AI is a tool that requires skill, despite what Rich Hickey says
LLMs are tools that amplify your judgment. Predictability is a metric for evaluating them, not a strict requirement for using them.
Tools · Source: Hacker News
What happened
Clojure creator Rich Hickey recently argued that AI is mere automation, not a tool. He claimed large language models cannot be wielded by humans because they lack intention and direction. Developer Huahai published a sharp rebuttal on his personal blog. He argues that LLMs are absolutely tools that demand serious operator skill. Outsourcing work to them might cost us learning opportunities, but that does not change their fundamental nature.
Huahai points out that tools simply amplify the existing judgment of the user. Picking up a saw does not guarantee a straight cut. Holding a chisel does not instantly make you a master sculptor. The usefulness of a model depends entirely on your ability to direct it. You must interpret its feedback and recognize its mistakes. A constant feedback loop of asking, inspecting, and refining is exactly how a human wields a traditional tool. Expertise makes the instrument far more useful.
The rebuttal also attacks the anthropomorphic language heavily used by the AI industry. Words like conspiring, escaping containment, or going rogue give software fake human motives. Unpredictable outputs or failures to follow instructions are just standard engineering problems. Software can produce unexpected outputs or cause harm when given access to consequential actions. A tool does not stop being a tool just because it is difficult to control or easy to misuse.
Key facts
- October 11, 2026 — Date Rich Hickey published his article claiming AI is not a tool
- yyhh.org — Blog where Huahai published his rebuttal
Why it matters
Treating AI as an independent agent makes you a lazy engineer. If you view a large language model as a tool, you take full responsibility for the final output. You learn to write better prompts. You rigorously verify generated code. You tighten your feedback loops. Builders who master this new instrument will ship significantly faster than those who blindly delegate work. The possibility of avoiding intellectual work does not mean every single use case avoids it.
This debate exposes a deep divide in how veteran programmers view modern probabilistic software. Rejecting AI because it lacks perfect predictability ignores how we already use complex systems. Compilers and spreadsheets perform substantial automated work without continuous manual direction. As AI gets embedded into every workflow, the gap between skilled operators and careless delegators will widen. Companies will stop paying for raw output. They will start paying a premium for the human judgment required to guide these models.
For builders
Stop using anthropomorphic language for software
Do not say your model went rogue or conspired against you. These are basic engineering failures, not complex personality quirks. Founders who blame the AI for bad outputs will quickly lose trust with enterprise clients who demand strict accountability. You pay the price when you treat code like a human.
Treat prompts as tight feedback loops
You cannot just dump a massive task on an LLM and walk away. You must ask questions, inspect the source, test the proposed change, and refine the request. Engineers who build internal tooling to shorten this exact feedback loop will win the market.
Do not confuse automation with autonomy
Compilers and spreadsheets automate tedious work without stripping ultimate control from the user. Build AI products that keep the human firmly in the driver seat. Users will happily pay for SaaS tools that amplify their existing expertise instead of attempting to replace their judgment.
My take
I am tired of veteran programmers moving the goalposts just because they do not like probabilistic software. A hammer is a tool, and if you smash your thumb, you do not blame the hammer for going rogue. Treat AI like a chisel, sharpen your skills, and stop whining about predictability.
Original reporting: Hacker News. This is my rewrite and opinion.