It’s just funny that I get to know about this sort of Korean policies in HN not from main media (I live in Korea).
I had a few chances to apply for government grants in a startup. There is a lot of blind money for new techs but people managing it are simply not competent enough to understand them. Also, like some comments mentioned, tech infra around their management system is often old and very insecure.
Location: Seoul, South Korea
Remote: Okay. Prefer Onsite
Willing to relocate: No
Technologies: ML, AI, Computer vision, Language model, Agents, Python
Résumé/CV: https://sbinnee.github.io/
Email: sungbin246@gmail.com
---
3 years of experience in start-ups right after PhD. I've followed main waves of ML/AI since 2018. Briefly led a small research team. Had experience traveling and pitching ideas around the globe. I am more into engineering which is the primary reason why I didn't stay in academia. Recent projects cover logical reasoning validation of LLMs and agentic optimization and evals.
Is it just an experimental tool by opencode team? If there is some article about this tool, I would love to read it. It’s not clear to me why I should use this instead of git worktree.
My take is that Mistral is not focusing on generating contents such as code, images, or videos. They focus on multi-lingual models, OCR, voice, and others I believe. Their model intro page manifests that although it always confuses me because it's too colorful and there are too many categories, not to mention model names. I hope their decisions will pay off.
When it comes to MoE, to me, I remember Mixtral model that showed the viability of MoE for the first time. I was impressed by their technical report. To be clear, MoE idea was already out there, if I am not mistaken. If they have pushed Mixtral model family further, who knows they might have achieved the reputation of what the current Qwen family has. A missed opportunity.
I believe that Mistral team is doing the best they can do. I like the directions they push; open models for various tasks, on-prem has a lot of potential. Sure, I use Claude code mostly for coding. But there are so many tasks other than just coding. Even for coding, eventually, I am certain they will catch up and Vibe becomes tolerable soon.
I learned about this plush toy recently. I am often confused with something else I cannot name though. In Korea where I live it’s so common to have key rings that sometimes are these types of plushes. I am not into Labubu but one thing I want to confess is that I like buying special anime related key rings. I restrain myself every time as much as possible. But sometimes it’s just a futile effort. I already own several key rings piled up in my closet.
I can share mine. It was a job interview with one of the fastest growing companies. They were expanding sales positions in APEC region, specifically Korea. I am not really into sales, but I thought okay because it was such a big opportunity to work for this company.
I got three rounds of interviews including technical ones, then I had an interview with my potential team lead. The first thing he asked was about my MBTI personality test, which I hate and didn't pay much attention to learn mine. It seemed every encounter in Korea began with this MBTI test, but common in a job interview? I honestly answered him that I don't know my MBTI and just described my personalities in general. Then he started describing his MBTI and told me that I may not be the best fit with him because this and that.
A few days after, I got an email "... sorry". I don't want to believe that his MBTI question attributed a lot to this decision.
I had a few chances to apply for government grants in a startup. There is a lot of blind money for new techs but people managing it are simply not competent enough to understand them. Also, like some comments mentioned, tech infra around their management system is often old and very insecure.
reply