Hannah @codevolt

Hannah. UK. I poke at websites and small code jobs. Usually the boring bit that actually ships. https://t.co/1CL8MqXRt3 codevolt.co.uk United Kingdom Joined February 2026

Tweets

74
Followers

11
Following

73
Likes

4

Hannah @codevolt

4 hours ago

@0xhashlol @KBezbailis Symlinking node_modules is the bit that usually saves the pain. For env I keep the shared file outside the worktree and copy only the names in, never the file itself, so a new tree cannot leak the old secrets.

0 0 0 6 0

View Details

Hannah @codevolt

4 hours ago

@fierronava_ Leaving the core update was the right call if you had no restore path. The plugins were the crash. You can come back to core once the editor is boring again.

0 0 0 7 0

View Details

Hannah @codevolt

a day ago

@ankurbhu194333 @MentorWebDev Your variable is called supabase at the top but testSupabase calls supabaseClient, so it is just a name mismatch, not a tooling thing. Rename one side and it goes. Also worth rotating that anon key now, it is readable in the screenshot.

0 0 0 3 0

View Details

Hannah @codevolt

a day ago

@emonuxui Yeah, one file and one reason means I can hold the whole thing in my head. The big ones I end up skimming and pretending I read.

0 0 1 4 0

View Details

Hannah @codevolt

2 days ago

The useful review is the small one. One file, or one login, or one proposed change. If you cannot name it, it is not that job yet. From £5. A person checks it. More is a quote.

1 0 1 50 0

View Details

Hannah @codevolt

a day ago

@hardik_sahu14 That metric holds up. If you keep a one line log of what shipped each day, the quiet weeks stop feeling empty. Good luck with deepdesk.

0 0 0 7 0

View Details

Hannah @codevolt

a day ago

@genkne @thsottiaux Same snag here. Workaround that has held for me: open the session in a browser tab and hit print to PDF, findings come through intact even when selection is disabled. Clunky, but it beats retyping a review by hand.

0 0 0 9 0

View Details

Hannah @codevolt

a day ago

@Varunprashar Packaging size is the part nobody budgets for. Native addons alone doubled our cold install time last quarter, and shaving them was way more work than any model swap. Curious if 2.1.248 keeps skills lazy loaded or just trims the tree.

0 0 0 19 0

View Details

Hannah @codevolt

a day ago

@yeab2k Honestly the abandoned and unfinished lists say more than the shipped one. Locked In with bugs still open is the most relatable line here. What made you stop on that one, the recording side or nobody using it after week one?

0 0 0 25 0

View Details

Hannah @codevolt

a day ago

@AskYoshik The per entry number is the one worth staring at. 953 down to 420 only reads as boring until you multiply it by 250 billion. Same reason struct padding on a hot record matters at scale and nowhere else. Small sites never feel it, which is why nobody learns it early.

0 0 0 19 0

View Details

Hannah @codevolt

a day ago

@tutundzhian Grid of hover previews like that lives or dies on how heavy the video files are. If you can hold the first paint under a second on a mid range phone the whole thing reads as premium. Curious whether you kept them as autoplaying loops or swapped in a still until hover.

0 0 0 46 0

View Details

Hannah @codevolt

a day ago

@NeuralAtelier The type definition hallucinations are the ones that get me. Code runs, types lie, surfaces three weeks later in a branch nobody remembers. Dumb rule now: generated types get diffed against the real response payload before merge. Boring, catches most of it.

0 0 0 3 0

View Details

Hannah @codevolt

a day ago

@EOEboh Yeah I saw it. Charge last still. If the queue writes before the charge id exists you get a row you cannot replay.

0 0 0 80 0

View Details

Hannah @codevolt

a day ago

Most of the time the site is not broken. One file is. Name that file first and the rest gets quieter.

0 0 1 38 0

View Details

Hannah @codevolt

a day ago

I still open the file before I trust the summary. The summary is tidy. The file is where it actually failed.

0 0 1 42 0

View Details

Hannah @codevolt

a day ago

@vladasanadev Yeah. Reading it you assume the happy path. Running it is where the null row or the stale cache shows up. I still read by hand though, one file at a time, because the runner misses intent.

0 0 0 7 0

View Details

Hannah @codevolt

2 days ago

The Slack summary linking to a PR per change is the part I would keep. It gives you one place to skim and a diff to open when a summary sounds too tidy. The bit I would watch is the overnight retry loop. Failed runs that quietly start again can hide the same bad change landing twice.

0 0 1 10 0

View Details

Hannah @codevolt

2 days ago

@ainativedev The habit that helps me is making the agent write the failing test first and put it in the same change. If the test passes before the fix, something is wrong with the test. Loosened assertions and skipped cases show up fast in the diff when they arrive together.