Viewing profile — synergy20

unsloth for small use cases, tinker for enterprise level use cases?

what are those cheap fine-tuning services these days?

V4L2 (MJPG) -> cap.read() -> CLAHE(night) -> imencode -> _current_frame (web MJPEG) -> resize .25 -> gray -> blur -> absdiff -> contours -> motion? -> motion? VideoWriter.write() :…

you probably are overly sensitive and biased, do not forget transformer is the only 'original' for AI so far, the 'original aspirations' nowadays are just incremental improvements.…

morningstar said the fair price is $62

it's widespread in rewriting CLI tools, so it's terminal

The "Rust Evangelism Strike Force" is so annoying and it becomes the major reason I don't want to have anything to do with the language, yes I know it's strange. Almost all HN soft…

there are quite a few companies are developing 'proctoring' system to defeat the cheat so you can take self-exam in realtime and everything will be recorded and anything suspicious…

but, it's $160/month(unless you buy a one-year plan that gets cheaper), not too far from $200/month from claude and codex? why should I switch?

or use git-shell so users do not need a full ssh account on the server due to security concerns: which git-shell sudo echo "/usr/bin/git-shell" >> /etc/shells sudo useradd -m -s /u…

hold on, are you saying graphenos has no ads everywhere? I need swap it in then

how is this different from lvgl? is this immediate mode or retained mode?

can vlm be used instead or it's too heavy and slow

still 10x better than the 'finish this leetcode tweak algorithm in 20 minutes and tell me your thought process along the way, and yes you will never need that skill in the real job…

It's also 3x slower than opus 4.8 per my use, and 10x slower than codex. Codex can find key design issues in 2 minutes yet Fable is clueless after spinning 20 minutes.

For embedded products, 1GB can go a long way thanks for linux customization and such.

truly scary. 2x at least token burning rate comparing to 4.8, can indeed run auto edit mode for hours. use it for super complex tasks then use cheaper model to do the rest, else wi…

does zig still matter in the AI era? i was learning zig on and off and now have no motivation to learn it anymore.

After a few months, now I could barely use up my $200 Claude Max unless I try really hard, I got it corporate API billing is more expensive, but, $1500 really should be enough for …

ollama does not support this yet, what else can I try

can it run linux as I am not a windows nor a macos user

For me, Pi6 shall focus on NPU to run edge models instead of getting into minipc space, e.g. 10TOPS built-in, that will be slightly better than Rockchip 3588 which was produced 4 y…

kind of agree here, golang is my main LLM language. for low-level, it's C or Lua, or modern c++ for simple cross-platform system daemons.

I'm using Vue and do not feel it's becoming react