Matt Berg

@zeddotdev This is great anyways!

Your Claude subscription still works in the terminal. And next week we're shipping Terminal Threads to Stable so terminal agents appear in the new Sidebar alongside your native agent threads.

Anthropic's Claude billing changes hit June 15. We wrote up what it means for Zed users and what your options are: zed.dev/blog/anthropic…

@kepano @falseparfaitt Matter of fact just upgraded :)

@kepano @falseparfaitt I’ve tried a few. It’s well worth the subscription in my view.

@falseparfaitt It doesn't... there are dozens of free ways to sync obsidian.md/help/sync-notes

@emollick Why always the same voice though and not something fresh?

Haven’t tried this but it seems very neat… Yet all of the demos (except maybe one) are the model being fun and/or annoying by correcting or reminding in real time. There are obvious uses for this sort of model in meetings, education, training, etc. Why not demo valuable cases?

I love when claude tells me it'll add a half day work. Lol

@czue Here's the best one I got yet

I don't know how email as a protocol can even deal with this problem. We'll have to curate our inboxes with AI to guard against the AI. Such a waste of tokens. Everyone loses except the model cos.

This AI slop cold outbound BS really needs to stop

@czue And the biz loan phone calls

@czue I thought you meant taking over from a subagent

@mberg Literally just opening a new terminal and saying "hey can you review this PR"

An agent workflow I've been finding quite useful lately when coding sessions get long is instead of compacting or dumping context, just ask a new agent to do a code review on the work. This avoids the problem where agents don't like critiquing their own output, and often reveals new things to explore or fix. As a bonus, you pick up work in the review session with a fresh context window and carry on. Seems to work quite well!

@czue Gotcha

@simonw 100% agree. Best just works asr model I’ve found

Microsoft's MIT licensed VibeVoice speech-to-text model (think Whisper with speaker diarization) is really good - my notes on running the 5.71GB 4bit MLX conversion on an M5 MacBook, using about 60GB of RAM at peak and transcribing 1hr of audio in ~9 mins simonwillison.net/2026/Apr/27/vi…

My guess with all the claude usage complaints is that when we got 1M and stopped hitting auto-compaction most of the usage is just bloated context now.

The Batman soundtrack by Prince is underrated

Excited to launch Pixel. Explore the world's most comprehensive population dataset in your browser. Featuring the @WorldPopProject 2 new datasets. geopixel.io

@trq212 @Scobleizer Ideally you could base it on people’s local time (or let them select their window) this is how ev charging work. Still sucks to have this change

@trq212 @Scobleizer Definitely favors those on the west coast. If you are east coast this is peak work hours from 8 to 2pm not ideal

To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged. During weekdays between 5am–11am PT / 1pm–7pm GMT, you'll move through your 5-hour session limits faster than before.

Really excited to see the release of channels. To leverage this I created agent-http that allows you to wrap claude code with an HTTP API. github.com/mberg/agent-ht…

@trq212 You can leverage channels to create an http api for claude code see: github.com/mberg/agent-ht…

We just released Claude Code channels, which allows you to control your Claude Code session through select MCPs, starting with Telegram and Discord. Use this to message Claude Code directly from your phone.

1 million context window: Now generally available for Claude Opus 4.6 and Claude Sonnet 4.6.

What movies, books, albums or shows would you bring to a Desert Island if you could only pick 5 each? desertislandpicks.com