Ask HN: What are you working on (August 2024)?

634 points by david927 ↗ HN
What are you working on? Any new ideas that you're thinking about?

1,459 comments

[ 3.0 ms ] story [ 435 ms ] thread
(See https://news.ycombinator.com/item?id=41132906 for a bit of recent context)
almost 1000 comments suggests that this is very popular. doing this monthly would really help, although people may want to repost, so maybe some rule is needed for that. probably marking a post as "REPOST" or "NEW" would help regular readers find new projects.
Adding "REPOST" or "NEW" does nothing to ameliorate the problem of the same comment occurring over and over in search results (particularly hn.algolia.com/?sort=byDate searches).

On the other hand, if each repost consists of just a link to the original post, or a link plus a few words, that would prevent the unpleasantness in search results.

fair point, but i think linking to the original post would make it more difficult to scan though the page, and people would simply not read those posts. might as well ban reposts altogether. so i think if reposts make sense then a tag is better than a link. if links are used they would at least need a short description, so for those that wrote a more detailed post at first, a link with a short desctiption would make sense, but that would benefit from a tag as well.

other tags such as "UPDATE" or "MAJOR UPDATE" could also be considered.

maybe something like this:

Please tag your posts as NEW, REPOST or UPDATE. For reposts and updates try not to repost the whole description but keep it short and link to the previous post instead

(comment deleted)
I am trying to get better at trading so I thought it would be nice chance to work on my coding skills. I wondered if financial indicators actually makes money for a given period.

https://indicatorinsights.co

Made it with React, Vite, firebase via Typescript. It’s a really different world for sure.

Your MACD seems to be inverse. It suggests a buy when it should be selling and vice versa.
Yes, there is a bug somewhere which I could not find so far… it is making me crazy. Yet, I will find it at some point. Other than that, I am still improving the site!
Recently released an iOS that helps you to see your Plausible Analytics data on the widgets: https://charsible.app/

Right now making improvements based on the feedback.

I’m back to working on an ML app pair (train/consume) for Mac/iOS, for finding a certain type of resource.

I tried it a while back, and wasn’t too thrilled with the way Apple has structured their ML stuff (they basically only afford very specific applications, which didn’t match my workflow).

It was suggested that I revisit it, so I am.

I moved to doing programming for some post production companies and I am working on a couple of patches for ffmpeg.

First time writing C and the task has been both daunting and rewarding, I am mostly working hls enc/dec and learning about av has been a ton of fun!

That's awesome! Do you know of any resources for learning more about audio/video compression?
Working on an update to `hiedb-mode`. It's an emacs package for querying an hiedb index; a database for building IDE tooling for Haskell. The update will use `compile` for the display buffer similar to how `find-grep` does it.
Despite not liking JavaScript as a language nor the Node runtime, learning about hybrid static and SSR apps with Astro.js - I have to say, I’m impressed so far. Seems like a sweet spot for certain types of sites.
Building more Amazon scraper websites. Working on a framework to minimize the effort of creating them, and currently testing to see if I can automate most of the process with Anthropic’s Sonnet.

The most recent site is https://computerdisplayprices.com

When I’m done, I’m probably going to make a blog post or two just talking through the automation I figured out so far. LLMs are amazing force multipliers.

what toolset do you use for scraping? puppeteer?
Nope, just selenium and pyautogui. The priority is to fly under the radar, rather than efficiency; for that I needed lower-level control. There's a bunch of other stuff I did to improve the anonymity, but to download I just open selenium at the url, wait for it to load, then use pyautogui to type 'ctrl-s' and type in the save directory.
(comment deleted)
I dusted off an old project today. It's a browser based CHIP-8 emulator with a built in editor.

https://chip8.dotslashdan.com/

I added the ability to import/export ROMs and finally added some of documentation.

...Plenty bugs left to fix

Working on: https://www.mylondonhomesearch.co.uk/

To help Londoners find their perfect home by filtering for proximity to parks, river, PoIs, commute time, etc.

The website is not optimized for mobile and customers have to pay to access the functional filters/properties.

working on a website showing people how micropayments could work today
can you share it? or at least what you know about the topic.
I'll share it when I've made something worth sharing :)

It'd be a basic guide that says do a,b,c,d,e,f,g to give me 10 cents, where a-g is

a) create a Coinbase account

b) fund it

c) buy some crypto

d) install a wallet

e) move the crypto to the wallet

f) connect the wallet to my website,

g) send 10 cents my way

it's not going to offer any choices of crypto and wallet, just a proof of concept to document that the rails do exist today.

are you going to cover the whole process end to end, from my cash to my friends cash assuming neither of us have a crypto wallet yet?

what about transfer fees?

all in all this is way to complex for the average person and inaccessible for many.

in some countries mobile money accounts come with every phone number, and you can transfer micropayments from one number to another even across networks and even internationally. and you don't even need a smartphone. any feature phone that can send USSD codes will work. that is a working micropayment system accessible to anyone who can afford at least a feature phone.

It's way too complex. But it's possible. I'm in the US, my site will be targeting US users. if you want to do something for EU users with USSD, I'd love to collab, email in profile.
ok, it is reasonable to focus on the situation in the US.

my point about USSD codes was that a globally applicable solution for micro payments does exist, and using a complex system with many steps really just begs the question of what prevents these simpler systems from being adopted.

the answer is probably politics and lethargy. the inability or unwillingness for big companies to innovate.

to be clear, i am not criticizing your work, but the state of micropayments in the US, and i am mostly interested in discussing what can be done about that.

showing people how it can work now may or may not be part of that. personally i am more inclined to look at this process and run away screaming than make any attempts to adapt it.

yeah, fair. I'm expecting most reader's eyes to glaze over so I'm expecting zero money out of this, given all the hurdles I hit to accomplish this task that should have been easy. but the first step is to document the current state of the world and get that out there.
Working on building a copilot for oncall engineers.

Goal is to automate or reduce the grunt work oncall engineers have to do.

Code is here: https://github.com/opslane/opslane

I've been working on similar lately. Changepoint detection over our metrics time series, and using that to inform automated root cause analysis.

I think your work was mentioned on our team channel this week.

I found these papers to be pretty interesting:

https://arxiv.org/abs/2305.10638 https://arxiv.org/abs/2402.02357

Anything else I should be reading?

Right now I'm mostly thinking about AI. Right this minute I'm sitting in a cafe reading a book on Multi-Agent Oriented Programming with a framework called JaCaMo[1], and later tonight when I get home I'll probably spend some time getting my Fuseki[2] server loaded with some base schemas (SKOS, FOAF, etc) as I slowly start working on getting things set up to explore some ideas around integrating symbolic logic with LLM's.

And I took my new quadcopter drone out and did some flying last night for the first time. As in, my first time flying a drone, ever. The results were... predictable. Let's just say, I bought a cheap (< $100) drone for a reason. This thing will wind up destroyed. In less than an hour I managed to crash it into fences, walls, bushes, cars, dumpsters, the ground, an armadillo, Elvis Presley, a 1974 AMC Gremlin, and Nickelback. Well, more or less.

It brought to mind this famous scene[3] from the movie Days of Thunder:

Harry: I want you to go back out on that track and hit the pace car.

Cole: Hit the pace car?

Harry: Hit the pace car!

Cole: What for?

Harry: Because you hit every other god-damned thing out there and I want you to be perfect.

[1]: https://jacamo-lang.github.io/

[2]: https://jena.apache.org/documentation/fuseki2/

[3]: https://www.youtube.com/watch?v=xll0VOsiE84

lol'd at the 1974 AMC,

As an experienced drone pilot - if you can, DJI is unbelievable at stopping you from crashing. Start there to get the feel of it (and more importantly not get frustrated and give up)

If you want to learn more freestyle without assistance, i can't recommend enough getting a simulator on your computer and crashing a digital drone a million times first. It's much cheaper. I think there is a lot out there now but i used Liftoff for many hours before i tried to fly a proper drone

I was wondering. I have a DJI and it is stupid simple to fly. I’ve yet to crash mine but I don’t fly it a ton. I’m certainly not trying to fly it through complex paths like an open doorway and into a house or something.
Yeah, I'm flying a super basic drone with none of the built-in crash avoidance / navigation / etc. stuff for now. I'll upgrade to something fancier later, but for now I'm totally satisfied with something I can crash repeatedly and when it breaks, throw it away (or salvage parts from it to build another drone).

I'll look into the simulator thing. I didn't even realize that drone simulators were a "thing" so I had not gone looking for anything like that.

I’m working on a screen play called DRONE RACER about kids who race human sized quadcopters through the subways of NYC. It’s a mix of Hackers and Fast & The Furious.

Using a lot of off the shelf AI and pushing the limits of what’s possible.

Really fun.

It’s a mix of Hackers and Fast & The Furious.

If HN supported images in replies, I'd totally be posting that "shut up and take my money" meme right now. As it is, I'll just say "that sounds awesome, and I hope it sees the light of day eventually."

Thanks God for HN not supporting inline image.
(comment deleted)
Same … no potential to over expose my brain to hyper VAD memetic warfare is a feature that keep me contributing here.
I would highly recommend getting a drone simulator off Steam. You can practice the controls and drill them into your fingertips during times when you can't fly the real thing (battery recharging, night time, weather, ...). Most advanced radios allow you to hook them up to the computer so you can fly with the exact same inputs, but a console controller is also acceptable.
A feature-rich, nice looking audio player for those who don't stream music: https://sakunlabs.com/muziqi

It has 90% of the features that I wanted, and now I am working on what my users want.

Was hoping this was for desktop, as ive been using winamp for roughly 10 years since it has everything i want, including android management support.

Regardless, keep up the good work!

my list of features needed for a usable audio player:

for audiobooks the ability to skip forward and backwards a few seconds. the best way to do that i found is to swipe left and right. short swipe is a short skip, swipe across the whole screen is maybe a minute. also the ability to quickly jump to a certain manually chosen timestamp.

also for audiobooks, the ability to bookmark positions. not all audio books have chapter markers.

for audiobooks and podcasts: list files by their names in storage sorted by timestamp or name. many podcasts and chapterized audiobooks don't set chapter or episode titles correctly. i have seen players where the title was always the title of the book or the name of the podcast, making it impossible to identify which file was which episode or chapter. also author names and even podcast titles are sometimes inconsistent, because spelling errors, different authors or multiple names for a podcast.

do not assume that the file metadata is consistent or correct and make it easy for the user to work around inconsistencies.

I think Muziqi ought to satisfy most of these for audiobooks. Podcasts is on the list.
I'm writing a new property-testing library for TypeScript. [1]

I wanted a way to define new Arbitraries that's easier than working with .map(), .chain(), .oneOf(), and other combinators that require you to think in terms of sets. It has those combinators, but you can also write code to randomly generate one value at a time. It uses an approach to shrinking that's inspired by Hypothesis.

Along the way, I ended up adding a Domain subclass, which also does validation like Zod. Not sure where I'm going with that, but it's also useful for generating unique keys.

(The documentation isn't done and doesn't explain what's interesting about it. Caveat: it's Deno-only, and will probably stay that way unless someone wants to help.)

[1] https://jsr.io/@skybrian/repeat-test

I'm currently building a mobile-friendly Postgres database tool. Think dbeaver but on Android.

Writing and editing SQL queries on the phone can get tedious. Rather than just giving users a tiny textfield, I am exploring if there are better ways to build SQL queries on a touchscreen

Here's the makeshift landing page: https://getselectable.com/

I'm putting qnx cross-toolchains into Debian packages because $DAYJOB builds qnx images in ubuntu containers. I've learned a lot about cross-toolchains and my appreciation for blackberry's technical prowess has never been lower.
I had this realization that, sometimes, people on my team are doing the same tasks in different ways, resulting each of them taking different amount of time. In certain scenarios, there is an optimal way of doing things, and I don't even know about it. Trying to test out if current AI-models are solid enough to do comparisons from a screen recording, or just general suggestions for improving white-collar workflows. I could see an eventual platform that could be built out of it.
Is it possible that you're describing design patterns here?
Hm, if you mean software design patterns, then not exactly. It's probably harder to detect in just coding world, but the easiest one I can think of in our realm "checking current CPU usage of X service". In my current company, there are quite a few ways - logging into AWS -> switching to production environment -> going to ECS -> Looking up the service -> Clicking on metrics, or opening DataDog -> Dashboards -> Service Name.

I understand it's very reductive and simplified, but for every task there are alternative ways to reach the goal, and from my historical experience, everyone does everything differently. Some things can be optimized significantly, if others just knew about the possibilities.

It's possible they aren't talking about code at all.

"This team member, and this other team member are both filing a particular kind of report but are doing so in very different ways, is one of them more effecient?"

I like it. Seems like there could be lots of ways to automate the creation of SOP for many businesses. I do wonder how high level of tasks it could pick up on. Though maybe that gets better with a bigger context window.

This is exactly it! I'm sure I do a lot of things slower than others as well, and simply never came across an "alternative, but faster" method.
That's fair. Assumption on my part. Hence couched in "is it possible..." weasel words.
A new CI/CD platform. Local CLI (run without git push), remote debugger, automatic content-based caching, DAG-based definition, and dynamic tasks. It’s taken a lot of work to build it, but I’m really excited about how well it’s working. https://rwx.com/mint
how does it compare with dagger?
I haven't tried this yet, but I wanted to say I think there's a lot of potential in this space. There's so much friction with the current popular solutions...and yet it's so hard to justify trying some of the newer and less popular ones.

I wish you luck because there are a lot of good ideas in here. Running locally and remote debugger are the most exciting to me.

I like the idea, but I have used Azure Pipelines, and the key differentiating it from GitLab.ci is its "service connection" that enables seamless integration with third-party services and a wide range of tasks (PowerShell, pscore, az, docker, etc).

I wish your dependency semantics yaml could be converted to Azure pipeline task, job, and stage similar to how typescript converts to javascript.

I am saying this because there are very complicated infrastructural problems that hundreds of people work on at big companies when creating a ci/cd I see you are not focusing on them yet but that DAG logic could be applied to existing azure pipeline yaml.

This seems really neat! I'm a huge proponent of incremental CI for any project that gets big enough to care about test speed.

I'd be pretty worried about an abstraction ceiling on the caching here relative to something like Bazel, though. Like would Mint get mad if I generated 4000 dynamic tasks?

I love Captain! I would have been over the moon to see this project ~1 year ago when I was porting our CI away from Circle CI.
Out of interest: CircleCI also has flaky test detection, rerun failed tests etc. Or did you mean that the CI that you were moving towards could benefit from Captains features?

(disclaimer: i'm a CircleCI employee)

CircleCI was incredibly expensive. We're at 1/30th of the cost with self-hosted GitHub actions.
I didn't make myself clear I think. You mention that you love Captain, and wish you had access to it when migrating away from CircleCI. I'm interested in understanding the specific reason(s) for your love for Captain's features, that seem to also be available in CircleCI?
When we moved away from Circle we had to come up with a replacement for flaky detection.
Created this website for a local rubble removal business.

https://legendsfreight.co.za

I’d love to get feedback on what can be done to improve the website to feature prominantly for local search when people need rubble removal near them.

Not an SEO expert but I've found that adding your business to Google Maps will get you many more eyeballs.
Thank you for this.

I created a Google my Business listing.

Will that suffice or is there a specific step required to make it available in Google Maps?

Google will review the listing and eventually add it to the map, no further action required by you. Once they do, you'll want to get your customers to write reviews.
So essentially just a waiting game then. I would happily do more dev or add more features, but I am not really sure what more to add / or do.
I am building a belt tension meter to attach my z-belt idlers to on my voron 2.4.

I'm still contemplating what I want the final form to look like, but currently it's a 5 kg load cell and a floating idler and an rp2040 reading values via hx711.

Are you looking at alignment too? Huge impact on the longevity of belt-drive systems.
No, I haven't considered alignment. I think they're pretty well aligned already, but honestly haven't thought about it much at all.