How do I still not have an AI that can draft/write emails in my voice that isn't horrible? Gemini/Superhuman AI writing is still slop and I need to edit after writing. Is this problem way harder than I think it is?

OpenAI's Codex and Jules don't feel as great to use as I initially expected. I think it comes mostly from. -Basically a docker build on every change (I don't think images solve) -An awkward make remote PR, pull down to try workflow. And yet parallel tasks is clearly good...

Thought: we might NEED AI to take our jobs. Today’s teens started on Instagram and developed on TikTok’s dopamine delirium. And now, AI eliminates any task that smells of intellectual depth. For today’s work culture, they are incredibly cooked. But not if work is reinvented.

“BREAKING” is Twitter’s way of telling me who to mute next.

Listen with eyes closed as the choppy highlighting makes it seem worse. Legitimately amazing - around 60% of this is indistinguishable from a real human to me.

Cursor 1.0 + Gemini 6-05 together = SOTA? Will wait for 24 hrs to see people’s response and re-evaluate Claude Code after.

Please one of them do a livestream now

iPhone 17 Pro next to iPhone 17 Air. Jobs would never.

My unorganized thoughts on Claude Code (Opus 4): For context, I've used it more than almost anyone - I've been rate limiting three 5 hour sessions on the 20x MAX plan on the average day, which is slightly degen-level. 1. It’s good. Addictively good. I just saw a post on

Cursor's background workers don't feel awesome for the same reasons. I really think I want local, git worktree based workers that use my existing environment setup instead of building a new one. Surprised that they went remote-first instead of shipping that later.

Puts on Apple - WWDC was extremely underwhelming against the backdrop of AI progress. They need to ship a cloud model (does NOT need to be SoTA), slap some verifiable privacy on it, and embed seamless AI directly into text entry, the screen, and apps on iPhone and Mac. Not for

Can somebody explain the mind virus of "they made the model way worse?" Shortly after any model release there are highly upvoted posts by people 100% convinced that the makers nerfed the model to oblivion. I'm fairly certain this doesn't happen, so why do people think it does?

Kind of interesting: the type of words people use to indicate “et cetera.” “Yada-yada,” “duh duh da,” “and so on and so forth,” “and you know,” “whoopity whoop,” “this that and the third” and my favorite “bish bosh bash.” I wonder what it says about where you grew up.