News for hackers. Stuff that matters.

renaudpawlak

Re:I think I have LLM burnout on Thursday July 09, 2026 @02:59AM (#48841953)
It looks that what you describe is partly a "burnout", partly a "sickness" of always the same LLM tricks and output (including errors). Of course LLMs tend to go back to their initial training and even if you "teach" them right, the attention mechanism make them forget things that are not often used (that's the KV-cache) even if they can be important for you (there is room for improvements here).

That said, your reaction is totally human. I personally get sick of how the LLM writes prose with always the same tricks and formulas (even if you prompt it not to). Humans need variants and novelty, that's why fashion exists. We get fed up with repetition and after seeing too much green shoes, seeing a red one is so relieving :) (quick note: I don't like fashion - I'd advocate diversity and personal styles, not fashion)

But that's also the way you work with AI that might be part of the problem. Personally, I don't review all the code the AI generates. I look at it, and I review only the code that matters. And with time on a given project I review less and less because I trust more the architecture and ability of the AI to follow it. In my settings, the AI gets confined to the existing architecture (that we define together at the beginning of the project), and has to ask for authorization to create new things (that's when I review the more). Hoping this could help to avoid burnout myself...

Show HN: LocalFlow – AI data analysis where the LLM never sees your data

https://github.com/localflow-ai/localflow-core

Hi HN, I'm the author of the JSweet project (Java-to-TypeScript transpiler), but today I'd like to show you something new: LocalFlow.There are two common ways to analyse your data with AI ("Talk to your Data"): 1. Use powerful frontier models on the cloud — but hand over your data and pay per token 2. Use on-premise or local models — but accept weaker results and significant hardware...

( Read More... | no comments | 2 points | on HN )
Re: on Monday December 16, 2024 @10:11AM (#42431723)
Hello HN, at LocalFlow, we have been working on a new kind of software component following lean methods, with the help of low-code tooling. I am curious to have you feedback.
Hello HN, at LocalFlow, we have been working on a new kind of software component following lean methods, with the help of low-code tooling. I am curious to have you feedback.
Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Tuesday December 12, 2023 @08:48AM (#38612001)
Did not find either :/
Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Sunday December 10, 2023 @04:26PM (#38595226)
You can take a look at https://github.com/j4ts, which contains some Java libs transpiled by JSweet to TS/JS. The AWT/Swing implementation is just a proof of concept, but some other libs are fully functional like awt/geom.

However, you have to be aware that the initial purpose of JSweet was not to port all Java libs to JS. JSweet allows the transpiler to be customized (with extensions) to map Java APIs to JS ones so that you don't necessarily need a JS runtime. It's all explained in the "Extending the transpiler" section of the core doc: https://github.com/cincheo/jsweet/blob/master/doc/jsweet-lan...

Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Sunday December 10, 2023 @04:16PM (#38595139)
Thanks :) I hope too. The project deserves it.
Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Sunday December 10, 2023 @12:40PM (#38593246)
Shutting down the projet and/or change the OS license to make it easier to fork is a potential option. However, I think it's worth asking if some organisation could be interested first. I believe JSweet could be more helpful if it was backed by some org that would have a use case of JSweet and maintain it for it's own purpose.
Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Sunday December 10, 2023 @12:32PM (#38593193)
JSweet comes with a TS definition files to Java translator (.d.ts to .java). So when you have the TS definition file, you can generate the correspond Java API on-the-fly or use a pre-translated/pre-compiled definition in so-called "candies". The candies are just here for typing in Java and ensure that you have the exact same API as in TS. Then JSweet generates intermediate TS code, which is compiled against the original TS definition. This 2-step transpilation process ensures that the generated code is correct and well-typed. In the end, you can work either directly with the JS code or the TS code depending on what you need.
Re:Ask HN: Anyone Interested in Taking over Jsweet.org? on Sunday December 10, 2023 @12:24PM (#38593141)
Yes indeed

Ask HN: Anyone Interested in Taking over Jsweet.org?

Hello HN community. JSweet.org creator/maintainer here. Being an individual and having other projects right now, I don't have much time to maintain JSweet anymore. I am looking for an organisation that would be interested in taking over the maintenance and / or ownership. Would anyone have an idea or connection to introduce? I would be grateful for any suggestion.

( Read More... | 16 comments | 39 points | on HN )
Re:The 8th "Forgotten" Principle of Local First Development on Tuesday October 10, 2023 @06:47AM (#37830487)
In this blog post, I propose to add Eco-Designed as an 8th principle to the original 7 principles of Local First.
Re:Ask HN: What happened to vanilla HTML/CSS/JS development? on Saturday September 10, 2022 @01:31AM (#32788072)
In a UI, you usually apply some sort of MVC pattern (Model-View-Controller), which means that you have a Model (your data, which usually comes from a server, or can be created by the user interacting with the UI), and the View, which is implemented with some HTML / DOM elements and reflects what the user sees. Then you have some sort of components (playing the role of Controllers) that are responsible (amongst other things) of syncing the Model and the View both ways (when the users interacts, with the view, it might update the data, but also, when the data changes for some reason, it might update the view).

The strength of the frameworks (React, View, Angular), is to provide a way to create well-scoped (local/small) components that take care of the Model (the data), the View (generally what is called a template), and have a built-in reactive capability that facilites the sync of the Model and the View. That reactivity is extremely powerful because it "watches" all changes of the Model and knows when a change is impactful for the View. It then refreshes the view automatically so that the user instantly sees the impact of its action or any state change in the application.

When you don't have reactivity, it is of course possible to build a UI, however, you will have to rigorously and explicitly update the view yourself. You can use an observer pattern to do so (btw, under the hood, reactivity is provided by an implicit and automated observer pattern, so that all data of the model become "observable"). In any case, when you don't have reactivity, it gets more complicated and you have to really know what you are doing because your UI code can become a big mess really easily.

You asked for an example. Imagine you have a UI with the check box and when the user checks it, it shows an input so that the user can enter some text to precise some data. In general, you can implement that behavior really easily with a reactive framework. Here is the pseudo code (not meant to be compiled).

The view: <input type="check" model="checkvalue" /><IF condition="checkvalue"><input type="text" model="textvalue" /></IF>

The model: { checkvalue: false, textvalue: undefined}

The important here is the IF tag, which depends on the checkvalue of the model. When this value is false, the text input does not show and vice-versa. It is fully automatic thanks to reactivity. If you don't have reactivity, you need to refresh the view manually when checkvalue changes, and you can imagine it can become quite messy when the view and the model get bigger :)

All that being said, reactivity, if employed poorly, can also become a mess and tigger a lot of unnecessary updates of the views. That's probably why some people complain about framework performances. Of course the role of the framework is also to give you a "frame of work"... if you go out of the frame, that's usually quite bad...

And finally, you can completely develop a small UI with no reactivity applying a MVC pattern with explicit update. That would just make the code more verbose, but IMO, it is doable if you know what you are doing. And it can probably be more efficient in some cases because you don't rely on the framework magic (you can decide in your program when to update the view).

Re:Ask HN: Which NoCode platforms are fine? on Tuesday October 26, 2021 @02:06AM (#28996950)
For the record: dLite.io is a low-code platform I have been working on to build front ends (similarly to bubble.io). It is ongoing work, but it already has many examples (and the website itself is made with dLite.io). It differentiates from others because it comes with a built-in support of the local-first paradigm. (web site: https://www.dlite.io, platform, with examples: https://plaform.dlite.io). (local-first paradigm explanation: https://cincheo.com/2021/10/03/when-low-code-meets-local-fir...) [disclaimer: I am also the author of JSweet.org, but right now my interests are moving towards low-code!]
Re:Ask HN: Which NoCode platforms are fine? on Tuesday October 26, 2021 @01:57AM (#28996889)
If you use a no-code front-end builder (like Bubble.io), I guess you could use any UI test framework to test your app. It is not the same as unit testing, but it is more global and in the end, UI tests are quite powerful to detect regressions that matter to the users.

That said, if you think that unit tests are always required, then I think that you are a victim of code complexity. Let me explain. I do believe that unit testing is good to master complexity, which often becomes out of hands with classical approches because the code is complex by itself (new lib versions popping up all the time, breaking refactoring). The thing with low-code is that it lowers that complexity by proposing more declarative approaches, which are more stable and less broken when changing versions (think of a static web page for instance, you don't really need to test it that bad although the browsers versions bump all the time).

I think that it is more important to be able to control complexity in order to avoid testing, rather than having to test "no matter what", like your reaction suggests. Plus, testing implies CI/CD and a lot of automation that are extremely complex to set up and consume a lot of resources.

I wrote a blog post about this stand-point lately https://cincheo.com/2021/10/03/when-low-code-meets-local-fir...

DISCLAIMER: I am currently working on a new low-code platform for frontend development based on the local-first paradigm. (web site: https://www.dlite.io, platform, with examples: https://plaform.dlite.io).

Re:Ask HN: What are you working on? on Friday January 15, 2021 @03:11AM (#25788252)
I have a web page for this :) https://cincheo.com/renaud-pawlak/ You will find all my (side) projects, including the ones in standby and failed/stopped ones. Not sure these projects all fit the "cool" criteria though! ;) Never applied to HN so far since I am not sure my projects fit.
Re:The other kind of JavaScript fatigue on Sunday July 17, 2016 @01:14PM (#12110843)
JSweet is not a magic bullet in the Java world, because JSweet is not Java. JSweet borrows the Java syntax to have well-typed JavaScript. Of course, not all transpilers have a focus on typing and not all of them have great tooling to back refactoring (not only renaming... you have plenty of other kind of help you will get when using a good Java IDE). But the focus of JSweet is typing and tooling (through Java).

I think that Christian Maioli Mackeprang is right. Some languages focus on instant productivity. They are really good and powerful to sketch a program in a few lines of code. But this quality comes at the expense of other important features that are really important for a language/development environment. People need to consider this or we will end up with poor programming environments and languages that are great for prototyping, but that don't scale up for complexity...

Older »