HNHacker News
TopNewBestAskShowJobs

the_duke

17,811 karma · joined June 4, 2016

submissionscomments
the_duke··on Getting AI to work in complex codebases
It absolutely can be, by a huge margin.

You spend a few minutes generating a spec, then agents go off and do their coding, often lasting 10-30 minutes, including running and fixing lints, adding and running tests, ...

Then you come back and review.

But you had 10 of these running at the same time!

You become a manager of AI agents.

For many, this will be a shitty way to spend their time.... But it is very likely the future of this profession.

the_duke··on Hyperion: Minecraft game engine for custom events
If the plugins are in Rust and tie into Bevy, then there probably is minimal overhead. That's one of the advantages of ECS. It's designed for efficiently adding additional layers of functionality.
the_duke··on WASM 3.0 Completed
The problem with multi-memory (and why it hasn't seen much usage, despite having been supported in many runtimes for years) is that basically no language supports distinct memory spaces. You have to rewrite everything to use WASM intrinsics to work on a specific memory.
the_duke··on PythonBPF – Writing eBPF Programs in Pure Python
So this is a "inline" Python to eBPF transpiler/compiler.

Which is cool!

But the description could be a bit clearer.

the_duke··on Chat Control faces blocking minority in the EU
Newsflash: it's the paid professionals that are doing this...
the_duke··on Chat Control faces blocking minority in the EU
The tactic here is sneaking legislation through the system by bringing it up again and again, hoping for the public to eventually lose interest, or to catch a time with a lot of other drama going on so they can avoid the public attention/backlash.

I do think there are procedural ways to support this, like: proposed bills that are very similar to previous rejected ones need a preemptive vote with 60%+ support to be considered - if brought again with a certain time frame.

I do see your point though, there can be unforeseen consequences.

the_duke··on BCacheFS is being disabled in the openSUSE kernels 6.17+
This is a tragedy, bcachefs has so many great features...
the_duke··on Strong Eventual Consistency – The Big Idea Behind CRDTs
Sure, but you can usually rely on database transactions to handle the hard part.
the_duke··on Strong Eventual Consistency – The Big Idea Behind CRDTs
Nice, will have a look!

I've had similar thoughts, but my concern was: if you have idempotent actions, then why not just encode them as actions in a log. Which just brings you to event sourcing, a quite well-known pattern.

If you go that route, then what do you need CRDTs for?

the_duke··on Strong Eventual Consistency – The Big Idea Behind CRDTs
There are many ways to solve each individual problem.

The point is that you always have to think about merging behaviour for every piece of state.

the_duke··on Strong Eventual Consistency – The Big Idea Behind CRDTs
The big problem with CRDTs IMO is that they make it incredibly easy to break application semantics.

Just a basic example for a task tracker:

* first update sets task cancelled_at and cancellation_reason

* second update wants the task to be in progress, so sets started_at

If code just uses the timestamps to consider the task state, it would not assume the task is cancelled, unexpected since the later user update set it to in progress.

Easy fix, we just add a state field 'PENDING|INPROGRESS|CANCELLED|...'.

Okay, but now you have a task that is in progress, but also has a cancellation timestamp, which seems inconsistent.

The point is:

With CRDTs you have to consider how partial out of order merges affect the state, and make sure your logic is always written in a way so these are handled properly. That is *not easy*!

I'd love it if someone came up with a framework that allows defining application semantics on top of CRDTs, and have the framework ensure types remain consistent.

the_duke··on Google can keep its Chrome browser but will be barred from exclusive contracts
Gemini 2.5 always provides a lot of references, without being prompted to do so.

ChatGPT 5 also does, especially with deep research.

the_duke··on U.S. Emissions Rise 4.2%, China's Fall 2.7%
> They don't care about the climate

I don't think it's that simple.

China is a signatory to Kyoto and Paris.

They do care about reducing pollution, and have managed to do so quite significantly in many cities.

China also has quite a bit to lose: many large cities on the coasts, and worsening water shortage problems.

National security probably plays a large role, and I reckon they would prioritize economy over climate, but the evidence implies that they do also care.

the_duke··on Bear is now source-available
We are also entering the age of "hey AI, take this repo, reimplement the same functionality".

Now, no LLM is currently anywhere near doing that for ElasticSearch.

But for a project with 4845 lines of Python code? (as per tokei)

Definitely doable, with a bit of handholding and manual fixing.

Would that be a derivative work? Maybe, but that would be a hard legal battle.

the_duke··on A failure of security systems at PayPal is causing concern for German banks
To be clear: I didn't mean they have to drag their feet. They just do it because they will lose some income.
the_duke··on A failure of security systems at PayPal is causing concern for German banks
The resistance against instant payments mostly has to do with technicalities around bank balance sheets and inter-bank settlement systems.

It's a bit complex for a comment, but the TLDR is:

* funds in transit (called "positive float") are held in the banks account, and can be used by the bank to earn interest

* liquidity management - there are a bunch of considerations here, but the longer settlement periods enable banks to do "deferred net settlements" (just paying each other the difference between all transactions in a batched way) and also helps balance sheets in other ways, making it easier to meet reserve requirements, smoothing out intra-day liquidity, etc

The delay also means systems have more time to catch fraudulent transactions, and to block them before they happen.

the_duke··on Go is still not good
Rust async makes it quite easy to shoot yourself in the foot in multiple ways.

Most users writing basic async CRUD servers won't notice, but you very much do if you write complex , highly concurrent servers.

That can be a viable tradeoff, and is for many, but it's far from being as fool-proof as Go.

the_duke··on Go is still not good
I personally don't like Go, and it has many shortcomings, but there is a reason it is popular regardless:

Go is a reasonably performant language that makes it pretty straightforward to write reliable, highly concurrent services that don't rely on heavy multithreading - all thanks to the goroutine model.

There really was no other reasonably popular, static, compiled language around when Google came out.

And there still barely is - the only real competitor that sits in a similar space is Java with the new virtual threads.

Languages with async/await promise something similar, but in practice are burdened with a lot of complexity (avoiding blocking in async tasks, function colouring, ...)

I'm not counting Erlang here, because it is a very different type of language...

So I'd say Go is popular despite the myriad of shortcomings, thanks to goroutines and the Google project street cred.

the_duke··on This website is for humans
You only need cookie banners for third parties, not for your own functionality.
the_duke··on Blog series on creating an OS in Rust
Even if you don't care about building an OS: the async/await post [1] is the best resource for understanding how async actually works in Rust.

Highly recommended.

[1] https://os.phil-opp.com/async-await/

the_duke··on You can now disable all AI features in Zed
I found Copilot tab completion completion to be VERY slow in Zed, for some reason.

It's fast in neovim.

Maybe they artificially slow down non-official clients somehow? (official neovim plugin, VS Code)

the_duke··on Hyperpb: Faster dynamic Protobuf parsing that's faster than generated code
The delta to the performance of C++/Rust Protobuf implementations would be interesting.
the_duke··on You can now disable all AI features in Zed
I have the opposite experience, tab completion by Copilot just got significantly worse for me recently (the last week or so), both for Rust and Python code.
the_duke··on When Is WebAssembly Going to Get DOM Support?
Sure, in Rust the web-sys [1] crate provides auto-generated bindings for pretty much all the browser APIs.

Has been used by most of the Rust web frontend frameworks for years.

It all has to go through JS shims though, limiting the performance potential.

[1] https://docs.rs/web-sys/latest/web_sys/

the_duke··on Why is AI so slow to spread?
I'd disagree with that, if given enough compute LLMs can have impressive capabilities in finding bugs and implementing new logic, if guided right.

It's hit or miss at the moment, but it's definitely way more than "UML code generators".

the_duke··on Why is AI so slow to spread?
Just on the coding side, tools like Claude Code/Codex can be incredibly powerful, but a lot of things need to be in place for it to work well:

* A "best practices" repository: clean code architecture and separation of concerns, well tested, very well-documented

* You need to know the code base very well to efficiently judge if what the AI wrote is sensible

* You need to take the time to write a thorough task description, like you would for a junior dev, with hints for what code files to look at , the goals, implementation hints, different parts of the code to analyse first, etc

* You need to clean up code and correct bad results manually to keep the code maintaineable

This amounts to a very different workflow that is a lot less fun and engaging for most developers. (write tasks, review, correct mistakes)

In domains like CRUD apps / frontend, where the complexity of changes is usually low, and there are great patterns to learn from for the LLM, they can provide a massive productivity boost if used right.

But this results in a style of work that is a lot less engaging for most developers.

the_duke··on Mistral Releases Deep Research, Voice, Projects in Le Chat
That's Gemini Pro now in general. The initial preview was pretty good, but the newer iterations are incredibly verbose.

"Be terse" is a mandatory part of the prompt now.

Either it's to increase token counts so they can charge more, or to show better usage growth metrics internally or for shareholders, or just some odd effects of fine tuning / system prompt ... who knows.

the_duke··on Postgres LISTEN/NOTIFY does not scale
Kafka is far from trivial to operate, for one thing, even post zookeeper.
the_duke··on LLMs and Elixir: Windfall or deathblow?
I have been arguing for a while that very strict languages with a heavy type system are ideal for agent coding.

The stricter the language, the harder it is for the LLM to produce nonsense, at least if it can get compilation errors and run tests. And the easier it is to validate that the output is correct, because the types already tell a lot of the story.

A language with dependent types, linear types, etc... would be ideal, but alas...

At the moment Rust is the sweet spot. Fairly popular (and hence known to LLMs and with a fairly good ecosystem), great error messages to guide resolution of problems, stricter type system and more compile-time guarantess than almost all of the other semi-popular languages...

Now Rust isn't trivial to write, for both humans and LLMs, and the output was pretty bad for a long time.

But with the ability to run `cargo check` and execute tests, even the current first iteration of agents is really quite good at iterating until it gets a working result.

the_duke··on OpenAI slams court order to save all ChatGPT logs, including deleted chats
The EU has a certain amount of jurisdiction over all companies providing a service to customers located in the EU.

A US company can always stop serving EU customers if it doesn't want to comply with EU laws, but for most the market is too big to ignore.

← PreviousPage 5 of 34Next →