Not only does it give you a super easy-to-grok understanding of the model quality just by looking at the image, but when you compare tokens and costs (both input and output), you really get a good, simple COST x QUALITY evaluation across models.
I built an iOS app for my sons who play club basketball. It’s designed to bridge the gap between practicing solo on the court and the pressure of playing in a game.
It uses mechanics and body movement, via visual detection through the cameras, to identify when you’re about to take a shot or make a pass.
It then pumps distracting noises (honks, cheers, jeers, etc.) into the headphones to try to distract you. It also runs a continuous 24-second clock so you build a natural sense of how much time is left before you need to get a shot off.
Building Basil (https://basilai.app), a privacy-first AI executive assistant / note taker. Everything is on-device: transcription, summarization, etc. So nothing is shared with any model provider.
Lately the interesting work has been less about raw transcription and more about making the output actually useful: recurring workflows, follow-ups, and personalized summaries that fit how someone runs their day. We are also pushing more toward reliable agent automation instead of one-off chat. And dealing with all the iOS limitations on background work.
Building a voice recorder/transcription/summarization app where everything runs on-device. No cloud, no uploads, no servers processing your audio.
Started with a web version using WebAssembly Whisper to validate the idea. Worked well enough to prove demand, but browser sandbox limits (no background recording, inconsistent model performance) pushed me to go native.
Rebuilt as an iOS/macOS app with CoreML. Apple Silicon handles Whisper-class models locally without issues, but its confined to recent devices.
I'm working on e2ee sync between web and mobile. I want to build a simplenotes but for audio transcription.
My experience share only supporting the latest OS:
I have launched apps focused on a new feature in the latest OS and regretted it. The # of people who have the latest OS is much smaller than the full install base for much longer than I thought. As a result, my marketing conversion was unnaturally low - people who liked the app idea but couldn't install because they had the wrong OS. This causes two problems: potential users I activated but couldn't convert and this signal gets internalized by the App Store, pushing down future impressions.
Now I always have a fallback implementation of the feature so I can target the prior OS. Both Mac and iOS.
As a side project, I just launched a privacy-first web-based meeting transcriber (https://basilai.app/app). Everything runs entirely in your browser — both the transcription and AI summarization — so no audio or text ever leaves your device.
I'm using the browser built in transcription service plus downloading a model and running it via webgpu. No login. At the end of your meeting, you get a zip file with the audio, transcript and summary.
I'm a big fan of digitpain (Jeffrey Scudder) and it is great to see him get airtime here. The tl/dr for Jeffrey is he make digital art, but first he makes his own digital tools to make that artwork.
For those that are new to him, be sure to browse to the top level (https://aesthetic.computer) and play around in the 'terminal'.
If you have a VR device, view Freaky Flowers in it.
The feature I like the most in your demo: the one page checkout. It comes very close to shopify's excellent check out. Nicely done!
WooCommerce should be filling the role of viable self host alternative to Shopify, but it falls short. Even with the plugin ecosystem, it is much harder to run an quality cart and checkout.
I am on a similar path with https://intro.wtf (AI product merchandising for online retailers). I am solo but am trying to make being a solo saas founder the 'water I swim in.' I am connecting with others in my network who are solo founders, joined microconf, listening to podcasts, etc. I am trying to connect with other people who are all focused on similar goals, combine that with my focus and work to get better over time.
I'm working on building better product merchandising with AI/ML for online retailers.
The problem: creating unique, engaging product content is expensive and time consuming for retailers. Hiring content writers can cost $50-100 per product. Multiply that by 1000s of products and then add in constant product churn.
With uniquely tuned AI models, combined with a retailers raw product data, I can generate content that improves the buyer's experience and leads to:
- higher conversions
- improved SEO
- standing out from competitors
The app includes:
- the unique LL models
- pipelines to connect to the retailer's catalog
- a feedback mechanism to monitor the products (add to cart, conversions, on page time, etc) and test new content
I did pilot with a large wine retailer. We generated nearly 4,000 product descriptions and saw a near immediate increase in on page time and conversions. It is 90 days out we are now seeing an SEO gain too.
This is really interesting. I use git to manage my deploys. I git pull from my servers whenever I deploy.
I have been toying with using github actions to automate the pull. My workflow would be something like:
1. work locally on a branch
2. test, merge with master
3. git push from local to github
4. a github action would see the commit to master and initiate a git pull on my server.
We should all get together and build one to rule them all.
My thoughts on this: The real question is whether tracking trends on reddit/twitter can reveal a coming price swing BEFORE it happens. AND, you have to have minimal/no false positives.
I have had some success generating alerts before the swings occur (https://cutt.co/track-record/) but still am working on sorting wheat from chaff.
I am tackling a similar problem at [the Cutt Newsletter](https://cutt.co).
I have taken a slightly different tactic than you. I am trying to surface actionable trade ideas from the social media mentions. Can I predict a stocks price trend from social media discussion?
I think that takes two forms:
1. Real-time alerts to stocks that are moving outside their normal 'mention' bands. Most times social media discussion of a stock is following public news. For example, a company releases earnings and the social media discussion is about those earnings. In that case, the 'mentions' rate will always trail the actual news. But, in some cases, an increase in mention rate is preceding price changes that day -- I am detecting that in real-time (sometimes, so far).
2. Over a long time (months, years) a company's social media mention rate will go up or down. A long-term trend (either positive or negative) could say something about the long-term financial prospects of the company. That could be tradable.
Thank you -- pyarmor is what I am looking for. It seems to provide for both code obfuscation AND licensing. Also, if you use the licensing, the code obfuscation is linked to the license, so you can easily track back code leaks.
I have just started playing with repl.it and am enjoying it. repl.it is using a similar IDE to the VSCode fork from the OP.
I am trying to build a small web app using repl.it as my dev environment and deploy on google app engine. The stack is flask/mysql/html5/bootstrap.
I have a workflow where I can commit to a dev branch inside repl.it, push to github and them merge to master. If merge is no-conflict, it uses github actions to deploy to google app engine.
There are trade offs, but I am really enjoying the ability to just start coding by opening a browser and having my full dev environment in a tab. Really fun!
Congrats on launching -- really impressive. One important issue that these self hosted analytics solve is ad blocking. Ad blocking by users really undermines the ability of a site or app to figure out what is working and not working. When you host your own analytics, you can get usability information for all of your users, not just those that don't block. That allows you to make a better product.
I have been working on something similar at https://argyle.cc -- we combine cloud analytics with a self-hosted analytics collector js. That gives you the best of both worlds: privacy focused, user respecting analytics, but full featured reporting in the cloud and ad-blocker resistance. It also allows event tracking to be done over js/web or in-line/server side.