich Hickey’s claim that LLMs are automation rather than tools, arguing that effective use requires skill.
Key points
- Rich Hickey published an article on October 11, 2026, arguing that large language models are automation, not tools, because they cannot be wielded by humans.
- A blog post responding to Hickey’s argument gained significant traction on the Hacker News front page, attracting over 160 points and 147 comments.
- The author of the response contends that LLMs function as tools when users apply expertise to direct inputs, interpret feedback, and correct errors.
Rich Hickey’s October 11, 2026, article asserting that artificial intelligence is automation rather than a tool sparked a substantial debate among technology professionals. The discussion unfolded on the Hacker News front page, where a blog post challenging Hickey’s premise received 163 points and generated 147 comments. The response argues that while outsourcing work carries risks, it does not disqualify large language models from being considered tools.
The core disagreement centers on the definition of a tool and the role of human agency. Hickey posits that large language models cannot be wielded, implying they lack the necessary characteristics of traditional instruments. The blogger counters that wielding a tool requires skill, judgment, and the ability to interpret feedback. Just as a saw or chisel amplifies both good and poor judgment, large language models depend on the operator’s ability to direct them and recognize mistakes.
Context on the Definition of Tools
The blogger argues that unpredictability and complexity do not remove an instrument from the category of tools. Qualities such as predictability, reliability, and bounded function are metrics for evaluating tools, not prerequisites for membership. A tool remains a tool even if it is difficult to control or easy to misuse. The author suggests that Hickey creates a false division between tools and automation, noting that automation can be a feature of a tool without negating its status.
Examples from software development illustrate this point. A compiler performs substantial work without continuous programmer direction, yet developers still wield it by specifying inputs and examining results. Similarly, spreadsheets recalculate data automatically, but users control the process by deciding what to do next. The blogger questions why automatic text generation should disqualify large language models from being tools when automatic code translation or number calculation does not disqualify other systems.
Who Is Affected by the Debate
The argument specifically addresses programmers and professionals who interact with large language models in their daily work. The blogger emphasizes that using these systems involves intention, feedback, and direction. A programmer can ask a model to explain code, challenge that explanation, inspect the source, and test proposed changes. This process creates a feedback loop directed by a human pursuing a specific purpose.
Expertise determines the usefulness of the instrument. Knowing what to ask, what evidence matters, and when an answer is wrong are skills that make the interaction productive. The blogger acknowledges Hickey’s concern that accepting summaries in place of reading can sacrifice learning opportunities. However, the author notes that users can also read source material and use large language models to question interpretations or identify areas for closer examination.
What should you do?
The debate highlights the importance of maintaining critical engagement with artificial intelligence outputs. Professionals should view large language models as instruments that require active direction rather than passive automation. Understanding the limitations of these systems helps users apply them effectively without losing essential skills or intellectual rigor. Monitoring discussions on platforms like Hacker News can provide insights into evolving best practices for tool usage.
Background: Large language models
Large language models are a specific type of neural network trained on text. Generative AI is the broader umbrella term for any system that creates new content, including images and code. Not all generative AI uses large language models, but most modern text generators do.
What to do and how to stay safe: Rich Hickey
- Evaluate large language model outputs critically by verifying information against primary sources rather than accepting summaries at face value.
- Develop specific prompts that encourage the system to explain reasoning, allowing you to inspect the logic and identify potential errors.
- Maintain manual inspection of code or text generated by automation tools to ensure alignment with your intended purpose and quality standards.
- Treat automation features as aids that require human direction, ensuring you retain the ability to override or correct the system’s suggestions.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is Rich Hickey’s main argument about AI in his October 2026 article?
Hickey argues that large language models are automation rather than tools because they cannot be wielded by humans in the traditional sense.
How does the blogger respond to Hickey’s definition of a tool?
The blogger argues that unpredictability and automation do not disqualify an instrument from being a tool, emphasizing that skill and feedback loops are key.
What examples does the blogger use to compare LLMs to other tools?
The blogger cites compilers and spreadsheets, which perform automatic work but are still considered tools because users direct inputs and examine results.
