Kyle McDonald

imagine twitter, but there's a bot is crawling all the tweets and when people post "add an edit button", the bot implements it. what would a fully crowdsourced social media platform look like? and what constraints would keep it from breaking apart?

if someone wants a good masters or phd thesis, they should point one of these tools at the entire github history of OF including all the issues, and ask "make a better toolkit that solves the most common problems and regressions that people run into, and make it easier to vibe code new apps"

these tools will often think there is a y-inversion when there is not, or hallucinate functions that it would make sense for OF to have if it worked like other toolkits, etc.

i use cursor (usually in auto, sometimes i switch between chatgpt, opus, and gemini for persistent bugs), or claude code if it's a web app. cursor can basically handle OF, but OF is so nonstandard in so many ways it would probably be faster to code against the raw libraries a lot of the time.

it is at least 2x harder to vibe code an openFrameworks app than anything for the web (or even python)

after 2.5 years of vibe coding, my biggest takeaway? native apps are dead. iterating for the web is so much faster, has better tooling, and lower overhead. low-latency, multi-projector, 3d, spatial audio, custom hardware—ai will continue to have trouble with these.

oh this app doesn't have an edit button either? ok

A graph of the monthly question count on Stack Overflow starting in 2009, ending in 2026, starting and ending at zero questions and peaking around 200,000 questions in 2014 to 2018.

RPI stack overflow (2009-2026)

i was just helping people in the south pacific with their phones. every single phone was infested with adware that made the phones unusable. if we managed to open facebook it was full of fake “phone cleaner” ads.

if you're enjoying these, please check them on another platform like instagram or youtube. bsky has a 100M upload limit so the videos look terrible.

www.instagram.com/kcimc/reels/
www.youtube.com/watch?v=J2Wf...

sailing from lumolumo to dawson island

water witches and solar power

rewiring my relationship to technology :) helping with various upgrades and teaching. but we are also still trying to capture this mysterious flash of light.

from the other side of the welcoming party

love the circle packing approach to sizing

betel nut, berries, apples and coconuts

i cannot even begin to explain my rage at how glitchy the instagram reels captions tool is

new cables and old ladders

a short trip from alotau to lumolumo with an incredible welcoming

“transformirror” with daito manabe will be at cinekid this year 🤗

i’m on day 19 of 50 days in the south pacific, helping upgrade power and internet for two traditional voyaging organizations—and trying to capture a rare flash of light called “te lapa”. i’m posting daily on instagram instagram.com/kcimc

the latest realtime video-to-video demos are wild

i think i see what you’re saying about there being a spectrum between treating AI-assisted search as a pure retrieval task vs a pure hallucination task. lots of hybrid in-between solutions that balance the strengths of each.

i tried this with o3: “do some research on the best rust proof nails. then provide a single recommendation along with sourced quotes justifying the answer. include quotes that come to a different conclusion and explain why they are wrong.”

re the original point: in theory you can file a complaint with CRCL but they've become toothless or complicit. the DHS OIG has forced some changes, but not individual accountability. personally i still feel identity is important, even if there is no clear route for accountability.

there is a precedent for takedowns of US-based sites (e.g. icelist) but icespy has been uninterrupted since 2018. at first i thought it was just under the radar, but two separate fox news hosts have painted a target on my back in the last week and.. nothing. so 🤷🏻‍♂️

first step is getting a database of headshots. in LA that was acquired through CPRA (like FOIA for california). once you have that ping me and we can build it out!

the first step is getting a database of photos that definitely includes these guys. then face recognition can help reduce the difficulty of searching through thousands of faces manually. but it cannot put a name to someone without an existing database.

there is currently no up-to-date database of ICE employees. but post-COVID there has been a lot of research on masked face recognition—and with new image generation techniques it should be possible to train even better masked face recognizers arxiv.org/pdf/2109.01745

next time someone asks why we don't have better pics of UFOs i'm sending them this 20 meter object that is literally broadcasting its exact location