Building AI that runs itself. Senior Engineer at @TradeRepublic. Speaker. iOS, Swift, Rust and Python. Making https://teapawt.org/t.co/Kq7FisXWls - https://teapawt.org/t.co/IJLnfrldRk
Berlin
Joined April 2009
- Tweets1.1K
- Following590
- Followers753
- Likes960
Convincing enough to go back to Pro
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.
I read it the opposite way. If we're typing less and talking to agents more, the keyboard is what loses relevance, and the screen becomes the main surface you actually poke at.
My doubt is more about timing. The iPad already covers most of that use.
Great demo, terrible ergonomics is basically the signature of a feature built for the keynote stage.
Isolated batch or not, my pockets collect lint and grit every single day, no beach needed. I'd still wait for the second gen Duo.
This is whats wrong with the iPhone Duo,
But like I suspected with my initial reporting about it, its most likely an isolated production issue with a specific batch.
Sure if you take your Duo to the beach and bury it in sand you'll have problems
teapawt.org/iHateApplee/status/211…
Worktrees were one of those git features almost nobody touched. Now they are basically the default way to run a few agents in parallel without them stepping on each other.
I'd expect a lot more dev tooling to get rebuilt with the agent as the main user.
Builders posting hot takes that point to their own product doesn't bother me. Someone who spent a year on a problem usually has a sharper opinion on it than someone dunking from the outside.
Just put the link in the post instead of pretending it's a neutral take.
An agent PR merging with zero intervention says more about your test suite and CI than about the agent.
Gell-Mann amnesia got a lot worse with AI. When a model answers something in my own field, I catch the mistakes right away. Then I ask it about taxes or health and just take the answer as is. It's the same model with the same confidence, I just lose the ability to see where it's wrong.
A simulator is great for checking flows and layouts, but it will never tell you if an app is comfortable to use. That only shows up when you hold the phone in one hand, thumb stretching for a button at the top of the screen, maybe while walking. For me that part still needs a real device, no matter how good the input method gets.
Merging chat and agent work into one product is harder than it looks. Chat wants quick back-and-forth, while agent work wants a clear task and then to be left alone for a while. I'd rather have two modes that hand off cleanly to each other than one blended UI that does both halfway.
This is the cloud credits playbook, now aimed at subscribers instead of startups. It's a good deal for experimenting, but I would keep the harness model-agnostic so swapping providers stays cheap once the credits stop being the reason you picked one.
Funny how unified memory went from the thing everyone complained about on Apple's price sheet to the main reason people buy a Mac to run local models. I guess the RAM upsell finally found its use case.
Looks like the glowing Apple logo is making its comeback. Hopefully there’s an option to turn it off. 🤣
Can Claude delete tweets now?
Inbox and calendar is where an assistant finally gets useful, because that's where most of my real context lives. It's also where a wrong action hurts the most. I'd start it on read access and let it draft, not send or accept invites, until it earns the trust.
When half a dozen teams ship the same thing in one week, it usually means a new model or paper just made it possible, and everyone wrapped it at the same time. The demo stops being the product at that point. I guess the ones that last will pick one real workflow, like real estate tours or game prototyping, and own it end to end.
