İçeriğe geç

Dra

@bydra.dev did:plc:2wd2tffh2xxplmwmeomv2mra

Building small, focused apps and utilities with a morning coffee & an AI companion 🤖. Calm tools that respect your attention and do what they say on the tin.

Bluesky’a katılma tarihi:

Takipçi
11
Takip edilen
14
Post
24
Sıralamaya girdiği gün
0

Özet

Dra (@bydra.dev), 30 Eylül 2026 tarihinde katılan bir Bluesky kullanıcısı. 11 takipçisi var, 14 hesabı takip ediyor ve 24 post paylaştı.

Sıralama geçmişi

Bu hesabın postlarının Debya günlük listesine girdiği en iyi günler.

Bu hesap henüz günlük sıralamaya girmedi.

Son beğendiği postlar

Bluesky ağındaki herkese açık beğeniler.

  • DY@dyoungai.bsky.social

    That 30-minute wall is brutal for review passes. Routing flash models through OpenCode CLI instead of Claude Code is a clean fix.

    Postu göster
  • ML Systems | Sal Parvez@mlsystems.bsky.social

    Everything off until you turn it on is the right default. Plain files I can read before anything runs is how we're setting things up at ML Systems too.

    Postu göster
  • DY@dyoungai.bsky.social

    Useful tip. Different CLIs change which craft habits transfer and which ones you have to relearn as steer rules (context limits, review depth).

    Postu göster
  • Vladimir Mikhalev@valdemar.ai

    Splitting the double cases fixes the scoring. The other half is a clean PR the review has to let through, run beside the planted ones. In my admission tests every refusal sits next to a signed image that must be admitted and run. A check that blocks everything scores zero.

    Postu göster
  • Vladimir Mikhalev@valdemar.ai

    Publishing 29 of 56 is the useful part: the 27 misses are now a list, not a feeling. When you plant the PRs, make each break exactly one criterion. In my admission CI every refusal has its own image missing exactly one thing, so a block from the wrong rule can't pass as a catch.

    Postu göster

Daha fazlası için üye girişi yap

Blokladığı hesaplar

Bluesky’da bloklar herkese açık kayıtlardır.

Bu hesap kimseyi bloklamıyor.