Emilien エミリアン

@EmilienKopp

Professional Human Captcha - Code Architect

Commit activity

63 commits in the last year

longest streak 4 days

1,502 / 2,500 XP998 XP to root

Sat, Nov 7

Alright this is coming together very very nicely! https://std-out.sh is the stage for your dev talks and meetups. Edit deck in markdown or good old WYSIWYG, run the deck-linter, insert polls with live results, get realtime reactions and comments... and even rehearsals and deck…
std-out.shstd-out.sh - The standard output for your talksA talk is a deploy, and std-out is its toolchain: write the deck, dry-run it with timed rehearsals, put it through code review with the speakers you follow, then ship it live to a room that streams back.

Tue, Nov 3

Fri, Oct 30

Wed, Oct 28

Tue, Oct 27

Fri, Oct 23

Thu, Oct 22

My recent philosophy: instead of "making AI do the repetitive tasks" I try to have AI build a deterministic tool to do the task, and leave AI the least responsibility possible. Parsing files and collecting data: script. Giving me insight on the data: AI. Not AI from the top…
Working on an automation tool myself, and based on many many benchmarks & many tokens burned after, I realized it's better to leave the least decision making responsibility to them and take care of what can be done deterministically.
I like this approach. And keep the build process iterative: ship a first version of the script then improve it when you hit new edge cases.
Yes, I've found the better Claude models are pretty good at taking this approach automatically - they will reach for a tool or build one to undertake most repetitive tasks where a deterministic outcome is possible. But then again, I work almost exclusively inside Claude Code -…

Wed, Oct 21