thefroh

it's like AI-generated images, though those tend to be more obvious than AI-generated text due to the "average" effect being visually obvious. Yes, these are normal patterns in English, but it is not normal to use them with the frequency that triggers people's llm-lese senses.

I don't see a way out of it until either generated content is either completely unable to be distinguished from authored content (unlikely?), or there's a silver bullet for identifying generated content (also unlikely?)

I'm a little surprised that preserve_thinking would matter here for cache purposes. for actual capabilities/intelligence, yes, I'd imagine it helps to have past reasoning traces in multi-turn setups.

but for caching, all you are doing is leaving off a fraction of the most recent assistant message generation, which will have little/no impact on cache hit rate.

possibly https://xkcd.com/353/

presumably it's the reverse engineered server that has most of the work put into it, and one would hope that's what is going to be released if the developer decides to

which is important to bear in mind if people are introducing a "drop earliest messages" sliding window for context management in a "chat-like" experience. once you're at that context limit and start dropping the earliest messages, you're guaranteeing every message afterwards will be a cache miss.

a simple alternative approach is to introduce hysteresis by having both a high and low context limit. if you hit the higher limit, trim to the lower. this batches together the cache misses.

if users are able to edit, remove or re-generate earlier messages, you can further improve on that by keeping track of cache prefixes and their TTLs, so rather than blindly trimming to the lower limit, you instead trim to the longest active cache prefix. only if there are none, do you trim to the lower limit.

not if call #1 is the file + the question, call #2 is the file + a different question, no.

if call #1 is the file, call #2 is the file + the question, call #3 is the file + a different question, then yes.

and consider that "the file" can equally be a lengthy chat history, especially after the cache TTL has elapsed.

because you can have multiple breakpoints with Anthropic's approach, whereas with OpenAI, you only have breakpoints for what was sent.

for example if a user sends a large number of tokens, like a file, and a question, and then they change the question.

here's the associated issue, I believe.

https://github.com/syncthing/syncthing/pull/10005

does this allow running applications which need FUSE? I know for termux that's only possible with root.

it's also the effect that lets you kinda know if you're near a wall (for example when you're fumbling around in the dark)

i've heard good things about using the 1 euro filter for user input related tasks, where you're trying to effectively remove noise, but also keep latency down.

see https://gery.casiez.net/1euro/ with plenty of existing implementations to pick from

while I'm a fan of TypeScript and using type hints in Python from an autocomplete and linting perspective, I am curious...

... has either language leveraged these to better tell the CPU what to do? presumably for perf.

for me, my desktop and laptop are the main go-to. the mobile is an extra device with different, more specific use cases

and so I've been a little disappointed with how these devices keep getting bigger and bigger. I was pretty happy with the size of the Pixel 3

I think I like to be able to access the whole screen comfortably with one hand, not fumbling it about. easy to manipulate, easy to pocket. the Pixel 8 shrunk a bit over its predecessors so I nabbed that, and it's probably at or just over the limit for me, size wise

is anyone aware of any GUI-driven tools that leverage SAM2 yet? Especially with video.

they're ambiguous terms, and the ruling will be "I'll know it if I see it!"

it's more useful to consider what people suggest for a track.

- "you're going to need a diff lock"

- "you're going to need high clearance"

I think it is more that no server side setup is needed beyond git to support this. That's to say regardless of the hosted git provider you may be using, they do not need special support for git-appraise.

As an aside, if I remember correctly you can also run git without any server at all with multiple contributors. But it isn't the most ergonomic!