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

https://colorango.com/

Now I release my knowledge in bite-sized chunks on my new YouTube channel to help others:

What's a battery management system? https://www.youtube.com/watch?v=_QsMoCrSTYc

What is the C-Rate? https://www.youtube.com/watch?v=dDu1fRtKfsA

What is battery balancing? https://www.youtube.com/watch?v=RGYLPOlT45A

etc. etc. I focus really hard on answering exactly one question in a concise and engaging way and trying to keep every video under 5 minutes. Oh, and to make the videos solution independent, so not specific to a product, but convey the underlying knowledge so it has a longer shelf-life.

Full list is here: https://foxev.io/batteries/ I am planning to turn this into a knowledge base with playlists for "learning paths" like "everything to watch about batteries" or "here is what you need to watch to make a motor spin on a bench". I will add interactive functionality like quizzes and widgets to make the knowledge even more sticky.

Ask HN: What are you working on (September 2024)? - https://news.ycombinator.com/item?id=41690087 - Sept 2024 (1041 comments)

Ask HN: What are you working on (August 2024)? - https://news.ycombinator.com/item?id=41342017 - Aug 2024 (1424 comments)

One thing I'd like to hone in on is that these threads aren't intended for promotion, but rather for the just-because sort of project, driven by idle interest or weird obsession—the sort of thing people might spend their free time on.

I'm not sure yet what the official "rule" should be (if any), but if you're working on a startup or have had attention via Show HN, maybe abstain from these discussions? It wouldn't be good for the thread to get taken over by things HN already has a place for.

Video on what I'm doing - https://www.youtube.com/watch?v=CCiuS1oHnKw

Picture of my home lab - https://x.com/koeng101/status/1844150979484319842

I'm trying to build a DNA assembly company right now (been lots of ups and downs lately...), and one thing I need to do is validate the specs of my oligo pool synthesis provider, Agilent, before I release to customers / raise a seed round. So as a stress-test run of my system, I'm synthesizing a genome, and am thinking about trying to livestream it. The unique technology is variety of ways to assemble and validate DNA from oligo pools for a lot cheaper, pretty much enabling a 10x reduction in DNA synthesis cost vs commercial suppliers. I've worked my ass off for nearly 2 years to get to this moment and am so excited!

                    ╓───╖
                    ║ ! ║
                    ╙─┬─╜   ┌───╖  ╔═══╗
                ┌─────┴─────┤ > ╟──╢ 2 ║
                │           ╘═╤═╝  ╚═══╝
  ╔════╗  ┌───╖ │             │
  ║ −1 ╟──┤ + ╟─┴─┐           │
  ╚════╝  ╘═╤═╝   │           │
          ┌─┴─╖   │    ╔═══╗  │
          │ ! ║   │    ║ 1 ║  │
          ╘═╤═╝   │    ╚═╤═╝  │
            │   ┌─┴─╖  ┌─┴─╖  │
            │   │ × ╟──┤ ? ╟──┘
            │   ╘═╤═╝  ╘═╤═╝
            └─────┘      │

I made this language over 10 years ago, but earlier this year I've been making YouTube videos in which I describe it in excruciating detail: https://youtube.com/playlist?list=PLkG32PHxWoJaetjKUMVRONWLg...

For my next video, I want to show in detail how the interpreter works. For this purpose I'm creating an elaborate animation. You'll notice that the latest video is already several months old; this is because this animation is more work than I bargained for, and I got a little burned out by it. Nevertheless, I persevere and the video will come out whenever I may finish it.

Language specification: https://esolangs.org/wiki/Funciton

Interpreter: https://codeberg.org/Timwi/Funciton

I'm building Pastmaps - striving to eventually be the world's largest online collection of old maps, aerials, and photos all packaged into a public historical research platform that's as easy to use as Google Maps. This has been a labor of love now for about a year, but I still have a huge mountain to climb to realize the full vision. Give it a try and give me your harsh criticisms - that's the greatest gift you could give me!

Even in it's current state, it's being used by geneologists, urban explorers, search & rescue teams, real estate developers, government agencies, etc. The number of exploding use-cases continues to astound me and keeps me motivated to continue.

It allows us to control the algorithm. It’s all LLM translating to YouTube search queries under the hood.

Visually it looks the same.

The suggested videos come from predefined buckets on topics they love.

E.g. 33% fun math, 33% DIY engineering, 33% creative activities.

Video recommendations that have a banned word in the title/desc don't get displayed e.g. MrBeast, anything with Minecraft in it, never gets surfaced.

For anyone interested in using it, send me an email.

jim.jones1@gmail.com

I drive a 21 year old Saab, and in my 2 years of owning it, I have replaced every single bulb in the exterior of the vehicle except a turn signal or two.

I decided to create a mobile service for vehicle lights. It's a simple website that even technologically-disadvantaged people can use. The website is nearly finished and I will likely come back here to write a post on it for how the website works.

Oh the best part, I get texted and emailed for each service order that comes in, and using my service is only $10 more than what it would cost you to go buy a bulb yourself at OReillys, AutoZone, etc.

I programmed everything myself and developed the idea as well. This is my first real-world project/solution I am bringing into this world that has been verified by others, to be a needed service. Pretty excited about it and I love changing bulbs or replacing light housings, it's fun and simple.

**Update: I perform the install.

- The idea: https://carlnewton.github.io/posts/location-based-social-net...

- A build update and plan: https://carlnewton.github.io/posts/building-habitat/

- The repository: https://github.com/carlnewton/habitat

- The project board: https://github.com/users/carlnewton/projects/2

I am a fullstack developer living in Norway. Last year I registered the Norwegian branch of the Architectural Uproar as a not for profit organization. With the support from paying members, I have been able to go on tour to most of the major cities in Norway. We organize large meetings were we discuss architecture and city planning with politicians, architects and property developers on stage.

I am strongly inspired by Create Streets in UK and Strong Towns in the US. I want to improve people’s quality of life, help saving the planet and make Norway beautiful again while doing it.

https://arkitekturopproret.no edit: typos

It is designed to solve the problem of "RPG hero just killed a dragon in front of the town and no one says anything about it." All the NPCs realistically react and talk about the Hero's exploits.

Visitors to the site can vote on what quest the hero undertakes next.

I'm running into the problem what the site isn't much fun. I'm honestly not sure what to do about that!

An only slightly buggy build is at https://www.generativestorytelling.ai/tinyllmtown/index.html

Importantly, I am aiming to have everything (except voice gen) working on a small model that can be ran locally.

There was a bit of noise to the sonar transducer since the stepper motor was so noisy, but I mostly eliminated it by routing the motor wires through liquid-tight flexible electrical conduit, connecting the conduit to ground.

I get custom character flaps printed and die-cut in bulk and then sell them in smaller sets. A full set of flaps for one module has 52 distinct designs (letters, numbers, symbols, etc) and I get them from the manufacturer grouped by design, so they need to be collated to sell as packs of 52 with 1 of each design.

My WIP robot will take a stack of one design and distribute them to a bunch of cubbies, then I'll swap in the next design, and so on, so each cubby ends up with a full set.

It's based on a cheap ~$110 CNC gantry frame from AliExpress and a ~$35 BTT SKR Pico 3d printer main board running GrblHAL. To detect whether the flaps feed successfully I use a visible light break-beam sensor (the typical IR sensors don't work because the PVC flaps happen to be IR transparent!) which acts as the "z probe" - the flap is fed via a G38.3 probe action which returns whether the probe was successful or not, and the "z" coordinate it was first detected.

I have a python script running on a computer to send the gcode to the machine.

[0] https://bsky.app/profile/scottbez1.bsky.social/post/3l737hme...

[1] https://github.com/scottbez1/splitflap

Im developing it with the use in mind of flying through your code to show others relationships, or edit with a visuospatial look at your code instead of basic 2D tabs and a mind map of which one had the thing you're working on. It's kinda fun to work on the project In the project!

It's built on Swift and Metal but can ready any utf8 text file, minus a few subsections of the Unicode spec (for now).

https://github.com/tikimcfee/LookAtThat

The result is a much more interactive way to present code than screencasts or blogs. Because at any point we can pause a session and freely explore and experiment with the codebase.

I put together a demo recently [1] and written much more about it here [2].

[1] https://youtu.be/Qp2GdLO5eSc

[2] https://codemic.io

I am writing a tool to collect and aggregate data about the processing times. This will help people plan around the delays. Knowing is half the battle.

The biggest challenge is that my readers find me at the start of the process, and I need their feedback at the end of it. I have to make it easy for them to provide partial feedback and complete it later after they get an email reminder.

This would be unnecessary if the immigration office collected and shared that information, but they don’t. They also don’t welcome any help because they “operate at peak efficiency”. I have stopped hoping for their collaboration.

Just doing plain OCR doesn't really work because the notes in the margin and the footnotes get mingled with the text, which results in gibberish.

But, when sent to Google Vision API, each page results in a json file that has an object for each word and the four coordinates of its bounding box.

That json file is pretty big (around 1.5 Mo when pretty printed, or 500 k with no indents or line breaks) but it can then be fed to Gemini, taking advantage of its large context window.

Gemini is pretty good at identifying each section of the page (headers, main text, margin comments, footnotes) but it takes a looong time to respond (2-5 minutes per page).

So another approach is to ask Gemini to write a python script to analyze the json result and group sections depending of the coordinates of each word, and then run that script against the json output by the OCR phase.

But it's quite difficult to have a script that works for any page; comments in the margin are always in the margin so that's pretty easy, but footnotes can start at any height of the page (some pages contain only footnotes running from previous pages) and Gemini likes to be pretty specific, giving hard 'y' coordinates for where footnotes should start, which obviously only works for the one page it's working on.

I'm iterating and making some progress but I feel like I miss a big breakthrough and it all should be simpler than it currently is. Information about OCR is pretty scarce online. Any pointer is welcome!

Macros in lisp are just normal functions that receive the code they wrap as argument and return some modified code. Typically they will just wrap the passed code into some more code. But then there are code walking macro: macros that will traverse passed code to modify it in depth.

What I'm working on is "code diving" macros. Not only will they traverse the passed code, but they will resolve called functions and macros, fetch their source code and traverse it too. And so on. All the redefined fns/macros are accumulated in a let/macrolet binding, topologically sorted by call/dependency order. Instrumented code will call these local redefinitions, shadowing the global definition lexically.

This allow the programmer to write truly local monkey-patches for existing code he doesn't have control over (e.g, code from another library for instance). I'm writing this in Clojure, and the traditional way to do this is to temporarily change the global definitions of targeted variables using with-redefs. This is problematic because other threads will see these redefinitions, and not just threads created within the instrumented code, but already existing threads too.

Another way to do it is to just redefine the targeted functions globally, but then your modifications are available to the whole program for the rest of its execution.

It's a very fun mix of hardware (for data collection), and crazy SQL queries to model energy flows between buildings, solar, batteries, etc. Considering just one building is pretty easy:

consumption = imported - exported + generated - stored + dispatched. carbon = carbon intensity * imported cost = tariff * imported

but then you add a site with a couple of buildings, solar on one of them, grid limited exports, etc modelling these flows is challenging. Like consider the case where one building got 10% of it's imported power from another building's excess solar, then calculating carbon becomes more difficult.

and once you've figured all that - then you have to figure out what makes commercial sense to do next.. install a battery, expand solar, move onto a TOU tariff, do nothing - and that's a whole other world of optimisation problems.

Started of with [1] which showed that there might be some strength to the idea.

Applied it for chunking [2] and web site analysis[3] and got pretty good results.

Just started trying out experiments on video [4] and surprised that the structure seems to hold for image embeddings as well.

I have no clue if this has any value, but it is fun to go down this rabbit hole :)

[1]https://gpt3experiments.substack.com/p/the-shape-of-stories-...

[2]https://gpt3experiments.substack.com/p/a-new-chunking-approa...

[3]https://gpt3experiments.substack.com/p/using-semantic-chunki...

[4]https://x.com/nutanc/status/1849661242027409723

Trick-or-treating at a door is so last decade; trying to catch a Snickers hurtling towards you in the darkness is the future.

Got sick of working in Google Docs and having to manually move days around and re-label dates, shift hotels, etc. Ended up creating Turas.app over a weekend in 2023 (and then let it loose on Reddit). But just recently created the Chrome Extension which feels like it is an even better tool because it lets you access all of the richness of Google Maps. It's a Google Maps powertool for people who like to plan their travel meticulously.

(Completely free and intended to be free forever; we tried to monetize it but realized that there's no reasonable way to do so that we ourselves would be happy with; seriously thinking about just open sourcing it, but needs some cleanup first!)

Extension: https://chromewebstore.google.com/detail/turasapp/lpfijfdbgo...

App: https://turas.app

The game was playable on the first try with only a few minor and very quick rule tweaks from me. Even I felt we went on a stressful quest in the forest. Honestly among the best games for this age group that I have ever played.

10/10 activity, will try again.

(Small note: I'm doing it in full-stack Rust, including web frontend, using leptos).

The work most of us do isn't tangible. You have nothing to "prove" that you made something. Creating something in meat space is really rewarding.

It's been busy:

- Reading books, currently "The Mom Test"

- Looking for a startup community and being disappointed with what my city has to offer

- Deciding between VC vs. bootstrapping, and taking a firmer stance to try the latter first. This includes rejecting ODF.

- Talking with like-minded people. I now have a better understanding of what the MVP should look like.

Receiving so much support has been very encouraging; I announced it more publicly at https://nullderef.com/blog/quit-job-2024/

It's scary. But working has never been so fun!

As a father, I wanted to capture all the little moments of our day-to-day family life to later share with my grown-up children. However, I did not have the discipline to journal regularly. So, I made Memzy to capture them easily on the fly!

https://apps.apple.com/us/app/memzy-easy-journaling-with-ai/...

https://ladsko.github.io/hexed/

It's hexagonal, compact and optimized for German and English thanks to noted (https://dariogoetz.github.io/noted-layout/)

Right now I'm testing key shapes and sizes and the manufacturing of key caps. I thought I needed a resin printer for best results but FDM printed caps with a blob of epoxy resin on top works surprisingly well.

There's no firmware yet, I have not even decided on a micro-controller or software. But I want to use Kailh Choc v1.

This is my first time exposing this project to the public so I'll be very happy about some feedback!

https://github.com/kviklet/kviklet

Essentially a PR review flow for production access, which allows you to enforce a second pair of eyes workflow. I was always a bit scared when I was on call and had all the power in my finger tips to ruin everyone's day. I think this helps alleviate the risk of human error significantly. Also helps with compliance of course.

Ride Sustainably with the World's First Repairable Battery

Refillable in 5 minutes (just buy $150 worth of new cells every 3 years or so, when they're depleted)

Be Worry-Free thanks to the Fireproof Casing! There's been waaaaaay too many lithium fires!

It's launched as an IndieGogo (the product already exists, but as a startup IndieGogo is convenient to get the cash upfront to buy the parts and build the batteries) and there is an offer for early-backers here https://get.gouach.com for a 25% discount on the battery!

Breathing Tips - a free web app to practice breathing exercises.

Tech stack: Ruby on Rails 7.2, sqlite3, BabylonJS for the 3D visuals, hosting on hetzner for 4$ and deployed with kamal 2

https://breathing.tips

If you're in the Atlanta, GA area and need bicycle work done, hit me up! No job is too big or too small. I particularly enjoy building wheels if you want a sweet custom wheel set, but I do it all (including Mountain Bike fork and suspension work that many shops won't do).

https://www.atlbikeshed.com/

I'm also documenting every step of the process and uploading it to YouTube, which means I am also teaching myself how to edit videos :)

If anyone wants to check it out: https://www.youtube.com/playlist?list=PL55VZ7oDEoRnXKSYTiGlN...

I’ve ran a start up and saw how SaaS-ified and expensive web tooling is (heroku, datadog, redshift, fivetran, etc), but how difficult it was to move off of them. We had a few years of over 1 million in infrastructure spend.

I’m hoping just making Kubernetes easier to use gives us a way out.

It’s fully open source and a hosted version is free to use! https://canine.sh.

Would love feedback on it, including how the overall pitch could be better, or if it actually solves people’s problems.

[1] https://www.researchgate.net/publication/381009719_Hydra_Enh...

Download: https://akkartik.name/carousel-cards.love (really just a zip file containing source code, 169KB)

Installation instructions: same as https://akkartik.itch.io/carousel

Repo for the core app: https://git.sr.ht/~akkartik/carousel.love

So far I have 111 little "levels", each good for between a few seconds and a minute. I think a full curriculum/game will need maybe 2000 levels?

If you are in an EPA-regulated (or equivalents in Canada and Europe) industry (such as mining, oil and gas, minerals, rare earths, metal processing, airlines, construction, shipping/marine, logistics, heavy industries, agriculture/farming/food production, data center, power generation including renewables, adjacent ones like consumer goods, real estate, large scale AI training, climate derivatives etc.) or require sustainability consulting support in general, we would love to talk to you: hello@carbonimpacthq.com (put “HN:” in the subject line so we know where you are coming from)

https://carbonimpacthq.com (the landing page is still a work-in-progress but you can check out our blog for more information https://carbonimpacthq.com/esg-and-your-business/)

Since my last update here, I've added more detailed personalized descriptions of recommendations (hit Describe to request), including a rating out of 10 for how well the item meets your preferences.

I've also added the ability to replace individual recommendations (this was the most requested new feature!). If you update your preferences, your replacements will use your updated preferences - pretty nice for fine tuning your results!

Try it: https://www.yogurrt.com/

What's holding me back from scaling is primarily my own resistance to marketing, plus some pending improvements to the ChatGPT-based transcript editing system. Once I finish optimizing the LLM integration, I'll have no more excuses to avoid sales outreach. One thread I want to pursue is a magazine of podcast transcripts that I inherited: https://podread.org

I'd appreciate advice on authentic outreach strategies for reaching knowledge-focused podcasters. If anyone here has experience in this space or wants to collaborate, I'm open to connecting.

[^0]: https://www.withfanfare.com/

[^1]: https://transcripts.volts.wtf/

Learning how to arrange things, navigation, and my own blog on my own site gives me the gratification of owning something fully. Everyone should have their own site is truly what I agree with.

https://spacesanjeet.me