It’s a partial driver, perhaps even meaningful in some way, but this is hardly an A/B test. Suffice to say lack of CarPlay is costing them in some way but model refreshes, steep price discounts, and other economical considerations are going to win out. Sure, 79% of some sample of car buyers said they wouldn’t buy a car without carplay but putting carplay in an otherwise crappy car isn’t going to magically make people buy that instead. There’s a lot of reasons Chevy is struggling. Blowing their own foot off with this decision can’t have been good for business but it’s certainly been hilarious for the quotes their execs have had to say in public with straight faces so they don’t get whacked for not following the party line. At some point I’m sure there will be a reorg where the opposite will become true.

If everyone took your position, then frontier labs can keep raising their price.

what trade off would one need to clear to justify the hardware and the work to get this running locally as part of a broader system? It’s a lot of work setting up and maintaining a production harness/system on a local device. I don’t personally repeatedly generate images at a scale where using a lab’s app somehow burns all my tokens. I like the ideas of local ai but I don’t see widespread adoption of it happening in commercial or customer situations anytime soon no matter how little/good enough they get. Even Uber- token burn whiplash but I doubt their answer will be “run some of it local”. IT nightmare, I’d imagine.

wow, this blows on so many levels.

anyone remember the whole “delete uber” thing from 2017ish? good times

I’m super envious. I can’t seem to do anything without a half a million tokens. I had to create a slash command that I run at the start of every session so the darn thing actually reads its own memory- whatever default is just doesn’t seem to do it. It’ll do things like start to spin up scripts it’s already written and stored in the code base unless I start every conversation with instructions to go read persistence and memory files. I also seem to have to actively remind it to go update those things at various parts of the conversation even though it has instructions to self update. All these things add up to a ton of work every session.

I think i’m doing it wrong

Something sounds very wrong with your setup or how you use it.

Is your CLAUDE.md barren?

Try moving memory files into the project:

    (In your project's .claude/settings.local.json)

    { ...
      "plansDirectory": "./plans/wip",
      "autoMemoryDirectory": "/Users/foo/project/.claude/memory"
    }

(Memory path has to be absolute)

I did this because memory (and plans) should show up in git status so that they are more visible, but then I noticed the agent started reading/setting them more.

This does kind of smell like the wrong way to use it. Not trying to self-promote here, but the experiences you shared really made me think I headed the right direction with my prompting framework ("projex" - I once made a post about it).

I straight up skip all the memory thing provided by harnesses or plugins. Most of my thread is just plan, execute, close - Each naturally produce a file - either a plan to execute, a execution log, a post-work walkthrough, and is also useful as memory and future reference.

Something seems wrong. A half-million tokens is almost five times larger than I allow even long-running conversations to get too. I've manually disabled the 1M context, so my limit is 200K, and I don't like it to get above 50%.

Is it... not aware of its current directory? Is its current directory not the root of your repo? Have you maybe disabled all tool use? I don't even know how I could get it to do what you're describing.

Maybe spend more time in /plan mode, so it uses tools and the Explore sub-agent to see what the current state of things is?

Two quick thoughts:

- Use the Plan mode, create a thorough plan, then hand it off to the next agent for execution.

- Start encapsulating these common actions into Skills (they can live globally, or in the project, per skill, as needed). Skills are basically like scripts for LLMs - package repeatable behavior into single commands.

If i had to guess i think you have probably overstuffed the context in hopes of moulding it and gotten worse outcomes because of that. I keep the default context _extremely_ small (as small as possible) and rely on invoked slash commands for a lot of what might have been in a CLAUDE.md before

Sorry, the point? isn’t the point of art pretty much what a person wants it to be?

Not OP and I won't dispute your point exactly but I'd like to point to a book called Pixel Logic wherein the author makes the same point regarding pixel art. Even though you'll be using stuff like the Lasso and Paint Bucket tools the big thing about pixel art is the manual control and precision of pixel placement (by hand) where you employ techniques like anti aliasing (again by hand). Advanced techniques like sub-pixeling when doing animation frames are another thing that makes sense only when you can place pixels one by one.

I have really struggled to get nano banana to follow size/proportion ratios for sprite art. any tips? I fed in a bunch of examples first and tried to write a really strict prompt. I wonder if any of the sw being discussed here can be programmatically controlled by claude code or similar to do sprite work

Like the comment above I split sprite sheets into grids with edges for NBP to follow. I have the option to add the canny edge map to the grid to enforce a lot of consistency as well. Then I specifically tailor the prompt to the task.

But even still it has issues sometimes.

entirely possible I’m just really bad at this stuff but I can’t get browser agents to do simple report pulls without running into a captcha or a dropdown menu that breaks its brain. hopefully this is the one!

good security will always be the eternal enemy of easy automation.

the realm of bots vs bots

this is a disgusting amount of money for this