Hacker News Reader

lemmingon Launch HN: Bitrig (YC S25) – Build Swift apps on your iPhone

With vibe coding, I generally find that it will get me 80-90% of the way to where I want to be, but I generally end up having to do some work too. How does that work here? Is there any way to later edit the app on my mac if I need to? Similarly, once my app is "done" for some definition of done, can I compile it into a native app rather than an interpreted one?

lemmingon Show HN: Stagewise – frontend coding agent for real codebases

Unrelated to the actual product (which looks nice), I'm always curious when I see these landing pages with logos from massive companies on them. I'd bet a beer that you didn't go through Oracle's processes for using their logo, for example. Are you just YOLOing it and hoping they don't come after you for unauthorised use?

lemmingon Show HN: Base, an SQLite database editor for macOS

So the OSS thing is a preference, obviously, if you like OSS tools, just use them - it's not like you're short of options.

But... expensive? Really? A one-off cost (not a subscription, mind) for a tool that might make an expensive developer/data person even marginally more productive? You have to be kidding. It blows my mind how much developers, who of all users should be the ones who appreciate how much work goes into something like this, complain so often about the cost of basic tools.

lemmingon AGENTS.md – Open format for guiding coding agents

You say they have someone drive it.

I didn't say that. But they're not fully autonomous.

We aren’t arguing about pure autonomy, we are arguing about the method by which humans resolve the problems.

This whole subthread started with the assertion:

Just like we will never have fully self-driving cars...

So we did start out by discussing whether current Waymo is fully autonomous or not. It then devolved into nit-picking, but that was where the conversation started.

FWIW I agree that Waymo is an amazing achievement that will only get better. I don't know (or care, frankly) if they will ever be fully autonomous. If I could, I'd buy one of those cars right now, and pay a subscription to cover the cost of the need for someone to help the car out when it needs it. But it's incorrect to say that they don't need human operators, when they clearly currently do.

lemmingon AGENTS.md – Open format for guiding coding agents

That’s irrelevant though. If the system requires human intervention, then it’s not fully autonomous by definition. See https://rodneybrooks.com/predictions-scorecard-2025-january-... for example:

The companies do not advertise this feature out loud too much, but they do acknowledge it, and the reports are that it happens somewhere between every one to two miles traveled.

That’s… not very autonomous.

lemmingon AGENTS.md – Open format for guiding coding agents

you are incorrect as they do not have a remote human operator

Yes, they do, the term to search is “remote assistance operator”. e.g. https://philkoopman.substack.com/p/all-robotaxis-have-remote...

lemmingon I accidentally became PureGym’s unofficial Apple Wallet developer

FWIW I enjoyed the article and the humour, and I don't know where the AI conspiracy is coming from – I wish I could get the AI to write copy this good. So thanks, that was a fun read!

lemmingon Study: Social media probably can't be fixed

Fortunately we don't have to rely on your anecdata, people actually study this stuff:

https://news.gallup.com/poll/467792/social-media-users-incli...

U.S. adults commonly engage with popular social media platforms but are more inclined to browse content on those websites and apps than to post their own content to them. The vast majority of those who say they use these platforms have accounts with them, but less than half who have accounts -- and even smaller proportions of all U.S. adults -- post their own content.

https://www.pewresearch.org/internet/2019/04/24/sizing-up-tw...

Most users rarely tweet, but the most prolific 10% create 80% of tweets from adult U.S. users

https://www.pewresearch.org/internet/2021/11/15/the-behavior...

The analysis also reveals another familiar pattern on social media: that a relatively small share of highly active users produce the vast majority of content.

lemmingon Study: Social media probably can't be fixed

LLMs aren’t people, and the authors have not convinced me that they will behave like people in this context.

This was my initial reaction as well, before reading the interview in full. They admit that there are problems with the approach, but they seem to have designed the simulation in a very thoughtful way. There really doesn't seem to be a better approach, apart from enlisting vast numbers of people instead of using LLMs/agent systems. That has its own problems as well of course, even leaving cost and difficulty aside.

There’s no option to create original content...

While this is true, I'd say the vast majority of users don't create original content either, but still end up shaping the social media environment through the actions that they did model. Again, it's not perfect but I'm more convinced that it might be useful after reading the interview.

lemmingon Artisanal handcrafted Git repositories

Git refers to the user-friendly commands as “porcelain”

Ahhhhahahaha… “user friendly”. When compared to coding the repo by hand, I guess.

lemmingon Show HN: I built an LLM chat app because we shouldn't need 10 AI subscriptions

I'll bite: what is the web search story like? This is the killer feature of e.g. chatgpt that none of the alternative or OSS options offer. Having the search be fast (i.e. not round tripping to the client on every search) and integrated into the thinking process is unbeatable, I use it constantly. The big API providers all provide search options in their APIs now, but they're very quirky - openai doesn't allow search via their API with thinking models. Gemini doesn't allow you to use search with any other tools available. Claude's just doesn't seem to work that well, even in their own web UI. I even paid for typing mind, which is nice, but I never use it and always just end up paying for chatgpt again because of this.

lemmingon Kite News

I'm very interested in this, I'm a long term kagi subscriber with the free t-shirt!

One thing that would be interesting to explore would be a feed that integrated over a longer timeframe. For example, I subscribe to the guardian weekly, and I like it much more than the daily homepage. News that survives a week tends to be more newsworthy, and there's more scope for the beginnings of retrospective analysis. It would be interesting to be able to play with that timeframe - what does a 3 day window look like, or a 1 month one?

lemmingon The End of Moore's Law for AI? Gemini Flash Offers a Warning

Can anyone explain the economics of Anthropic's Max plan pricing to me? I have friends on the $100/month plan using well over $800 of tokens per month with Claude Code (according to ccusage). I certainly don't use Claude Code as much if I'm not on a flat rate plan, the cost spirals out of control very quickly. I understand that a subscription makes for more predictable revenue and that there will be people on the Max plan not using Claude Code 24/7, but the delta between what the API costs and what using the Max plan with Claude Code costs just seems too great for that to be an explanation. I don't think that user/mindshare capture can fully explain it either, Code is free and the cost of switching to something else if pricing later changes is just too low. I don't get it.

lemmingon Lateralized sleeping positions in domestic cats

Wait till I tell my daughter that someone's job is watching cat videos on the internet for science!

lemmingon Show HN: Gradle plugin for faster Java compiles

The underlying Elide tool looks amazing, and potentially solves all sorts of pain for me. Does the Kotlin compiler support plugins, e.g. serialisation? I can imagine they might be too reflective.

lemmingon Cloud Run GPUs, now GA, makes running AI workloads easier for everyone

If I understand this correctly, I should be able to stand up an API running arbitrary models (e.g. from Hugging Face), and it’s not quite charged by the token but should be very cheap if my usage is sporadic. Is that correct? Seems pretty huge if so, most of the providers I looked at required a monthly fee to run a custom model.

lemmingon Consider Knitting

Do you have any recommendations for online resources? A few years ago my wife and I started with Justin guitar, but then we switched to piano to tempt our daughter. Is that a good choice, or are there better options?

lemmingon Hacker News now runs on top of Common Lisp

Chez compiles, I think, its not a VM. It’s not as fast as SBCL of course, but it’s not interpreted.

lemmingon Show HN: Generate SVGs with AI

The results actually look pretty impressive! Can you go into any details about how the generation works? Are you generating images and then vectorising?

lemmingon Show HN: SVG Animation Software

I have no problem at all with your licensing model, I currently use the same one myself. I just think your messaging needs to be clear.

lemmingon Show HN: SVG Animation Software

This isn't true for JetBrains any more, and hasn't been for ~10 years. They run on what is essentially a pure subscription now, with the caveat that if you stop paying, you can continue to use a version of the software. However, it's not the version you're using at the end of your subscription period, it's the version from 1 year prior to that. So if you stop paying, you will probably have to downgrade. See: https://sales.jetbrains.com/hc/en-gb/articles/207240845-What....

lemmingon Show HN: SVG Animation Software

I sell software with a similar licence. I'd say that get access to all features and upcoming updates is incompatible with Then, the updates will stop. At a very minimum it's confusing, but it sounds fairly deceptive to me. I think being really clear about what users will receive is paramount, even if it ends up making the messaging confusing.

lemmingon America is in danger of experiencing an academic brain drain

I see zero upsides of reducing brain power

You have fewer annoying people pointing out logical problems with your policies.

lemmingon Codebase to Easy Tutorial with AI

The code behind this was posted a while back (https://news.ycombinator.com/item?id=43739456), but this project now has a nice easy to use UI. Here's the accompanying video: https://www.youtube.com/watch?v=wJ42uIVGFA4.

I'm not affiliated with the project, but I think that PocketFlow is really nice, and I love the example projects Zachary uses.

Edit: There was also some interesting critical feedback on the produced documentation: https://technicalwriting.dev/ml/pocketflow/index.html

lemmingon Windsurf SWE-1: Our First Frontier Models

The company that is best placed to collect tons of high quality data of this type is undoubtedly Google. They’ve had publications talking about how they capture data from their in house SWE tools and use it to improve their tooling.

lemmingon Show HN: I’ve built an IoT device to let my family know when I’m in a meeting

Mine is even simpler - I just control an off-the-shelf smart bulb via a couple of macOS Shortcuts, which is trivial using Homekit. Those are driven by Hammerspoon which triggers the shortcuts based on the camera events. No custom hardware and like 20 lines of code (not including the shortcuts, I guess). I have the bulb in a normal lamp base outside my door. It's properly event driven rather than polling, too.

lemmingon Web search on the Anthropic API

Trafilatura to markdown? But yeah, likely to be clunky.

lemmingon Web search on the Anthropic API

I would really love this too. However I think that the only solution for that is to give it a Kagi search tool, in combination with a web scraping tool, and a loop while it figures out whether it's got the information it needs to answer the question.

lemmingon Web search on the Anthropic API

I'm also interested to know if there are other limitations with this. Gemini, for example, has a built-in web search tool, but it can't be used in combination with other tools, which is a little annoying. o3/o4-mini can't use the search tool at all over the API, which is even more annoying.

lemmingon Google Gemini has the worst LLM API

Additionally, there's no OpenAPI spec, so you have to generate one from their protobuf specs if you want to use that to generate a client model. Their protobuf specs live in a repo at https://github.com/googleapis/googleapis/tree/master/google/.... Now you might think that v1 would be the latest there, but you would be wrong - everyone uses v1beta (not v1, not v1alpha, not v1beta3) for reasons that are completely unclear. Additionally, this repo is frequently not up to date with the actual API (it took them ages to get the new thinking config added, for example, and their usage fields were out of date for the longest time). It's really frustrating.