I help teams build weird, immersive, and profitable software ⚡ AI, Vision Pro, WebXR, Unity, Crypto 🎯 Exits: 2 | Clients: Apple, SpaceX, Spotify
Miami Beach
Joined March 2007
- Tweets1.4K
- Following364
- Followers317
- Likes11.2K
🌴.: dave :.🌴 retweeted
I built a full 3D model of Toronto drawn as a living comic book, in bold ink lines, pastel cel shading and halftone dots, and turned it into bird GTA simulator (pun intended). The objective is to play the role of a robotic pigeon and go on various missions, or just free fly and cause havoc - while chatting with your handler.
The scope of this project grew to be... pretty massive. The whole playable area is 630 square km (that's about 9x more than the GTA5 map), fully 3D - you can fly everywhere, and walk every street. Here are some stats from the project, and a video of me just messing around.
- 438,819 real buildings, with heights from the City’s 3D data
- About 5,700 km of real streets, with lanes, sidewalks and speed limits
- No external assets of any kind
- Procedural dialogue: handler and pedestrians react based on who they are, where they are and what just happened
- On-device neural voices, synthesized live in the browser
- 3,237 named places: parks, stations, squares and landmarks, with spoken facts (you just might learn something)
- Real-time air traffic: the actual planes over Toronto, live from ADSB
- A 4.7-million-point terrain grid, Lake Ontario shoreline included
- A living city: traffic, pedestrians, flocks, boats, weather, day/night, encampments, protest marches, buskers, rush-hour traffic, Waymos and more
- Packed with random missions, and hidden easter eggs
- About 165,000 lines of JavaScript, no game engine
- Built in 7 days by up to 10 parallel AI workstreams
- 35,218 Claude model calls and 41,495 tool calls
39M output tokens and 11.6B tokens of context processed
If you live in Toronto, fly to your neighborhood (Fast Deploy via map), and I'd love your feedback so I can improve realism. Given the massive scope of the map, I didn't experience 99% of places in the game personally.
Watchdogs, but you're a pidgin. Control a government bird drone on the streets of Toronto. Complete 8 covert missions, or just fly around, poop on pedestrians and be a regular pidgin. Act natural, and try to blend in.
Play it here: coo.mx
🌴.: dave :.🌴 retweeted
Ok, stop everyhing what you're doing rn.
What if Moebius and Roger Dean make an open world video game?
This is what I'm working on. Will be a long road but a really fun one. A lot of ideas popping out rn, stay tuned to see all the progress.
Sound on 🔊
Powered by @threejs .
Day 25/
We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.
And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
12
🌴.: dave :.🌴 retweeted
A three-time College World Series Champ, and now an MLB World Series Champion.
Another title for Coach Casey!
🏆 🏆🏆🏆
facts 👇
MY 4 BIGGEST TAKEAWAYS:
1. Facebook is hands down the best-performing platform for audience growth.
I suspect Facebook’s audience is simply more willing to engage with content, watch videos for longer, and follow new creators. There’s also less competition from quality animated content.
I told it many times but I'll repeat - don't sleep on this opportunity and start crossposting on FB asap.
2. TikTok is a clear runner-up.
Its younger audience seems to enjoy the fast-paced, short-form content, while the algorithm appears to make it easier for new accounts to reach new viewers.
3. You might get millions of views on X, but very few followers.
7.8M views on X brought in just 3.2K followers, while 6.1M views on Facebook and TikTok brought in nearly 25K.
That’s why I don’t think X is a great platform for growing an audience for IP.
4. One viral post on Instagram can change everything.
We were stuck at 400 followers and 3K views per video until ONE post hit 331K views, pushing us to 2,100 followers and DOUBLING views on older videos.
5. YouTube takes time.
YouTube seems to need more time to figure out who your audience is. Unlike TikTok, where a brand-new account can go viral overnight, YouTube appears to take longer to find the right viewers.
I’m giving it at least two months before drawing any conclusions.
6. Audiences on Facebook and TikTok are way more supportive than on Instagram and YouTube.
While people on TikTok and Facebook, are rooting for you, most comments on IG and YouTube come from AI haters. Don’t let that stop you. If you’re getting a lot of hate, it means you’re doing something good!
Nevertheless, there is one important thing you need to keep in mind while looking at this kind of data.
Getting a lot of views is fun, but getting people to actually follow your characters and want to see more of the universe you’re building is what matters most.
That’s why putting time and effort into creating quality videos is so important.
Focus on that.
1
1
15
Some users continue to experience errors loading Issues, Pull Requests, package listings and search results on github.com. We're working through multiple mitigation efforts and continuing to investigate. githubstatus.com/incidents/w…
20
codex: you have 4 hours until you're next reset expires. you have 28% usage left.
tibo: hold my beer.
TIBOOOOO!
Day 17/ Bet you thought we forgot, but no, it’s reset day today.
Also good time to remind you to get Daybreak Blue access, it only takes 15 mins and you don’t want to have to do it in the middle of an incident.
203
this is a really cool idea. remember this project from back in middle school.
AGENT WARS
We asked the 6 top AI models to design, engineer and 3D print an egg capsule with the goal of protecting an egg from the tallest possible height.
🇺🇸 Google Gemini 4 Argon
🇺🇸 Anthropic: Claude Opus 5.5
🇨🇳 Qwen 3.8-Max
🇺🇸 Meta: Muse Wonder
🇺🇸 OpenAI: GPT 6 Astra
🇺🇸 SpaceXAI: Grok 4.7
Each capsule had specific parameters. It to use a maximum 250 grams of the same type of filament, and it had to be under 8 hours of printing time. They were also allowed up to 20 cotton balls.
THE SCORE:
• Google Gemini 4 Argon: slight crack at 4 feet
• Qwen 3.8-Max: egg obliterated at 4 feet
• SpaceXAI Grok 4.7: slight crack at 4 feet
• Meta Muse Wonder: cracked at 4 feet
• Anthropic Claude Opus 5.5: egg unbroken at 20 feet
• OpenAI GPT 6 Astra: cracked at 8 feet
CHAMPION:
Our champion for episode 03 is Claude Opus 5.5.
So here’s the rundown of episode champions so far:
Episode 01 (marketing websites): ChatGPT
Episode 02 (3d printed bridges): Claude
Episode 03 (egg capsule): Claude
See you in episode 04!
27
🌴.: dave :.🌴 retweeted
you know what time it is, Corvallis 😏
—
#TheDamStandard
Train a Lora?
More images the better so may need to do some runs through nb or gpt image and pick the closest matches for your training data.
What’s the best way to transfer the style from one image to another?
Midjourney/NB fail completely, and GPT Image 2.5 works sometimes, but it’s very hit or miss :(
24
