Cooperation Cube (https://cooperationcube.com/) — A strategic 4-player memory/semi-cooperative board game I designed, played on a rotating 3D cube. Just added a daily puzzle (https://cooperationcube.com/daily) you can play without signing up. Place sticks, complete patterns, and try to beat the day's challenge.
Live Kaiwa (https://livekaiwa.com/) — A real-time Japanese conversation assistant. It listens, transcribes, translates, and suggests responses so you can follow along in conversations you'd otherwise get lost in. I built it because I live in Japan and needed something for the situations where missing a nuance actually matters — PTA meetings, bank appointments, neighborhood councils.
The landing page looks nice, but it’s very technical. It might be helpful to include some real-world use cases that show what people could do with the tool (at the 30,000 ft level).
1. Live Kaiwa — real-time Japanese conversation support
I live in a rural farming neighborhood in Japan. Day-to-day Japanese is fine for me, but neighborhood meetings were a completely different level. Fast speech, local dialect, references to people and events from decades ago. I'd leave feeling like I understood maybe 5% of what happened.
So I built a tool for myself to help follow those conversations.
Live Kaiwa transcribes Japanese speech in real time and gives English translations, summaries, and suggested responses while the conversation is happening.
Some technical details:
* Browser microphone streams audio via WebRTC to a server with Kotoba Whisper
* Multi-pass transcription: quick first pass, then higher-accuracy re-transcription that replaces earlier text
* Each batch of transcript is sent to an LLM that generates translations, summary bullets, and response suggestions
* Everything is streamed back to the UI live
* Session data stays entirely in the browser — nothing stored server-side
2. Cooperation Cube — a board game that rotates the playing field
Years ago I built a physical board game where players place sticks into a wooden cube to complete patterns on the faces.
The twist: the cube rotates 90° every round, so patterns you're building suddenly become part of someone else's board. It creates a mix of strategy, memory, and semi-cooperative play.
I recently built a digital version.
Game mechanics:
* 4 players drafting cards and placing colored sticks on cube faces
* The cube rotates every 4 actions
* Players must remember what exists on other faces
* Cooperation cards allow two players to coordinate for shared bonuses
* Game ends when someone runs out of short sticks
I found the apk file and I played around with Android Studio today but was having trouble getting the Nexus device with KittyKat or Lollipop to not crash when trying to open the device. Will keep trying different options.
Can you explain more what you mean by “do whatever you want without as much effort”? Is it because text-davinci-003 accepts more tokens for the prompt? Something else?
Ambiki is a web application for pediatric speech, occupational, and physical therapists which will be launching later this month. The product has been incubated and built alongside therapists for 3 years at a private practice in Tennessee. Ambiki touches on almost everything including teletherapy, billing, scheduling, tracking patient outcomes, as well as providing resources/tools for therapists to use in their treatment sessions.
Send your CV (PDF) and a short cover letter (PDF) to [email protected]. In your cover letter please answer: What interests you about working on Ambiki?
Nice work, this looks nice. I signed up and was writing a post. I was nearly finished and switched to the tutorial post (which I had archived). I clicked the hamburger and clicked "permanently delete" to remove the tutorial post. However, nothing happened. I then clicked on "All posts" to go back to the post I was working on and it had deleted that. Any way to recover that? I don't see a help email.
We give schools and patients access to the therapies and resources they need to improve their communication and overall quality of life. Our experienced team is made up of 100+ specialized therapists in Speech, Occupational and Physical Therapy. Our development team is currently 3 full-time employees.
See my comment below for some of the reasons I've had issues trying to test the commonly used segmentation corpora. I completely agree it would be great if there was a free (as in both speech and beer) common training set. One key would be that this common training set either provide the exact text that should be run in the segmenter or exact instructions on how to produce the text to run in the segmenter (re: see the issue I mentioned below of the ambiguity around how to actually test the Brown corpus).
Good points. I'd love to test it on some of the typically used corpora. The issues I have are:
1) Most segmentation research papers are done by Universities which have access to the Penn Treebank data (WSJ and Brown corpus). However, the cost of that data is $1,700 https://catalog.ldc.upenn.edu/LDC99T42
2) The Brown corpus is available for free in NLTK (http://www.nltk.org/nltk_data/). However it is the tagged corpus. I've contacted the researchers for all of the top segmentation libraries but never received an answer to any of the following questions:
a) I’m assuming you preprocessed the text by removing the tags. Is this correct? Or did you use the untagged version, and if so do you have a link to that as I only found the tagged version in the NLTK data?
b) When removing the tags did you also remove each carriage return and newline so the text was one long string, each sentence separated by just one whitespace?
c) The download contains 100+ files. Did you analyze each individually? Or did you create one combined file? If you created a combined file how did you space each individual file within the larger file? Also, if you combined them what order did you combine them in?
So sure, all of these papers use the same data, but we have no idea if they are actually using that data in the same way, as none of the papers actually release their code and tests, or tell the steps they used to preprocess the corpus.
To test more broad coverage on my library I added the full text of Alice in Wonderland https://github.com/diasks2/pragmatic_segmenter/blob/master/s.... A grad student from Stanford offered to test my library on the WSJ corpus a few months ago which was very kind, but I'm still waiting to hear back on that.
I did an analysis of different sentence segmentation tools when I was working on my own rule-based segmenter. The results can be found in this README (https://github.com/diasks2/pragmatic_segmenter).
I think this blog post almost hits on the key in the middle - in my opinion it is important to test (all of) the edge cases. The problem with most corpora typically used to test segmenters is that 80-90% of the sentences are the same (i.e. a regular sentence ending in a period). Thus if a segmenter just simply split the sentence at every period it would still show a 80-90% accuracy rate. This is why I am trying to develop a standardized set of edge cases: https://github.com/diasks2/pragmatic_segmenter#the-golden-ru...
I've found the best way to make upgrading easier is having good test coverage. I think one of the biggest fears when upgrading is having something break but not knowing that it is now broken. Although tests won't always catch everything they are a great start to ease the upgrade process.
Michael, great work as always. Your tutorials helped me to get up to speed on Rails quickly. Just curious, is there any particular reason you switched from GitHub to Bitbucket in your tutorial?
I think this shows one reason it can benefit young people to join a company in the industry they are targeting before jumping into startup life. One of the best/easiest ways to find pain points is to be knee deep in the industry. Obviously in this specific case I am not recommending going to prison, but hopefully you get the point.
Great job. Nice execution. I would recommend adding some type of simple tutorial walk-through after the new page is first created. It took a little while for it to sink in that this is a blank canvas and I can start typing, etc. Maybe the fact that the canvas is the same color as the background of the site, it is not easy that this is a "word doc" where I can start building. I know you probably want to keep it minimalistic, but it might help to outline the canvas so this fact is more obvious.
I'm not quite sure I understand the purpose of the app. I would have to actually check the app during an earthquake to have it tell me yes? Well, in that case I would already know that there is an earthquake near me.
Maybe a better title would be "Was there an earthquake near me?" and you could show nearby earthquakes in the past hour, day, etc. However, there are already a lot of sites that do something like that such as http://earthquake.usgs.gov/earthquakes/index.php, so I'm not quite sure what additional value you could add.
Live Kaiwa (https://livekaiwa.com/) — A real-time Japanese conversation assistant. It listens, transcribes, translates, and suggests responses so you can follow along in conversations you'd otherwise get lost in. I built it because I live in Japan and needed something for the situations where missing a nuance actually matters — PTA meetings, bank appointments, neighborhood councils.