Charles Cushing

How did Siri get dumber?

My rice vendor is implying incoming food inflation

There's a huge product opportunity downstream of Tailscale etc too. At @tldraw we've started designing for "high trust networks" and it's been a beautiful thing. 1. Features that would be a nightmare on the internet are okay on my computer and mostly okay on private networks. Websites that execute user-generated scripts, control production systems, allow for users to send desktop notifications, spend tokens, or directly prompt my agent. Crazy stuff that would only make sense if everyone was in a committed polycule. 2. When everyone on the server can be trusted to have _external_ reasons for good behavior (eg keeping their job, or maintaining friends, family, church, mafia org), the software can be designed in a default social way. Most apps default to isolation, low horizontal visibility, highly restricted collaboration. Makes sense if the social contracts are inside of the app, but if you can trust that they live outside of the app then you can get wild. Example being small group chat names as an emergent comedy phenomenon because anyone in the group can change the name. 3. This can work for commercial products, too, not just internal stuff. Security matters more than ever but the way we've seen that manifest is that more companies are more confident in their own solutions than anything we could provide, to the point where owning data and access is a better value proposition than having it managed. Always has been true for big companies but now even small companies can get a good secure network set up pretty easily, and the hooks we need to build are easy enough. tldraw offline got us started down this road (and shout out to @pendev for first introducing me to the idea of desktop + mcp). We've been bringing that product to private networks and it's been exciting to see how much we can get away with. Keep an eye out

I encourage anyone who doubts this to try and make a real AAA game and apply your own real critical game quality standards to it. I am making one right now and while Opus 5.5 (Fable and Astra can't do it) can do many of the coding tasks, it simply isn't able to make gameplay design and UI design decisions at the level of detail and correctness demanded by a GAME. My theory is that it's because the quality requirement in a game is ironically higher than for e.g. a SaaS application. SaaS or other "work" applications provide economic value for using it, so as long as your economic value gained is high enough, you'll tolerate bugs and UI imperfections and lag, even outright illogical app behavior - as long as it does the thing you need it to do. But when you're playing a game, your willingness to continue doing so rests entirely on the quality of the experience. And that's impacted by gameplay, fun, UI, story, etc. If there's any sort of "irritant" like a part of the interface is janky, or something consistently looks wrong, or the gameplay does something illogical, or it's laggy, etc - you don't have any incentive to keep going. To keep someone playing a game, the experience has to be far better than that - maybe not perfect - but it still has to be great. What I'm finding is the Opus 5.5 will make a game at roughly between 50% to 85% quality level. If you give it a lazy prompt, it'll puke out something with blocky graphics and barely-working gameplay: 50%. If you iterate a bit and give it more details, you can get it up to 85% or so - playable, can generate impressive screenshots and maybe even gameplay clips - but that last 15% or so still needs hand-tuning and human taste-making. You can have it read every game design and UI/UX guide you can find, and it just won't build anything better than "kind of workable." And "kind of workable" is often enough for a work app. I have like a dozen vibe-coded utility apps that roughly do the thing I need them to do (track stuff, manage some kind of data, visualize something) but they're not worth polishing because they're good enough to perform the function. But with games I can imagine someone else looking at it with the same critical eye that I would, if I were playing it: - these buttons are a little out of place, the alignment isn't perfect - why does this icon which does the same thing as this other one look different? - this tree or shadow looks weird from this angle - this game element acts a little janky under these conditions - there's lag when stuff gets busy In particular, AI doesn't seem to be able to create really good custom UIs. If you have a bunch of commonplace functions then yeah, it'll use a UI template. But if you have a game, and the game involves doing a bunch of unusual (fun!) things, you want the UI to engage the user's intuition about it, show the right responsiveness or progress of an action, act correctly in certain unusual corner cases (because fun often involves corner cases) etc, and AI can't quite get all of it right yet. It's workable, you can often see the vision - but it's not a game. No AAA game studio would ship what AI produces today. Much of this seems to be about paying attention to the right details. Maybe subsequent models will be better (they often are). But right now a human still has to look at every little corner of the game, decide if it's acting just right, then specify in extensive detail exactly what should happen and why, and what overall effect is intended. The good news is that one person can be the gameplay designer, the artistic director, the engine developer, the interface designer - you can do all of these jobs and let the AI carry out the tasks. The bad news is that one person has to be the gameplay designer, the artistic director, the engine developer, the interface designer... and I'll probably have to be the marketing guy later when/if the game is released. Consequently, I've had to quickly learn the fundamentals of all these disciplines. Unsurprisingly given my SWE background, the engine on my game is quite well-optimized even with tons and tons of moving things and physics. But the gameplay is another thing and I don't even know the right terminology to describe a lot of the art/graphics things I want to see, and... oh my god, UI design is an AI clusterfuck... Back in the day, tech companies sort of had a hierarchy of respect when it came to technical staff - I'm a little ashamed to say - it was roughly: 1. backend engineer 2. frontend engineer 3. visual designer 4. UI/UX designer Yes, the lowly UI/UX designer, at the bottom of the totem pole - not even considered "real" designers - they were ones responsible for deciding where to put the buttons and what fonts to use. Well, today the AI does everything else but it still can't design a great interface worth a damn. So yeah, to the UI/UX designers I knew: I owe you an apology. I wasn't really familiar with your game. Of all of the tasks, this is the one that AI can't seem to do - I have to tell it how to do things and it's extra hard when I don't even have the terminology. I can describe how I want things to work, but the AI doesn't have the interiority of a human user - there's a lot of intuition and feelings involved - it does boneheaded things, ignores critical details that would be obvious to a human, doesn't realize which things it absolutely mustn't overlook on a UI and worst of all: even when I feed it a precise sketch of the interface I want, it ignores critical details because it simply didn't realize those details were key. (example: when I draw this thing next to this other thing, it literally means they need to be next to each other) If anyone wants a benchmark: I've been working on a game for about 6 weeks now - coincidentally I got started right before Opus 5.5 came out (when attempts with Astra just couldn't cut it). Granted, I have a day job so a lot of that was after-hours and weekends but I'd estimate that I have gotten done the equivalent work of a 10-person studio working for ~6 months. So there's definitely leverage here, but there's definitely a lot more work to go before this is a real game.

planes are so fucking dope

the fact that grass can become a sweater through the intermediary of a sheep ought to astonish us more than it does.

the @every team, nov 2026 so proud of this crew!

This is probably obviously AI complete already, but: you get *really* good fiction recommendations, including not-the-usual-suspects, from SOTA LLMs that have a handle on you either from the Internet or through interactions.

OpenClaw deleted around 400k LOC of its own tests without much change in code coverage. Modern models just love writing tests for every tiny change, even if they aren't useful. This skill helped. github.com/openclaw/openclaw…