The confusion comes from conflating "repetitive in structure" with "repetitive in specifics."
Traditional automation handles tasks that are repetitive in specifics - the exact same input produces the exact same output every time. Press this button, get that result.
What AI handles is tasks that are repetitive in structure but variable in specifics. "Review this pull request" is structurally the same every time (read code, check for issues, suggest improvements), but the actual code and context varies. You can't write a script for it because the inputs are too variable, but it's not intellectually novel work either.
Same with writing boilerplate. The pattern is repetitive (create CRUD endpoint, add validation, wire up to database), but the specifics change each time (different fields, different validation rules, different table schemas). Traditional code generation works when you can parameterize everything upfront. AI works when the parameters are implicit in context you'd have to explain to another developer anyway.
The real threshold isn't repetitive vs. novel. It's "can the task be fully specified with formal rules" vs. "does it require judgment calls based on fuzzy context." AI is good at the latter in ways traditional automation isn't.
You _could_ write non-AI tools for those examples. That's basically what sonarqube, coverity, lint, etc... are. They might even do a better job, with better correctness and deterministic results.
There are also plenty of code generation / templating tools available. Maybe there are other tasks where AI offers novel capabilities, like interacting with humans, but at least for those two examples I think a lot of people are really just distracted by the novelty of AI, and perceiving the results of non-deterministic / black box system as "magic" or "creative".
You're right that static analysis tools handle well-defined rule violations - null pointer dereferences, buffer overflows, style violations. Those are formally specifiable.
But "review this pull request" isn't just about catching bugs the linter would find. It includes things like "does this architectural choice make sense given where the codebase is heading" or "is this abstraction going to cause problems when we add feature Y next quarter." Those require contextual judgment that's hard to encode in rules.
Same with boilerplate - yes, templating tools exist, but they require you to anticipate the parameterization ahead of time. AI handles "I need a thing like X but slightly different in ways that are obvious if you look at the existing code." That's a fuzzy spec problem, not a rule-based generation problem.
Whether that's "magic" or just pattern matching on training data is a fair philosophical question. But pragmatically, it solves a category of tasks that neither linters nor template generators handle well.
I am generally skeptical about AI but I do see the benefit here.
I write a bunch of widgets for my website. They're little calculators that use common components and apply simple logic. Think unit conversion or date arithmetic.
These currently take a few hours to write, and most of the work is just wiring things together in a predictable way: template, tests, common form controls.
I think that this would be a very good case for AI.
I've automated the writing of many of these kinds of things away with LLMs over the last year. I'd recommend giving it a shot.
Another tangential use is to ask it to see if there are better abstractions that can be applied that maybe you haven't thought of yet. I find it's not the majority case that it offers a suggestion I like, but I have found it to be good some of the time. Worth the time of asking, at least.
This is cool! I set some time aside to experiment with it in February. If I can keep the human aspect while removing the tedium, I would be very happy.
Perhaps your supposedly unique work is more repetitive than you thought: it just has a decision tree that's difficult to model with a regular algorithm, and annoyingly, it turns out you can just brute force that decision tree if you have enough electricity.
Unless your job is cutting-edge research where you are truly making new scientific discoveries and methods, you're just combining other peoples' ideas into a new unique package and selling it.
The truly valuable work is to notice that there is an underserved market and figure out how to meet their needs.
Seeing a lot of claims about using AI to write "boilerplate" and other repetitive bits of code, I was somewhat surprised, as I have historically written my own code generation tools to spit out repetitive, formulaic code. I didn't need AI; I just needed to understand what I wanted and write a script for it.
I suppose that generative AI was seen as such a boon to writing boilerplate because it could do so without you having to specifically program anything; it was trained on enough sufficiently-close examples that it could pull it off without a thorough description.
The confusion comes from conflating "repetitive in structure" with "repetitive in specifics."
Traditional automation handles tasks that are repetitive in specifics - the exact same input produces the exact same output every time. Press this button, get that result.
What AI handles is tasks that are repetitive in structure but variable in specifics. "Review this pull request" is structurally the same every time (read code, check for issues, suggest improvements), but the actual code and context varies. You can't write a script for it because the inputs are too variable, but it's not intellectually novel work either.
Same with writing boilerplate. The pattern is repetitive (create CRUD endpoint, add validation, wire up to database), but the specifics change each time (different fields, different validation rules, different table schemas). Traditional code generation works when you can parameterize everything upfront. AI works when the parameters are implicit in context you'd have to explain to another developer anyway.
The real threshold isn't repetitive vs. novel. It's "can the task be fully specified with formal rules" vs. "does it require judgment calls based on fuzzy context." AI is good at the latter in ways traditional automation isn't.
You _could_ write non-AI tools for those examples. That's basically what sonarqube, coverity, lint, etc... are. They might even do a better job, with better correctness and deterministic results.
There are also plenty of code generation / templating tools available. Maybe there are other tasks where AI offers novel capabilities, like interacting with humans, but at least for those two examples I think a lot of people are really just distracted by the novelty of AI, and perceiving the results of non-deterministic / black box system as "magic" or "creative".
You're right that static analysis tools handle well-defined rule violations - null pointer dereferences, buffer overflows, style violations. Those are formally specifiable.
But "review this pull request" isn't just about catching bugs the linter would find. It includes things like "does this architectural choice make sense given where the codebase is heading" or "is this abstraction going to cause problems when we add feature Y next quarter." Those require contextual judgment that's hard to encode in rules.
Same with boilerplate - yes, templating tools exist, but they require you to anticipate the parameterization ahead of time. AI handles "I need a thing like X but slightly different in ways that are obvious if you look at the existing code." That's a fuzzy spec problem, not a rule-based generation problem.
Whether that's "magic" or just pattern matching on training data is a fair philosophical question. But pragmatically, it solves a category of tasks that neither linters nor template generators handle well.
I am generally skeptical about AI but I do see the benefit here.
I write a bunch of widgets for my website. They're little calculators that use common components and apply simple logic. Think unit conversion or date arithmetic.
These currently take a few hours to write, and most of the work is just wiring things together in a predictable way: template, tests, common form controls.
I think that this would be a very good case for AI.
I've automated the writing of many of these kinds of things away with LLMs over the last year. I'd recommend giving it a shot.
Another tangential use is to ask it to see if there are better abstractions that can be applied that maybe you haven't thought of yet. I find it's not the majority case that it offers a suggestion I like, but I have found it to be good some of the time. Worth the time of asking, at least.
This is cool! I set some time aside to experiment with it in February. If I can keep the human aspect while removing the tedium, I would be very happy.
Perhaps your supposedly unique work is more repetitive than you thought: it just has a decision tree that's difficult to model with a regular algorithm, and annoyingly, it turns out you can just brute force that decision tree if you have enough electricity.
Unless your job is cutting-edge research where you are truly making new scientific discoveries and methods, you're just combining other peoples' ideas into a new unique package and selling it.
The truly valuable work is to notice that there is an underserved market and figure out how to meet their needs.
Seeing a lot of claims about using AI to write "boilerplate" and other repetitive bits of code, I was somewhat surprised, as I have historically written my own code generation tools to spit out repetitive, formulaic code. I didn't need AI; I just needed to understand what I wanted and write a script for it.
I suppose that generative AI was seen as such a boon to writing boilerplate because it could do so without you having to specifically program anything; it was trained on enough sufficiently-close examples that it could pull it off without a thorough description.
Use the AI to make the “machine” that does the repetitive work?
I’m not a programmer but that’s what I’ve done. In the past I would’ve needed either to learn how to code or hire someone.
AI automates deterministic tasks, in a non-deterministic way. Repetitiveness just gives you a pipeline.