Ask HN: Tell us about your project that's not done yet but you want feedback on
A lot of times with side projects I wished I had gotten feedback early on, before I spent a lot of time on an inefficient direction. I wonder if people wait too long to publish something before it is fully polished, then realized that the polishing wasn't needed.
I'm interested to see things that people would have never published otherwise. I know a lot of my projects never make it to a published phase, but I still would have been interested in knowing the general reception. Please drop your projects here!
471 comments
[ 5.7 ms ] story [ 331 ms ] threadLink: https://carvemeetings.com/
What features would you add that set Carve apart?
[1] https://workspace.google.com/resources/appointment-schedulin...
The additional features would let users set filters on:
* What times they wanted to schedule meetings
* Schedule meetings at least 24 hours in advance
* Set some people as optional
* etc..
https://support.google.com/chat/answer/7655908?hl=en
Find 45 minutes with Cynthia next week, title it "1:1”
These two flows are the ones I'm interested in building out. Something low friction where you don't have to spend time trying to find a common free slot on each other's calendar
https://github.com/hofstadter-io/cuelm (working name)
Our hypothesis is that a unified config space across our industry will bring many benefits.
Our open questions are largely around whether we
(1) replace / alternative
(2) just a skin, still same reconciliation loop from the tools
(3) replace, but support existing modules & charts
Very open to feedback and ideas. Contact in my bio. Thanks HN!
https://www.hazysurf.com/
People that run equipment rental stores probably have a facebook group or professional association. If you can befriend someone influential in one of those, it might be a good place to get started.
Though to be honest I have no idea what kind of info is available from Strava, but I feel that the more you can do on your end and the less involvement required by users, the further you could get with a greater number of people. Relying on people coming to you to sign up for yet another social network seems like a lot to ask of people.
Lots of things that could be done with this but a lot of robotics teams and projects got laid in the past year and it doesn't have as many users as it once did. There is also Foxglove which is a full-blown funded startup working on something similar and the more progress they make, the less motivation I have to make my FOSS project better, as a one-person team.
* Air quality monitor -- https://dheera.net/projects/airmonitor/
Unfinished due to the PM2.5 sensor in it (PMSA003I) always reporting 0 and Adafruit not caring.
* BotParty -- telepresence robots -- https://dheera.net/projects/botparty/
Unfinished since COVID isolation ended and telepresence parties aren't needed anymore
* Luxo -- jumping Pixar lamp -- https://dheera.net/projects/luxo/
Unfinished due to lack of time but it does scoot and jump.
* Digital 4x5 camera back -- https://dheera.net/projects/4x5/
Unfinished due to issues with CRA optimization on the microlens array on the Pi HQ camera and unavailability of a real HQ camera with an actually large (e.g. 35mm) sensor.
* E-ink picture frame -- https://dheera.net/projects/einkframe/
Pending to be done: (a) Get rid of the unsightly USB power wire, use a battery or super-capacitor and a thin strip of solar panel on top of the frame, and ESP32 deep sleep mode and have it update once every hour or so. (b) Update it to use a diffusion model in the cloud for image generation (c) Add a microphone and a mode where you can yell a prompt at it and it draws it
You could get a calibration frame by removing the lens with the sensor centered, and a black frame by putting a lens cap on, to help with corrections.
That looks like a really fun project. With the HDR mode of the sensor you should get some awesome images. Good luck!
[1] https://www.sony-semicon.com/files/62/pdf/p-13_IMX477-AACK_F... - (PDF) Flyer
[2] https://www.uctronics.com/download/Image_Sensor/IMX477-DS.pd... - (PDF) Data Sheet
> I think the plastic lens mount might be obstructing the light going to the sensor
It's actually a low-profile, 3D printed thing housing a glass UV/IR cut filter. The housing is flush with the glass filter unlike the stock Pi HQ camera lens holder. Removing the UV/IR cut results in full-spectrum photos that manifest the same color shift problem which is characteristic of CRA optimization (where they deliberately put the microlenses off-center toward the edges of the sensor.
> As far as I can tell, there's no lens array on the sensor[1][2]
I thought there is a lens array on almost all Sony sensors? [2] for example lists a "CRA characteristics of recommended lens" section on page 47, but this is not the full datasheet and only goes upto page 25. I can't find the full datasheet on the public internet unfortunately.
[1] https://maxmax.com/maincamerapage/uvcameras/raspberry-pi-hq-...
If you've ever seen a ball cutter attachment for a lathe, or one of those parallax eliminating camera mounts, you could step the sensor along a sphere instead of a plane.
(No tutorial yet, click on the castle and move around with WASD. E to undo a move.)
Little multiplayer web game made with Elixir and Svelte. Based on the game generals.io.
Not mobile friendly atm.
https://www.mirroredfables.com/?proxyKey=hellohackernews
source: https://github.com/mirroredfables/mirroredfables
``` [worker] Searching for music... ```
If it helps, my prompt was:
> I would like a story set 2000 years in the future, where humanity has achieved interstellar travel. Humans prioritize exploration of extra-solar planets and the search for intelligent life
It was based on the prompt I shared previously.
The protagonist was Aeyela who was rendered as an anime/manga school girl. Her voice was portrayed as deep and masculine, which was pretty entertaining.
I did encounter a couple issues:
* the 2nd song was age-restricted, and I was unable to skip it, so the music stopped. I looked at another tab and returned, and another song loaded
* potentially related to the above: the voice narration stopped and I couldn't figure out how to get it to resume, even after toggling a few of the related settings
I think the problem is it generates content, but not very high quality content. It's not gripping enough from the beginning, just sort of drags out. It could be that the engine works right, just that the input is garbage, as often happens with AI.
But maybe something about it could be engineered a little better. From the code it looks like it's just generating most of it in a single blast. But I wonder if it can be iterated into chunks.
e.g. split it into the Three Act Structure.
1. Protagonist really wants something. Faces an obstacle. Tries everything and gets stuck. 2. Protagonist makes sacrifices. Gets even more stuck. Something that the protagonist faces in Act 1 ends up being the key, often a mentor or a sacrifice that Protagonist didn't expect. 3. Denouement.
You could probably generate a want and an obstacle. I kind of did a prototype of that here (feel free to steal ideas): https://random-character-generator.com/
Then do the same with the other characters. Generate relationships.
Then some iteration where the AI writes those characters into those scripts.
Probably another where it writes up scenes/places/setting later. The characters are then colored in and put into these places.
An editing run where it removes unnecessary scenes or scenes that didn't work. Or leave this to the humans.
Anyway, it is very good work. It's nice that it's open source, may contribute to it some day.
You have got a nice generator there! I am very curious what's powering it, as it seems to be instant. I dropped you an email, let's chat!
Re: structure and cohesion- this is where GPT4 vastly outperforms GPT3. For example, it is much better at using a hero's journey and managing plot twists. You can go into the AI settings window and set the model to use GPT4 instead [I'll leave how to get there as an exercise ;)], and give it a try.
Currently the generator works in a very similar way to how you described it: 1. It takes your input and generate a world setting and the characters in it, including relationships 2. Based on the information from above, it generates a overall plot outline, broken down into chapters 3. Based on the information from above, it then fills out the chapters with dialog
Currently the biggest limitation is context length and cost, and I haven't focused on optimizing that yet.
You are absolutely right that a human editor would vastly improve the quality. You can see a lightly guided early pre-generated example here [https://www.mirroredfables.com/?gameUuid=302fe7bf-4ff0-4a77-...]
An AI story editor is a very interesting idea that I haven't thought of before, and it makes a lot of sense. I am going to play around with it!
Stripe's mobile app and others kinda do this already but some of their numbers can be inaccurate (as detailed in the repo's readme) so that made me open-source + solve an issue my own SaaS has with Stripe:
https://github.com/hbcondo/revenut-app
As mentioned in the challenge section, getting data from Stripe is slow so I've been reluctant to put this app in front of others even though it is fully functional.
Eventually I'd like to add auth and a backend, it's using localstorage for now.
Myself and a friend have been working on this for quite a while. It's frustrating because it has come so far (showing trail data worldwide interactively on a tiny budget is challenging), and yet it is so far from being something super useful like AllTrails due to low data quality and a lack of relevant features (photos, reviews, etc etc.)
Catnip is an RSS web app with two main differentiators: - it's an RSS 'aggregator', not a reader mode. I believe the browser itself is the optimal reader. Every RSS app has a bespoke reader mode that tends to suck in comparison to the browser itself. - it has a frontpage that's an aggregate of some popular feeds, making it useful without an account. I plan on adding some boosting of popular feeds once I have users.
https://rooms.substack.com/
Would love any general feedback or ideas!
Could monetize this eventually fairly easily with a premium version with expanded suite of features and improve the UX.
If anyone wants to help me take on some of the more technical aspects.
--> I need a clean HTML/CSS to PDF export functionality, but it's proving challenging.
Any interest or feedback?
- working on cleaning up the source for open sourcing
- the first task is failing right now due to a bug
- doc upload is bugged
- tools are disabled
- tons of loose ends in the UI
- Actively developed: database is likely to be wiped frequently until launch, along with any user accounts.
And a skeleton application that uses the bundle. https://github.com/SeleneSoftware/SeleneCMS
Issues and Pull Requests welcome.
It's a simple (WIP) ISA and corresponding simulator. I want to get experience adding a new backend target for LLVM, and this is will be the target.
The idea is to be similar to other LLVM targets, but different enough that I can't just copy-and-paste existing LLVM code.
I am interested in knowing what are the common problems you face with command line approach that can be solved via a Web UI
[1]: https://github.com/ananthakumaran/paisa
[2]: https://paisa-demo.ananthakumaran.in/
I haven't worked out a coherent UI for this quite yet, but ideally you'd be able to give yourself a notification that a branch you were working on hasn't seen any updates in some time, or that you did commit but haven't pushed, or haven't changed the status of the Jira ticket back into review.
If anyone has these problems, I'd love to hear if anything like this would be useful.
Been working on this as a new way to find, explore, and view old historical maps and aerials of my area. Still heavily in development but have been surprised by some early traffic stats (1-2K organic uniques / mo and growing 200%+ m/m right now)
Hoping to add in more advanced map tooling within the next week or 2, including new basemap options, 3d terrain view, and then a proper search box which I've been pushing off for far too long
In that same vein, you could add a feature where users could contribute georeferenced map files for community review and approval--I think that would really increase your scalability. I see you have an email for that currently.
I have found the ui for map difficult to use in the past.
Really love this suggestion, I could see it making map exploration far easier since now you just need to explore by year instead of by map and by year (reducing dimensionality by 1!). Thank you, I'm going to chew on this a bit more but I really think there's something here
I waited too long to publish this but it is still nowhere near done and I just had to publish early.
Missing most features I want to add but hey, it’s live now!
That's double meta in this thread, I know.
I don't have an MVP yet but I've been testing and tweaking prompts. An early strategy might be to release this as a free open source app and then offer extra business value later. I created an email form in case there's any interest here.
https://tally.so/r/mK5J6K
- You don't submit your ideas through us; they stay between you and your LLM.
- It will focus on tech founders.
- It will be open source and free with additional paid features for the 1% who may need more.
- It will show you results without asking for more information.
- It will have a much more narrow focus; at front at least.
I don’t want to entirely replace the data gathering + annotation workflow for everyone, but I believe that ready-to-go data can reduce the barrier of entry for ML projects, make _some_ people realize that they don’t really need their own data collection, and even help large businesses reduce their dependency on Scale / Appen.
I am working on a dedicated GPT AI Dungeon Master. Results have been promising so far. Though it has a ways to go before it is even remotely playable outside of "yeah, it'll break if you do that, let me load up the jupyter notebook that can handle that situation."
https://github.com/justinlloyd/retro-chores
A physical chores list with a digital component. Unfortunately it has languished for a year due to circumstances of life (family health issues). Now that things are settling down I can return my attention to it.
And I found a functional 4K 55" TV on the street last night that was being disposed of, so I'm trying to decide what to do with that (apart from watching TV on it). I am thinking another cat toy, since the cats trashed their last one.
I also have plans to add a feature to save and share your projects, but not implemented yet and I'm always curious what other features or musical capabilities people would enjoy playing with.
I'm also working on cli time tracking tool that is still very basic but I'm sort of wondering how best to add more functionality. The current main branch allows you to start a task with `$ start -t` and you can add details and then list all the events you've tracked, all saved to sqlite database. I like how immediate it is to track things from the command line, but I'm still thinking about how to display the events and tasks you've tracked (I'm building with rust and started tinkering with Ratatui to build a TUI, though I prefer just having a cli most of the time). I'm also still thinking about cli UX. I kind of prefer the git hub cli style that has commands and sub-commands because it seems more orderly than just having long lists of arguments. Anyway it's very in the beginning stages but would love recommendations on reading for UX for command line interfaces. https://github.com/farmeroy/Time-Bandit
You might want to look into adding a DynamicsCompressorNode near the end of your web audio stack; it should help with the clipping issues you're getting whenever there's polyphony going on.