Sung Kim

Sung Kim

@sungkim.bsky.social

A business analyst at heart who enjoys delving into AI, ML, data engineering, data science, data analytics, and modeling. My views are my own. You can also find me at threads: @sung.kim.mw

7,636 Followers 1,153 Following 6,699 Posts Joined Jan 2024
36 minutes ago

One more thing. If your app has a chat UI then your app is an AI app, not agentic AI app.

2 0 0 0
39 minutes ago

Are moving toward Zero UI for some Agentic AI app? It is appealing.

3 0 3 0
57 minutes ago

Priority Pass in 2025 was iffy, but Priority Pass in 2026 is unusable. It feels like every airport lounge has 1 hour wait.

1 1 0 0
16 hours ago

Better than Claude.

1 0 1 0
16 hours ago

Codex is better at code review. Just my opinion.

12 0 2 0
1 day ago

OK. Temporal is good. Painful, but good so far.

3 0 2 0
1 day ago
Preview
Temporal: The 9-Year Journey to Fix Time in JavaScript JavaScript's Date object has been a source of bugs for three decades. Temporal, which just reached Stage 4, is a modern replacement with immutable types, first-class time zone and calendar support...

Massive 3rd blog post is out: "Temporal: The 9-Year Journey to Fix Time in JavaScript" by @jason-williams.co.uk ->

4 4 1 0
1 day ago

The first six days of excursion costed Pentagon $11.3B.

8 0 0 0
1 day ago
Preview
Introducing Nemotron 3 Super: An Open Hybrid Mamba-Transformer MoE for Agentic Reasoning | NVIDIA Technical Blog Agentic AI systems need models with the specialized depth to solve dense technical problems autonomously. They must excel at reasoning, coding, and long-context analysis…

developer.nvidia.com/blog/introdu...

1 1 1 0
1 day ago
Post image

The keyword is NVFP4 (NVIDIA 4-bit floating point). Only NVIDIA Blackwell or newer GPUs can perform inference on NVFP4 models natively; other GPUs or ASICs must rely on software translation layers.

Nvidia's Nemotron 3 Super: An Open Hybrid Mamba-Transformer MoE for Agentic Reasoning

12 3 4 0
1 day ago
Preview
GitHub - github/copilot-cli-for-beginners: Learn how to get started using the GitHub Copilot CLI! Learn how to get started using the GitHub Copilot CLI! - github/copilot-cli-for-beginners

By the end, you'll confidently use AI to review code, generate tests, debug issues, and automate workflows: all from your terminal.

github.com/github/copil...

3 0 0 0
1 day ago

GitHub Copilot CLI for Beginners

This hands-on course takes you from zero to productive with GitHub Copilot CLI. You'll work with a single Python book collection app throughout all chapters, progressively improving it using AI-assisted workflows.

7 0 2 0
1 day ago
Post image

I love how this works. Cannot login to Claude Code, but status bar is green.

15 1 3 0
2 days ago

Our AI agent workers call server services for most thing and do not do what Claude Code does, so we have no issues.

1 0 0 0
2 days ago
Post image
1 0 1 0
2 days ago
Post image

Does AI agent consume a lot of memory? Yes.

10 0 2 0
2 days ago

tool execution, data processing, and API calls.

…and there are a lot of long-running agent workers.

7 0 1 0
2 days ago

While you are building agentic AI application, it becomes clear that much of the runtime infrastructure will be CPU-driven rather than GPU-driven.

Most agent workers run inside containers and spend the majority of their time on CPU-bound tasks such as orchestration,

20 0 2 0
2 days ago
Image

Say hello to Gemini Embedding 2, our new SOTA multimodal model that lets your bring text, images, video, audio, and docs into the same embedding space! 👀 x.com/OfficialLoga...

58 7 5 1
2 days ago
Post image

Cloudflare introduced a new /crawl endpoint - one API call and an entire site crawled.

No scripts. No browser management. Just the content in HTML, Markdown, or JSON.

developers.cloudflare.com/changelog/po...

64 2 3 1
2 days ago
Post image

Reverse Engineering Binaries With AI by Nicolas Krassas

landaire.net/reverse-engi...

18 0 0 0
2 days ago
Reverse engineering Claude's CVE-2026-2796 exploit

Reverse engineering Claude's CVE-2026-2796 exploit

This blog post will deep dive into how Claude wrote an exploit for CVE-2026-2796 (now patched).

red.anthropic.com/2026/exploit/

8 2 0 0
2 days ago

Yes, testing in general.

0 0 0 0
2 days ago

Building an agentic AI app is really painful. 😭

10 0 3 0
2 days ago

Chinese AI companies need OpenClaw to do well. They need training dataset. 😀

1 0 1 0
2 days ago

😂

1 0 0 0
2 days ago

🤣🤣🤣

Some people are now offering OpenClaw removal services, with remote or in-home options priced at RMB 20 to 299, or about $2.80 to $41.50. Sellers say people are asking mainly over security concerns or because the tool is getting too expensive to use.

23 1 2 0
2 days ago
Preview
Quickstart - Upstash Documentation

This + Claude Max 20X = Your app has been agent-ified. Enjoy your millions from Series A funding!

Upstash Box is a secure, isolated cloud container with an AI Agent built-in (i.e., Claude Code, Codex, and OpenCode)

upstash.com/docs/box/ove...

10 0 2 0
2 days ago
Post image

I wonder who will join them in near future at Meta’s roof?

Alexandr Wang
ManusAI
Moltbook

13 0 0 0
2 days ago

I guess I am moving on to Temporal.

2 0 0 1