Mati Hong

⚡Tech PO & Developer, Software Educator, 🌏Lover #RAG #LLMs #Flutter #Elixir #DevOps

[Mati Hong @MartinHong17]

cmux 쓰다가 파일보고 싶다고 vscode 열면, 나만 정신없어? 그냥 터미널에서 딱 필요한 세팅만 vscode단축키로 쉽게 했어 https://t.co/sG1uoOjF35 금방 커스터마이징 가능하니 각자 더 잘해봐. cmux만들어줘서 고마워 @austinywang

[Mati Hong @MartinHong17]

I am hosting 31회차 AI로 사업강화 : 당근모임 확장. Join me! https://t.co/MM5T82Yz1c 출처 @LumaHQ

[Mati Hong @MartinHong17]

TalkMode — a real assistant on your Mac: On-device AI voice assistant for your Mac https://t.co/z1dWrv0uzt via @producthunt

[Andrej Karpathy @karpathy]

Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights:

The first theme I tried to push on is that LLMs are about a lot more than just speeding up what existed before (e.g. coding). Three examples of new horizons:

1. menugen: an app that can be fully engulfed by LLMs, with no classical code needed: input an image, output an image and an LLM can natively do the thing.
2. install .md skills instead of install .sh scripts. Why create a complex Software 1.0 bash script for e.g. installing a piece of software if you can write the installation out in words and say "just show this to your LLM". The LLM is an advanced interpreter of English and can intelligently target installation to your setup, debug everything inline, etc.
3. LLM knowledge bases as an example of something that was *impossible* with classical code because it's computation over unstructured data (knowledge) from arbitrary sources and in arbitrary formats, including simply text articles etc.

I pushed on these because in every new paradigm change, the obvious things are always in the realm of speeding up or somehow improving what existed, but here we have examples of functionality that either suddenly perhaps shouldn't even exist (1,2), or was fundamentally not possible before (3).

The second (ongoing) theme is trying to explain the pattern of jaggedness in LLMs. How it can be true that a single artifact will simultaneously 1) coherently refactor a 100,000-line code base *and* 2) tell you to walk to the car wash to wash your car. I previously wrote about the source of this as having to do with verifiability of a domain, here I expand on this as having to also do with economics because revenue/TAM dictates what the frontier labs choose to package into training data distributions during RL. You're either in the data distribution (on the rails of the RL circuits) and flying or you're off-roading in the jungle with a machete, in relative terms. Still not 100% satisfied with this, but it's an ongoing struggle to build an accurate model of LLM capabilities if you wish to practically take advantage of their power while avoiding their pitfalls, which brings me to...

Last theme is the agent-native economy. The decomposition of products and services into sensors, actuators and logic (split up across all of 1.0/2.0/3.0 computing paradigms), how we can make information maximally legible to LLMs, some words on the quickly emerging agentic engineering and its skill set, related hiring practices, etc., possibly even hints/dreams of fully neural computing handling the vast majority of computation with some help from (classical) CPU coprocessors.

[きしだൠ(K1S) @kis]

mlx-vlm
https://t.co/jeDx4R7BHj

[How To AI @HowToAI_]

The entire RAG industry is about to get cooked.

Researchers have built a new RAG approach that:

- does not need a vector DB.
- does not embed data.
- involves no chunking.
- performs no similarity search.

It's called PageIndex. Instead of chunking your docs and stuffing them into pinecone, it builds a tree index and lets the LLM reason through it like a human reading a book.

hit 98.7% on financebench. beats every vector RAG on the leaderboard.

no embeddings. no chunking. no vector DB.

100% open source.

[Mati Hong @MartinHong17]

vibecamp ai horizon #2 · 2026-05-06

• 美 CAISI, 구글·MS·xAI와 프런티어 모델 사전 검수 협약 — 출시 전 정부 평가
• 한국, 국가성장펀드 5.7조 AI 투자 — 삼성·SK하이닉스 'AI 주권' 선언 #AIhorizon #vibecamp #AI https://t.co/UbKSUuMacd

[そう｜Claude Codeで始めるAI自動化 @so_ainsight]

無料なのが信じられないオープンソースのAIツール10選：

1. Fooocus
Midjourneyの代替。自分のノートPCで動いて、3クリックで初画像まで到達できる。
https://t.co/GpU6nzoxX5

2. ComfyUI
プロのAIアーティストが本番制作で使う画像生成ツール。スタジオでも採用されている本気仕様。
https://t.co/uDdXEinLVB

3. Ollama
ChatGPTを自分のマシンで動かせる。ネット不要、API課金ゼロ、コマンド1つで導入完了。
https://t.co/wvdidJsTO9

4. OpenVoice
10秒の音声からどんな声でも複製できる。MITとMyShell製、ライセンスは商用利用OK。
https://t.co/y7cruci2Zv

5. Penpot
Figmaの代替。リアルタイム共同編集つき、ブラウザで動いて、自前ホストも可能。
https://t.co/FAH6DlULfj

6. AppFlowy
Notionの代替で、データは全部自分のマシンに残る。Rust製のローカルファースト設計。
https://t.co/B9Iopyikr8

7. n8n
Zapierの代替。1600以上の連携と専用AIノードを搭載。自前ホストならワークフロー無制限。
https://t.co/pvJGzLdiyo

8. Cal(.)com
Calendlyの代替。有料版より機能が多く、クライアント向けにホワイトラベル化もできる。
https://t.co/G67KBojy2o

9. Supabase
Firebaseの代替で、しかも自分で所有できる。Postgres・認証・ストレージ・Realtime・Edge Functionsが全部入り。
https://t.co/UWh80N52tL

10. Cline
VS Code内で動くCursorの代替。ファイルを読み、コードを編集し、機能を完成まで運んでくれる。
https://t.co/A1AQmzRgB2

SaaSで全部揃えたら月300〜500ドル（約4.8〜8万円）かかるレベル。それが全部、無料で永久に使える。

保存して後で使い倒して。

[Alex Imas @alexolegimas]

Excellent essay from @danbjork on the implications of AI for low and middle income countries—a critical topic that deserves way more attention.

https://t.co/cyclCQwsAU

[Mati Hong @MartinHong17]

📬 vibecamp 뉴스레터 #1
▸ 🛠 트렌드 (3건)
· GitHub Copilot, 6/1부터 토큰 기반 'AI Credits' 과금으로 전환
... https://t.co/F8LKjk6ghs
#AI #Newsletter #VibeCamp

[Sebastian Raschka @rasbt]

April was a pretty strong month for LLM releases:
- Gemma 4
- GLM-5.1
- Qwen3.6
- Kimi K2.6
- DeepSeek V4

All are now added to the LLM Architecture Gallery.
More details once I am fully back in May!

[mizutani @m_mizutani]

最近の開発環境、ついにほとんどエディタを開かなくなり、Claude Code + cmux および 仕様mardown確認用のmo ( https://t.co/f8eymAh2Vj ) とコード差分確認用のdifit ( https://t.co/JB8ISayBN7 ) で完結するようになってきてしまった

[Mati Hong @MartinHong17]

@m_mizutani 나도 비슷하게 md preivew가 필요하더라고 https://t.co/v2tHDneDs8

[Mati Hong @MartinHong17]

cmux 쓰다가 파일보고 싶다고 vscode 열면, 나만 정신없어? 그냥 터미널에서 딱 필요한 세팅만 vscode단축키로 쉽게 했어 https://t.co/sG1uoOjF35 금방 커스터마이징 가능하니 각자 더 잘해봐. cmux만들어줘서 고마워 @austinywang

[stevibe @stevibe]

Qwen3.6-27B on RTX 5090, 4 power limits tested:

🔴 400W → 66.58 t/s · baseline (but fluctuates a lot — frequent dips)
🟢 450W → 69.79 t/s · +4.8% speed for +12.5% power (much more stable)
🟡 500W → 71.48 t/s · +2.4% speed for +11.1% power
🟣 575W → 72.64 t/s · +1.6% speed for +15.0% power

➡️ 400W → 575W: +44% power, +9% speed.

Conclusion: 450W is the real sweet spot. 400W looks great on the average but the t/s curve is jittery; 450W trades 50W for consistent throughput. Above that, you're just heating your room.

[そう｜Claude Codeで始めるAI自動化 @so_ainsight]

寝てる間も稼ぎ続けるGitHubリポジトリ10選：

1. AutoHedge
AIエージェントが市場分析からトレード実行まで全自動。数分で自分専用の自律型ヘッジファンドが作れる。
https://t.co/o0w4RFVdQp

2. Vibe-Trading
自然言語で取引戦略を指示できる個人用AIトレーディングエージェント。29の専門チームが分析から実行まで動く。
https://t.co/sNrgblv8e2

3. Claude Ads
Google・Meta・YouTube・TikTokなど7媒体の広告を250以上の観点でAI監査。広告費のムダを最大35%削減できる。
https://t.co/vG0kAnudwD

4. Toprank
Search ConsoleとGoogle Adsに直接アクセスし、トラフィック低下の原因を診断して修正案まで出してくれるSEOツール。
https://t.co/47PcFuXsfF

5. Fincept Terminal
Bloombergのオープンソース代替。37のAIエージェントが市場分析・投資調査を無料で担ってくれる。
https://t.co/m8rRo1vvQc

6. Agentic Inbox
Cloudflare Workers上で動く自己ホスト型AIメールエージェント。受信→分類→下書き作成まで自動。送信前に確認できる。
https://t.co/nd8rFqQGm9

7. ClawRouter
41以上のLLMモデルを1ms以下でルーティング。AIのAPI費用を最大92%削減できる。
https://t.co/hAZGZl1ZAE

8. Camofox Browser
AIエージェント向けのアンチ検出ブラウザ。Webスクレイピングや自動化でブロックされなくなる。
https://t.co/MYnUuRaOWC

9. Open Higgsfield AI
Higgsfield・Krea・Freepik AIなど複数サービスの無料オープンソース代替。200以上のモデルでフィルターなしの画像・動画生成が使える。
https://t.co/xaiWWxrIho

10. Hyperframes
HTMLを書くだけで動画をレンダリングできるフレームワーク。エージェントとの連携に最適。
https://t.co/vSJhdJX4df

保存して後で使いこなして。

[Mati Hong @MartinHong17]

소프트웨어를 공격하는건 쉬워도,
사람이 의도한 대로 소프트웨어를 만드는 건 어렵다

[José Valim @josevalim]

At this point, this is just irresponsible.

Yes, coding agents are leading to an increase of software production, but we are not seeing a similar push or increase in software quality.

If Anthropic focuses on safety and it believes software engineering is going away, then it needs to be doing much more to improve how we design, build, test, and maintain software (aka software engineering). Increasing the production of unreliable, poorly designed, and unverified software directly undermines safety.

Claude Code is claimed to be "fully written by AI". In the last two months, it took three separate postmortem-worthy failures and user complaints to surface what their own testing missed. Yesterday users were being over billed by hundreds of dollars. Software engineering isn't ready to go away and there is not enough progress to argue that case.

I am certain Anthropic would argue that AI progress in other domains is strongly dependent on having proper safeguards in place. I can't wrap my head around the cognitive dissonance when it comes to software.

PS: Mythos (may) improve software security, but that is only a subset of safety.

[José Valim @josevalim]

At this point, this is just irresponsible.

Yes, coding agents are leading to an increase of software production, but we are not seeing a similar push or increase in software quality.

If Anthropic focuses on safety and it believes software engineering is going away, then it needs to be doing much more to improve how we design, build, test, and maintain software (aka software engineering). Increasing the production of unreliable, poorly designed, and unverified software directly undermines safety.

Claude Code is claimed to be "fully written by AI". In the last two months, it took three separate postmortem-worthy failures and user complaints to surface what their own testing missed. Yesterday users were being over billed by hundreds of dollars. Software engineering isn't ready to go away and there is not enough progress to argue that case.

I am certain Anthropic would argue that AI progress in other domains is strongly dependent on having proper safeguards in place. I can't wrap my head around the cognitive dissonance when it comes to software.

PS: Mythos (may) improve software security, but that is only a subset of safety.

[Fraser Cottrell | Ad Creative For Meta & TikTok @sourfraser]

https://t.co/o5osBJ933J

[sysls @systematicls]

https://t.co/wBaKpAI5Vl

[superwhisper @superwhisper]

Founding member of OpenAI and former Director of AI at Tesla, Andrej Karpathy (@karpathy) shows how he uses Superwhisper ✨