Sung Kim
@sungkim
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
It seems I personally owe $1.9T to AWS. Should I setup a GoFundMe page, so I can pay this bill. health.aws.amazon.com/health/status
Windows 11 running in a browser, developed using Kim K3 agent swarms by chetaslua windowos.kimi.page
This is on point for China, which has always seen itself as the center of the world, except during the “Century of Humiliation.” China’s message is essentially: We will keep our AI models open-weight and work with other countries to advance AI development.
macOS 27, running in a browser, developed using Kim K3 agent swarms in 4 hours, by Max Weinbach. macos27.kimi.page
🔹 Built for long-horizon agentic coding and self-evolving workflows Tech blog: kimi.com/blog/kimi-k3
Kimi K3 (Open Weights by July 27, 2026.) 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
I predict someone will weaponize this. An air-to-air kill of a flying moth by an autonomous micro-drone. tornyol.com
SaaSOCALYPSE seems to be happening... This is great news for software developers. Instead of customizing an off-the-shelf solution, they can finally build an application tailored to their organization’s needs.
Single-rollout Asynchronous Optimization (SAO), is able to train stably for one thousand steps and consistently outperform GRPO and its variants on agentic coding and reasoning benchmarks, such as SWE-Bench Verified, BeyondAIME, and IMOAnswerBench.
The Little Book of Reinforcement Learning by Alexandre Torres-Leguet github.com/alxndrTL/lit...
Google's Open Knowledge Format (bit like llm.txt, i guess...) An open specification that formalizes the LLM-wiki pattern into a portable, interoperable format. cloud.google.com/blog/product...
Tibo Sottiaux of OpenAI: Working on Codex must have been a stressful experience.
LLMs can't jump. The thought experiment is this: Take an LLM with a 1905 knowledge cutoff. Feed it every paper, every dataset, every equation of that era. Could it invent general relativity? No.
Meta's muse spark 1.1 is an industry-competitive agentic and coding model. across many agentic evals it rivals gpt-5.5 and opus-4.8. available now through the new meta model api and in meta ai. ai.meta.com/blog/introdu...
A Brown professor gave his students a take-home midterm exam. After suspecting many cheated using AI, he made the final in-person. The orange dots are the midterm scores and the gray dots are the final scores. I applaud S22 for honesty because I would've cheated.
Benchmarking Coding Agents on Databricks’ Multi-Million Line Codebase They ran a comprehensive evaluation on our tasks, code base, infra. It's been produced by more than 3,000 software engineers, spans 3 hyperscalar clouds and many languages and tasks. www.databricks.com/blog/benchma...
Goodfire AI's Block-Sparse Featurizers (BSFs), a new way to find concepts in model activations - using multidimensional “blocks” instead of single directions. www.goodfire.ai/research/bsf...
How to Build a Diffusion Language Model by Volodymyr Kuleshov "An introduction to diffusion language models and the research advances that underlie today's diffusion LLMs. We describe the building blocks of recent open-source models, starting from simple masking diffusion,