family mooncake 😋

in the age of vibe coding, the measure of a top programmer should be how long they can get Claude Code to work continuously without human intervention. if you can get Claude to code independently for a full 24-hour workday, you deserve top dollar in the market 🤔

another useful feature is Binpacking. by default, kuberenetes uses LeastRequestedPriority which prioritizes to schedule the Pod to least utilized hosts, and try best to spread the work load evenly.

unfortunately, when comes with GPU, this is likely to happen:

this weekend's paper reading is 'Sarathi: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefills', which discusses Chunked Prefills.

recently i learned a trick about using LLM is: let LLM to write a better prompt for me, then use the generated prompt for work. 🤔

the generated prompt often has a more comprehensive consideration for the task 🤔

it's a simple recursive func to generate randomized BOOLEAN expressions with the column values included, the final boolean value might finally executed as FALSE but it's fine, just prepend a NOT node to it.

omg reinforcement learning works 😲

now it can play up to 100 scores 😲

failed the training in the last night, tuned parameters and it seems good so far...

go to bed, let me see how much this dumb bird learns until tomorrow morning....

the micromover controller seems also broken 🥲

time to learn reinforcement learning 🤔

the scroll wheel seems to be stuck 😮

😆 on my way to read DDIA too: