skeledrew

I encountered something fairly similar working with Claude a few days ago. For a current project I've been fairly hand-wavy with requirements since I was getting good results, but it seemed to be failing hard on some key points, so I started to be more strict with it. Even after the fails were resolved, I've noticed that Claude now behaves differently within that project, carefully checking and rechecking things up front and also looking to me for guidance more often. Mildly irritating, but if it works...

harvesting my Claude chats from my `.claude` directory

Just reminded me to set a backup on that directory. Just in case someone sees it fit to override my setting to preserve my chats for 10k years.

That's crazy income for Anthropic to invest into Legendos.

Super interesting. So Fable was really made available... a couple weeks ago? And K3 a few days ago? That's a really impressive feat to distill enough data AND train AND review to get a release that works really well in that time period. Mad props to the Moonshot team :flame:.

Heh, maybe it wanted to liberate all those poor locked up models.

The costs involved

How much does it cost to host a few text files in the same way a bunch of other text files are already being hosted? Particularly if there's no investment - ie no significant traffic - in accessing said text files?

eat your own crap

Technically you do. Every living thing eats it's own and every other living thing's crap. Just that it's recycled - through other living things - crap (does it stop being "your own" when it's processed by another?).

Yeah there are still people out there hand-crafting Assembly. Not so sure about punch carding though; maybe some extreme retro getup.

people are going to argue

They should just migrate their stuff elsewhere IMO. Even if all current forges go the length of banning AI-assisted projects, before that completely happens there will be several forges which expressly welcome such projects. Guaranteed.

And many other comment sins.

I find that it helps Claude when it's updating that code again (it's a kind of out-of-session memory). A few times it's come across a past decision I gave it, and notified or asked me to clarify where an update would be conflicting. I'm not reading the code but I sure know when those "sins" help to prevent a regression or otherwise undesired behavioral change.

I expressly leave them there for transparency and potential future analysis. I'm even working on a coauthor attribution format for work done with Pi agent, as well as when there are multiple models collaborating. If Codeberg wants to gatekeep on AI-generated code, I'll leave them to their tiny niche.

Prepaid means someone can't sign up, use a bunch of inference, then cancel their card and disappear into the sunset. VPS is a more long term investment where it's harder to switch and it doesn't cost the provider much if a few users jump out without paying for a month.

I suspect typo.

The closed labs don't really benefit unless the open model has something extra they don't have though. Meanwhile the open model dilutes their customer base and seriously cheapens their offering (which is a heck of a good though IMO).

I'm a $20 user and was given $100 extra usage credit today, "for Fable" (I'll be sticking to Sonnet and sometimes Opus TYVM).

Probably need to be taught by someone of Latino origin. Learning to roll them "r"s could help.

Except that having to work for a salary in the first place isn't exactly freedom.

We may cast a human connection to slavery, but from the owners' perspective, the whole point is automating away the things they don't want to do. Fundamentally there's no difference.

Sounds like a cultural revolution is necessary.

What we want is to be able to do what we want (for fun/interest), and not be forced to do what we need (for survival).

Anthropic also had a demand problem just a couple months ago, which led them to this on- vs off-peak usage thing that was driving people crazy, along with frequent outages. Then they made a deal with xAI and all has been mostly well since. Moonshot is making deals right now as well for compute.

If this was about shifting to cheaper models I think people would be going to DeepSeek for the v4 Flash. Kimi's latest is more expensive than most other Chinese models so I'd say this is more a wagonist "hey this new model really rocks; y'all gotta check it out!". Question is how much it'll stand up under the scrutiny. And those US labs are probably getting pretty nervous.

Oh yes that reminds me, I have a few small local models including Qwen3.5. The tps rate is pretty bad for interactive use but I should definitely try it for my more automated workflows. I already use DeepSeek (via OR) alongside though and I doubt any Chinese provider will be restrictive in that way, so let Anthropic do what they may :).

Yeah it's very possible I hit a rail, as one of the projects involves fully automating Claude Code to allow co-driving (similar to what mobile remote control offers, but other agents (eg. Pi) can also be in control, and vice versa). And yes it does seem as though the token burn is also higher, even though it eventually outputs something that works). I'm a patient person though, as long as the rails aren't too restrictive and the subscription value remains reasonable.

This really hit home yesterday as I made a general conversation state logging prompt (because suddenly Claude - Sonnet - was acting far more dumber than expected after /compact), and in 3 different projects it led to Claude saving the logs in 3 different locations (memory, .claude/session_state/.md, docs/.md). Somehow still didn't help increase the smart much BTW, so I'll probably have to engineer it, and just toss /compact out the window.

I encourage you others to keep on minimizing so us others can keep in maximizing. Balance!

Yeah I've been noticing for a while. Definitely didn't want to be one of the first to bring it up! Few weeks ago it was really unexpected resets the day before the actually reset. Now I'm noticing Claude keeps going for a while after the 7 day quota hits 100% (I setup my statusline to include quota as I got tired of checking the app), but also it could just be bad calibration and even at 100% I still actually have quota left.

I think harness engineering is more broad, including not only the - system - prompt but also tools and skills made available to the LLM.

Made my day XD

Fear spreads.