Also: kids can code using cardboard cut-out blocks and just take a photo to import, generate Arduino firmware, compile and flash to ESP32 - all from a single docker web application.

Hoping to bootstrap a cheaper, open source alternative to current systems (one of which my son is using at his school). I have had success doing this with another project on Patreon so that's where it is going to be launched. https://www.patreon.com/ChonkyBlox - so far just a couple of demo video's but code will be released in full once I have added more blocks and a tutorial.

Ultimately this is aimed at the South African education sector, where many schools may have only one computer available for a classroom to share, hence the cardboard cut-outs. There is also a hardware component still in development - a simple pcb breadboard break-out with connectors for components instead of kids having to use jumper cables on the breadboared. We also have a working Desktop application which does code assistance from within the Arduino IDE - using the same web server with api, so literally everything from very small kids to advanced is covered.

DoodleMate is a multimodal storytelling canvas that exposes students to the basics of animation, character design, and story creation. Starting from paper drawings, it creates rigged versions of your characters that can talk, walk, dancing, sing, etc and you can create scenes out of it.

It doesn't use any generative AI whatsoever. Just computer vision and animation algorithms.

It was a pain point from a start up I cofounded before where we needed Kubernetes flexibility but developers hated using it. I figured there was no reason why it had to be hard to use and thus Canine was born.

Lately been playing around with setting up remote dev containers on your own compute as well -- basically like github codespaces, except self hosted.

Built in Rails + Stimulus

Will be live 24/7 during October. Wrapping up technical work and building out the scene currently!

It's like a fixed up TypeScript specifically tailored for Wasm GC that can produce very small and fast binaries, but also has features like pattern matching, ownership / borrow checking for resources, pipeline, tail call elimination, multi-value returns, direct WASI component integration and a lot more.

Working with remote large data sets as if they where local.

I made the project public on May, and now as September has arrived, you can work with post production videos anywhere in the world. You can run the daemon on your own NAS and access the data on-demand from your working machines.

The same applies to data acquisition in digital forensics flows.

Anyhow it's basically just twitter but as a chat app. Everything is tag-based and there are no channels. Instead, users create and share lenses, which are just saved searches. That way, you can "speak in multiple channels" at once, or carefully filter out specific people or topics.

The best part is that I can speak, and then organize after the fact.

https://pipoca.chat

    {:find [?title ?director ?rating ?card ?shout]
     :where [[?m :movie/title ?title]
    
             [?m :movie/vote-average ?rating]
    
             [(str ?rating) ?rating-str]
             [(concat ?title ": ") ?title-colon]
             [(concat ?title-colon ?rating-str) ?card]
    
             [(lower ?title) ?lowered]
             [(concat ?lowered "!!!") ?decorated]
             [(upper ?decorated) ?shout]
             
             ;; joints
             [?cr :crew/job "Director"]
             [?cr :crew/movie ?m]
             [?cr :crew/person ?p]
             [?p :person/name ?director]]}

I’ve always had the same problem with website blockers: I might need YouTube for a tutorial related to what I’m working on, but allowing YouTube also means allowing every unrelated video. Same with Hacker News - I might need to look something up for work, but I shouldn’t end up reading the front page for the next 30 minutes. Most blockers operate at the domain/app level, but whether something is distracting really depends on what you’re trying to do.

With Nono, you describe what you intend to work on at the start of a focus session. It then uses an on-device model to look at the active window in that context. So a relevant YT video can stay open, while an unrelated video gets blocked.

The tricky part right now is the model. It needs to be small and fast to run continuously, while still being capable enough to understand the context and not annoy you with false positives.

On the side, https://infrabase.ai (hand-verified AI infrastructure directory) launched a Media Generation category, image and video APIs after a run of requests for it. Also the submission spam is way up here latley, been improving the detection here as well. Also seeing real interest in paid listings now, including one that started as a link-buying request I turned down :)

Been doing a lot of AI assisted emailing, so I have started to work on a little review UI for AI generated text, comments on specific words, inline edits, Claude picks up both and revises in place. Like a PR review interface for text. Turns out pretty good workflow with Claude watching the files for new comments, editing and commenting back right away.