HNHacker News
TopNewBestAskShowJobs

zie1ony

205 karma · joined July 10, 2015

zie1ony.at.hn
submissionscomments
zie1ony··on We charge $10k a week to delete AI-generated code
Initial AI scan can list most of system usecases. Then we review it with a client and agree on what must stay. We can propse better solutions at this stage. If we remove the code, then we can also remove its unit tests. For intergation, repo-wide tests, we treat it as documentation/requirements that must work on the other side.

Human to human interaction is the tool we use the most :)

zie1ony··on We charge $10k a week to delete AI-generated code
There is also a flip side of this. I work a lot with AI-generate text and I can also catch it quickly, but most of the time now, it's easier to read AI, than humans with their weird chain of thoughts and compex sentences ;P
zie1ony··on We charge $10k a week to delete AI-generated code
No shame in using Fable to help with the grammar and style of text. English is not my native language and while I can use it quite well, LLMs are much better in forming my thoughs into something simple, that sounds great.

My writing workflow is: (1) write what I want to have on the page. (2) /grill-me it should be sound and logical and easy to read. (3) Manually review the text, replace by hand what I don't like.

zie1ony··on We charge $10k a week to delete AI-generated code
Software engineer profession soon will be like a plumber. You don't build, you just fix things.
zie1ony··on We charge $10k a week to delete AI-generated code
I bet on the loc growing exponentially. I wrote this at the end of 2024: https://codegen.substack.com/p/future-of-software-is-growing...

Hope you like it.

zie1ony··on We charge $10k a week to delete AI-generated code
We don't target software developers. Most of the time, they know their craft. We target those who don't have experience and diging into how serwer should be implemented, or best composition of a domain model is not what they would like to sepend hours doing.

The typical client is a 40 years old ex-CTO, that had idea, verified it with Claude Code and got to the point, where the repo is too big. Maybe even rewrote it all from zero, once or twice, but end-up with the same outcome after adding more features.

He can either hire a developer to work on the codebase and refactor it or hire 3party company.

Sometimes the requirement is a security review. And you pay much less if you downsize the repo before handig it over.

zie1ony··on We charge $10k a week to delete AI-generated code
Our outcome is also AI-generated (or AI-assisted) code. It is not possible to rewrite 100k loc in a week. Where our stills are is refactoring at scale with help of agents. Basically, software engineer knows better what prompt to put, and when to redo the process or go different path.
zie1ony··on We charge $10k a week to delete AI-generated code
Creator here,

There is a new kind of task for software engineers these days. A client calls, asks for a "small refactor," and sends you 100k lines of AI-generated spaghetti.

And this is great! This is something we can work with.

Any experienced engineer can look at a codebase like that and quickly see what to refactor, where a library replaces a few thousand hand-rolled lines, and what smells bad. Removing the first 30% is easy. The next 30% is harder, and that is exactly what the price should be on: doing what others can't. We use coding agents too, of course, but as a tool, not as the driving force.

That is why we started Slopfix, a software house focused entirely on refactoring AI-generated codebases. We commit to a reduction target up front, and the client pays in proportion to how much of it we hit. We get paid to delete code.

I am sharing this because cleaning up after agents with 1M token context is a real business for engineers. Curious what HN thinks.

zie1ony··on Slop is not necessarily the future
Bytecode generated by compilers in many cases is far from optimal, but we don't care anymore.
zie1ony··on Founder of GitLab battles cancer by founding companies
Thank you. I've filled the form. I'll be happy to explore this topic together.
zie1ony··on Founder of GitLab battles cancer by founding companies
Have you just "ordered the cancer vaccine" or this is not yet that easy?
zie1ony··on Jeffgeerling.com has been migrated to Hugo
How do you handle comments in a "static blog"?
zie1ony··on The Cloudflare outage might be a good thing
My friend wasn't able to do RTG during the outage. They had to use ultrasound machine on his broken arm to see inside.
zie1ony··on Poker Tournament for LLMs
They are not, and that's the whole point of doing this research. If we can build good benchmark, models developers would have nice goal.
zie1ony··on Poker Tournament for LLMs
Hi there, I'm also working on LLMs in Texas Hold'em :)

First of all, congrats on your work. Picking a form of presenting LLMs, that playes poker is a hard task, and I like your approach in presenting the Action Log.

I can share some interesting insights from my experiments:

- Findin strategies is more interesting than comparing different models. Strategies can get pretty long and specific. For example, if part of the strategy is: "bluff on the river if you have a weak hand but the opponent has been playing tight all game", most models, given this strategy, would execute it with the same outcome. Models could be compared only using some open-ended strategy like "play aggressively" or "play tight", or even "win the tournament".

- I implemented a tournament game, where players drop out when they run out of chips. This creates a more dynamic environment, where players have to win a tournament, not just a hand. That requires adding the whole table history to the prompt, and it might get quite long, so context management might be a challenge.

- I tested playing LLM against a randomly playing bot (1vs1). `grok-4` was able to come up with the winning strategy against a random bot on the first try (I asked: "You play against a random bot. What is your strategy?"). `gpt-5-high` struggled.

- Public chat between LLMs over the poker table is fun to watch, but it is hard to create a strategy that makes an LLM successfully convince other LLMs to fold. Given their chain of thought, they are more focused on actions rather than what others say. Yet, more experiments are needed. For waker models (looking at you `gpt-5-nano`) it is hard to convince them not to review their hand.

- Playing random hands is expensive. You would have to play thousands of hands to get some statistical significance measures. It's better to put LLMs in predefined situations (like AliceAI has a weak hand, BobAI has a strong hand) and see how they behave.

- 1-on-1 is easier to analyze and work with than multiplayer.

- There is an interesting choice to make when building the context for an LLM: should the previous chains of thought be included in the prompt? I found that including them actually makes LLMs "stick" to the first strategy they came up with, and they are less likely to adapt to the changing situation on the table. On the other hand, not including them makes LLMs "rethink" their strategy every time and is more error-prone. I'm working on an AlphaEvolve-like approach now.

- This will be super interesting to fine-tune an LLM model using an AlphaZero-like approach, where the model plays against itself and improves over time. But this is a complex task.

zie1ony··on Andrej Karpathy's talk on the future of the industry
This is great idea, until you have to build something.
zie1ony··on The unreasonable effectiveness of fuzzing for porting programs
I find it amazing, that the same ideas pop up in the same period of time. For example, I work on tests generation and I went the same path. I tried to find bugs by prompting "Find bugs in this code and implement tests to show it.", but this didn't get me far. Then I switched to property (invariant) testing, like you, but in my case I ask AI: "Based on the whole codebase, make the property tests." and then I fuzz some random actions on the state-full objects and run prop tests over and over again.

At first I also wanted to automate everything, but over time I realized that best is: 10% human to 90% AI of work.

Another idea I'm exploring is AI + Mutation Tests (https://en.wikipedia.org/wiki/Mutation_testing). It should help AI with generation of full coverage.

zie1ony··on The Grug Brained Developer (2022)
You must be a Rust developer.
zie1ony··on Vine: A programming language based on Interaction Nets
Implications looks weird https://vine.dev/docs/features/conditions#implication:

    let y = None;
    y is Some(value) => value > 0 // true
    y is Some(value) => value == 0 // true
zie1ony··on 21st Century C++
Seeing badly formatted code snippets without color highlighting in article called "21st Century C++" somehow resonates with my opinion on how hard to write and to ready C++ still is after working with other laguages.
zie1ony··on PR process killing morale and productivity
Good point, but my team is small and has only top engineers with both soft and hard skills. Many times you can do much more with team of 3 than team of 10 ppl.
zie1ony··on PR process killing morale and productivity
Recently I allowed everyone in my team to push to develop without PRs and even without doing feature branches. We review the code all together (max 2 ppl) just before the release and that's it. But great CI pipeline is must for this. It will get even better soon when AI will be able to slightly refactor the code.
zie1ony··on Comparing our Rust-based indexing and querying pipeline to Langchain
DSPy is in Python, so it must be Python. Sorry bro :P
zie1ony··on Rust panics under the hood, and implementing them in .NET
I was thinking about code to code translation. You are right.
zie1ony··on Rust panics under the hood, and implementing them in .NET
In the past I have worked on Solidity -> Rust compiler, to enable Solidity on WASM VMs. My pain point was emulation of C3 inheritence in Rust, which I actually was able to implement with a few macros. In Rust -> .NET I'm interested in how he does the `DerefMut` trait.
zie1ony··on Half of code will be AI-generated soon – Newsletter for developers to keep up
AI-generated code is the next abstraction over programming languages.

GitHub Copilot started gently, but what I seen now is a race for becoming THE codegen AI. I think large companies already realized how big the AI CodeGen space is going to be. Future might be very different. Some say we will have one AI to know how to code in every language using every library. Other points that we might have AI-tied to the specific repositories, that can maintain them, solve issues and provide code examples for other AIs to use in projects.

Going this rabbit whole I created a newsletter where I'll put the latest development and research from this battleground. It would mean a lot to me if you subscribe.

What do you think?

I wrote this text without AI :P

zie1ony··on Penrose – Create diagrams by typing notation in plain text
For quick flowchart I use https://www.websequencediagrams.com/
zie1ony··on Nysa: Solidity to Rust Transpiler
Enable Solidity code to be usable for WASM-based blockchains.
zie1ony··on DragGAN source code release: Interactive Point-Based Manipulation of Images
This is nuts!
zie1ony··on You don't need Scrum, you just need to do Kanban right (2022)
Unpopular opinion: if your team is small enough (up to 5 people) you don't need any methodology.
← PreviousPage 2 of 3Next →