Sigrid Jin | Jin Hyung Park 🪄 👧

@sigridjin_eth

✯ @thisissigrid ★ ☄  CS @UBC ☄ ★ Ultrathink Engineer @sionic_ai 🐟 digital nomad 💻

Social Media Influencer

The Earth

github.com/sigridjineth

7월 2019에 가입

12천게시물 2천팔로워 8천팔로우 중

내가 좋아할 만한 콘텐츠

@bambben

@Hyunxukee

@heysoahri

@MoneyStack9

@decentra1ized_

@gohameye

@dev_chooble

@fkemgod

@subinium

@DecipherGlobal

@0xboosik

@JonghyunChun

@JuhyukB

@web3vibe

@Crypnovice88

고정된 트윗

Sigrid Jin | Jin Hyung Park 🪄 👧

@sigridjin_eth

2024. 5. 19.

Read this tweet every day

Andrew D. Huberman, Ph.D.

@hubermanlab

2023. 12. 13.

What many people perceive as a focus problem is often a thoroughness problem. One route to improving focus is to *direct your efforts to thoroughness without the expectation of focus arriving first*. Put simply, thoroughness improves focus, no matter how slow the process may be.

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

AstraKernel 💫

@AstraKernel

. 10. 31.

This is a side effect of Rust 🙂

$tanya1866's profile picture. i code in rust. building @fractylxyz$

Tanya

@tanya1866

. 10. 30.

unfortunately, learning rust ruined almost every other language for me

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

Armen Aghajanyan

@ArmenAgha

. 10. 31.

jackass leaking company IP at 2AM

Maciej Kilian

@kilian_maciej

. 10. 31.

remember: do not re-normalize MoE router scores post topk if k=1 weighing MoE outputs by the router scores is how the task loss can influence router params. when k=1 score=sum(score) so the update makes the scores=1 meaning it doesn't get gradients since its a constant!

kilian_maciej's tweet image. remember: do not re-normalize MoE router scores post topk if k=1

weighing MoE outputs by the router scores is how the task loss can influence router params. when k=1
score=sum(score) so the update makes the scores=1 meaning it doesn't get gradients since its a constant!

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

Haze

@Nebu1eto

. 10. 30.

Kafka는 훌륭한 도구지만 PostgreSQL로도 어지간한 지점까지 MQ를 쓸 수 있다는 글. OpenAI조차 아직도 단일 PostgreSQL 쓰기 인스턴스로 운영하고 있고, Figma도 2022년까지 샤딩 없는 PostgreSQL로 서비스를 운영했다. 실행 가능한 최소한의 인프라에 굉장히 동의한다. topicpartition.io/blog/postgres-…

Nebu1eto's tweet card. Why you should just use Postgres instead of Kafka for small-scale message queuing and pub-sub patterns. Benchmarks and practical tests included.

Kafka is fast -- I'll use Postgres

출처: topicpartition.io

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

Pak Youngrok

@pakyoungrok

. 10. 31.

어, 나도 스타트업들 컨설팅 및 기술상담하면서 카프카 쓰지 말라는 이야기 정말 많이 했는데. 아니 그냥 DB에 들어가야 하는 데이터인데 write 비용 크다고 카프카 거쳐서 넣는 사람이 왜 이렇게 많은지. 제발 그냥 RDBMS로 다 하세요. 도저히 비용효율이 안 나온다 싶을 때 카프카 가도 늦지 않아요.

Haze

@Nebu1eto

. 10. 30.

Kafka is fast -- I'll use Postgres

출처: topicpartition.io

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

marcus

@mjovanc

. 10. 27.

Come one Rust people, let's help the man be free! 🦀

이 트윗은 더 이상 사용할 수 없습니다.

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

Omar Khattab

@lateinteraction

. 10. 27.

Guido van Rossum builds a python package for RAG

Pamela Fox

@pamelafox

. 10. 21.

At #PyBay25, @gvanrossum demo'd a Python package for "structured RAG". During ingestion, it uses LLM to extract structured data (entities/topics/verbs) and stores in standard DB, and then retrieves by structuring the user query as well. Try it out at: github.com/microsoft/type…

pamelafox's tweet image. At #PyBay25, @gvanrossum demo'd a Python package for "structured RAG".
During ingestion, it uses LLM to extract structured data (entities/topics/verbs) and stores in standard DB, and then retrieves by structuring the user query as well.
Try it out at:
github.com/microsoft/type…

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

The Spectator Index

@spectatorindex

. 10. 22.

Household debt as share of GDP. 🇨🇦 Canada: 103% 🇬🇧 UK: 80% 🇺🇸 US: 73% 🇫🇷 France: 63% 🇨🇳 China: 62% 🇩🇪 Germany: 52% 🇪🇸 Spain: 48% 🇮🇹 Italy: 39% 🇮🇳 India: 37% 🇿🇦 South Africa: 34% 🇧🇷 Brazil: 34% 🇸🇦 Saudi: 32% 🇷🇺 Russia: 22% 🇮🇩 Indonesia: 16% 🇲🇽 Mexico: 16% 🇹🇷 Turkey: 11%

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

asbubam

@asbubam

. 10. 20.

11월부터 연락주셨던 분들 만나서 이야기 나누려고 약속을 잡고 있어요. 세컨비 만들면서 DevOps, SRE 포지션이 아닌, 프러덕트 엔지니어에 도전하고 싶다는 생각이 들었어요. 내가 지금 일하고 있는 팀이 너무 좋은 팀이라고 생각되는 분 계시면 알려주세요! github.com/asbubam/resume

asbubam's tweet card. Resume of Seungwoo Lee. Contribute to asbubam/resume development by creating an account on GitHub.

GitHub - asbubam/resume: Resume of Seungwoo Lee

출처: github.com

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

Crystal

@crystalsssup

. 10. 20.

Kimi K2 is up to 5x faster and 50% more accurate ：）

Guillermo Rauch

@rauchg

. 10. 18.

We're benchmarking a bunch of models for one of our internal agents. 🤯 Kimi K2 is up to 5x faster and 50% more accurate than frontier proprietary models… And it took very little work to switch out the models via @aisdk / @vercel ai gateway 😬 (served by @groqinc)

rauchg's tweet image. We're benchmarking a bunch of models for one of our internal agents. 🤯 Kimi K2 is up to 5x faster and 50% more accurate than frontier proprietary models…

And it took very little work to switch out the models via @aisdk / @vercel ai gateway 😬 (served by @groqinc)

Sigrid Jin | Jin Hyung Park 🪄 👧 님이 재게시함

jason liu

@jxnlco

. 10. 18.

Why do frontier models still fail at PDF parsing in 2025? Evan Vogelbaum, @hu_yifei and @AlvinRyanputra from @reductoai showed their autoregressive layout model: outputs hundreds of boxes with zero hallucinations. Traditional OCR for simple text. Custom VLMs for handwritten…