3M free tokens on GLM 5.3 Flash — 2M plus another 1M, no card, no trial clock. 
Spend them in this order — each build ships before the next one starts
README generator — points at a repo, writes the intro
changelog bot — reads commits, writes the release line
CSV cleaner — the spreadsheet you keep avoiding
commit-message writer — git hook, two words in, a real message out
regex helper — describe it, get the pattern with tests
bug-report summariser — twenty comments into three lines
docs translator — runs in the background, you never watch it
prompt library — on your own disk, versioned
code-comment rewriter — legacy files, readable again
release-notes drafter — tags in, customer-facing notes out
quiz maker — any PDF becomes questions
chat UI backend — the smallest thing you would actually use
Why they survive — batch the calls · temperature low for code, higher for writing · keep every prompt that worked. The library is the asset; the tokens were only fuel.
!