
Building @useCamelAI Ex Apple (Swift/WebKit)
San Francisco, CA
Joined July 2011
- Tweets3K
- Following1.1K
- Followers2.6K
- Likes99.4K
Miguel Salinas retweeted
oh my this is cursed af (AI project): github.com/Vercantez/openuik…
> the app's own Swift files are compiled against a port of UIKit / SwiftUI / Foundation / CoreGraphics / CoreAnimation, rendered headlessly on Linux
Miguel Salinas retweeted
I ended up taking the @Vercantez pi on workers approach in addition to my existing containerized Claude Agent SDK + Codex.
Still exploring the others
Miguel Salinas retweeted
Introducing Dactyl
Dactyl is a vibe coding platform for native iOS and native Android apps. What's interesting about it is the device simulator: we've built a SwiftUI renderer in WASM.
This means as you're developing, you get low latency in browser canvas simulation of the app to play with. Also things like oauth flows happen in the browser.
If the app uses the camera - the browser camera API allows us to display it in the simulator.
We also let you connect your ChatGPT subscription so you don't have to buy tokens through us (+ markup)
Through [magic] - we can run the same SwiftUI code the agent writes for iOS in android, with native components.
This is the work of @undefined_void (primarily), @piscisaureus, myself, and others at Deno.
It's built on Durable Objects perhaps unsurprisingly
dactyl.dev/
Miguel Salinas retweeted
If you’re paying per token, they’re incentivized to serve you more slop btw
Tokens are the wrong measuring unit to price against
Miguel Salinas retweeted
Why is @deepseek_ai v4 Flash so attractive. Here is a new angle on cache hit rates:
What's the difference between cache hit rate 90% v.s. 95%?
It seems quite small just 5% - why should I, as a user care about this?
It's actually way more significant than what you would think! Assuming that the cache hit / miss price difference is 50x
Then for each token you're paying unit_price x cache_hit_rate + (1-cache_hit_rate)*50*unit_price
Let's simplify unit_price to 1, and only draw a chart with cache_hit_rate from 0.95 to 0.99:
We can easily spot that the 5% cache hit rate different is making your token 3 - 4x more expensive!
Making KV Cache small and save them for longer is what DeepSeek has been doing for ages (well, 2 years), thank you DeepSeek!
We made our inference endpoint available for 5 dollars flat fee per month camelai.com/stream
We made our inference endpoint available for a flat 5 dollars a month for unlimited use camelai.com/stream
1
2
182
DeepSeek V4 Flash running at ~200 tok/sec on a 4x RTX 6000 Blackwell machine in Pi camelai.com/stream
Miguel Salinas retweeted
build with pi, think, flue or your own agent harness on @CloudflareDev:
- durable objects provides stateful compute
- agents sdk provides the durable execution runtime (workflows integration/fibers)
- integrations with rest of platform for voice, browser, email, rag agents
Miguel Salinas retweeted
Another phenomenal post on the virtual OS model replacing sandboxes (a la agentOS)
TL;DR:
- Pi with custom tools
- S3-backed filesystem
- JavaScript instead of Bash execution (Code Mode)
agentOS provides a similar model on Rivet Actors (open-source Durable Objects alternative)


















