Posts

I don’t use Codex currently, but OpenAI seems to have a use cases page which shows how to steer agents into making native Mac + iOS apps.

I assumed this was proprietary and in their Codex app, but I’d be curious to see people importing these into OpenCode/Claude Code, etc.. 👀

an interesting point about “agentic engineering” right now (more professional vibe coding) is that experienced devs who can control the agents better seem to have a better experience, i wonder which app/company will be the first to put an LLM in between you and the agent though, to turn natural language into technical language for the coding agent… i imagine this would improve results hugely?

for the past hour or so I’ve been playing with skycoll - my quick, vibe coded atproto equivalent of twecoll (an old thing I saw for twitter a few days ago on gh)

can see the source code here and can install with PyPI!

has atproto OAuth, but I’ve only been using it with an alt for safety atm!

Aside from the obvious features, like replacing Spotlight search, I think Raycast Focus is one of my favourite features. It lets you block apps and sites through categories (e.g. socials) and adds a pomodoro-like timer to the menu bar. Actually helps me study! Forgot about it until recently!

They finally added sync to Dia!

I’m switching between Firefox and Dia as my main browsers currently, wonder if this’ll change that… I loved Arc’s full tab sync feature, it was so sick!

Two screenshots of a web browser interface show different views, one with a simple search bar and the other displaying a detailed list of files and folders. “Here we have two separate Dia instances that have been synced to show changes in real time. This will also happen across two separate machines!”

ok Blip is very cool. i’m currently sending my friend who’s on holiday atm files at nearly 100mbps because of peer to peer tech!!

wait what, Meta just announced a new LLM - their first non OSS one(?)

why does it actually seem good tho??

Muse Spark is the first step on our scaling ladder and the first product of a ground-up overhaul of our AI efforts.

A table compares performance across various benchmarks for different models in categories such as multimodal tasks, text reasoning, and health.