This is emphatically not fundamental to LLMs! Yes, the next token is selected randomly; but "randomly" could mean "chosen using an RNG with a fixed seed." Indeed, many APIs used to support a "temperature" parameter that, when set to 0, would result in fully deterministic output. These parameters were slowly removed or made non-functional, though, and the reason has never been entirely clear to me. My current guess is that it is some combination of A) 99% of users don't care, B) perfect determinism would require not just a seeded RNG, but also fixing a bunch of data races that are currently benign, and C) deterministic output might be exploitable in undesirable ways, or lead to bad PR somehow.
This. Thanks for saying that, because now I don't need to read the article, since if the author doesn't even get that, I'm not interested in the rest.
I'm also reminded of the old software called Formulize, which could take in a set of arbitrary data and find a function that described it. http://nutonian.wikidot.com/
But it raises an interesting question about where the fitness function comes from. In traditional GAs you define it explicitly. With LLM-generated code, the fitness function is often just "does it pass the tests" - which means the quality of your tests becomes the actual bottleneck, not the quality of the code generation.
I wonder if that shifts the core skill of programming from "write correct code" to "write correct specifications." And if so, is that actually a new problem, or is it the same problem formal methods people have been working on for decades, just wearing a different hat?
So then, as you say, being able to specify exactly what you want becomes the central skill of programming - I mean, describe the behavior not in terms of the final code, which is an implementation detail, but how it interacts with a given environment. That was always the case since in higher-level languages, including C, what we write is not the final code, which is technically the compiled result.
A difference I notice is that, now, even junior devs are expected to be the "mentor" to language models - teaching and guiding them to generate well-written code with plenty of tests, asserts, and other guardrails. In another comment someone said, breaking down a large program into smaller modules is useful - which is common sense, but we now have to guide an LLM to know and apply best practices, design patterns, useful tricks to improve code organization or performance, etc.
That means, it would be valuable to codify best practices, as documentation in Markdown as well as described in code, as specs and tests. Programming is becoming meta-programming. We're shifting emphasis from assembling genetic code manually to preparing the environment for such code to evolve.
I'm glad to see others talking about it. One day we'll look back on this era the same way folks look back at the time before we validated inputs.
https://www.stevenathompson.com/effective-vibe-coding-best-p...
I'm finding code falls into two categories. Code that produces known results and code that produces results that are not known. For example, creating a table with a pagination component with a backend that loads the first 30 rows ordered by date descending from the database on page 1 and the second set of 30 rows on page 2. We know what the code is supposed to output, we know what the result looks like. On the other hand, there is code that does statistical analysis on the 30 rows of data. This is different because we don't know what the result is.
The known result code is easy to use an LLM with. I have a skill that will iterate with an OODA loop — observe, act, and validate. It will in the validate step take screenshots and even without telling it, it will query the database from the CLI, compare the rendered row data to the database data. It will more surprisingly make sure that all the components are responsive and render beautifully on mobile. I'm orders of magnitude past linting here which is solved with Biome.
The statistical analysis is different. The only way I can know for sure of the result is by writing the code painstakingly by hand. The LLM will always produce specious lies. It will fabricate and show me what I want to see, not the truth. This is because until it is written manually by hand, there is no ground truth. In this case, there is no code checking code.
So can't you just save the conversation transcript and replay it with the tools? Seems a lot more efficient that regenerating the whole thing. And, also, no risk of branching when a tool reply is slightly different. (Of course, errors can occur on subsequent runs.)
Obviously this won't work if your tools are not deterministic, but reproducible builds is a well-trodden discipline.
LLMs really cause diminished reasoning, or in terms that LLM people might understand: Your minds have been quantized!
As the state travels across the graph, I keep a trace of the steps which were executed, which means that when an error happens, the agent has a lot more information than it normally would, it can see what decision points the code passed through already, it can cross references that with the declared workflow, and quickly find where it screwed up.
The idea of workflow engines has been around for a long time, but they feel too awkward to use when you're writing code by hand. Writing conditional logic directly in the code keeps you in your flow, and having to jump out and declare it in config somewhere feels awkward. Coding agents completely change the dynamic though because they don't have that problem. If the LLM is writing the code, then I can just focus on ensuring the code meets the contract, while the agent can deal with the implementation details.
it goes on for ages just to reach the point of "write the tests first"