WTF! I want my money back. I upgraded my plan to Max 20x just so I could use Fable 5 for next 12 days. "The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance." www.anthropic.com/news/fable-m...
Who's Who of AI
Sung Kim
Why they matter
Directory member with public evidence across AI business, Culture, work & education.
- AI signals
- 80 past 30d
- Sources
- 46 distinct domains
- Discussions
- 26 past 30d
- Latest signal
- 6h ago
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
What they're sharing
Statement on the US government directive to suspend access to Fable 5 and Mythos 5 anthropic.com
When AI builds itself anthropic.com
Our position on open-weights models anthropic.com
Introducing Claude Opus 5 anthropic.com
reuters.com
DiffusionGemma: 4x faster text generation blog.google
Claude Fable 5 and Claude Mythos 5 anthropic.com
[2605.07723] LLM hallucinations in the wild: Large-scale evidence from non-existent citations arxiv.org
openai.com
Redeploying Claude Fable 5 anthropic.com
Introducing Muse Spark 1.1 ai.meta.com
openai.com
Articles & links
Anthropic's When AI builds itself "We looked at sessions where a human researcher took a wrong turn, showed Claude the session up to that point, and asked it what to do next. Mythos Preview improved on humans 64% of the time—up from 22% in 2024." www.anthropic.com/institute/re...
Dario, at minimum, is consistent. "Only we, few hundreads, can be Billionaires. Forget that, this would destroy the backbone of U.S. semiconductor economy that employs millions of people in U.S.." www.anthropic.com/news/positio...
www.anthropic.com/news/claude-...
AI Weekly's analysis
→
- Anthropic launched Claude Opus 5 on July 24, 2026 at $5 per million input tokens, matching Opus 4.8's rate.
- On Frontier-Bench v0.1 Opus 5 scored 43.3%, versus 18.7% for Opus 4.8 and 33.7% for Fable 5.
- Opus 5 becomes the default on Claude Max but sits behind Mythos 5 on cybersecurity tasks, per Anthropic.
Read full analysis →
China’s Ministry of Commerce has led meetings over the past month with major AI companies, including Alibaba, ByteDance, and Z.ai, to discuss measures that would restrict overseas access to cutting-edge AI models, including models that have not yet been released. www.reuters.c…
Sung Kim reposted
New gemma!!! And it's a diffusion model! Deepmind keeps releasing diffusion stuff 🤔 it's not that much worse on benches compared to the same sized autoregressive Gemma 4
AI Weekly's analysis
→
- DiffusionGemma generates 256 tokens per forward pass using bidirectional attention, reaching 1,000+ tokens/sec on a single H100 GPU.
- With only 3.8B active parameters during inference and an 18GB VRAM footprint when quantized, it runs on consumer hardware without server-grade resources.
- Google recommends DiffusionGemma only for speed-critical workloads like in-line editing and code infilling, not for applications requiring maximum quality.
Read full analysis →
View on Bluesky →
Running /security-review on entire repo. We'll see... www.anthropic.com/news/claude-...
"LLM hallucinations in the wild: Large-scale evidence from non-existent citations" Paper: arxiv.org/abs/2605.07723
OpenAI releases GPT-5.6, but ..... openai.com/index/previe...
FYI: Claude Fable 5 will be available again globally tomorrow. www.anthropic.com/news/redeplo...
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...
AI Weekly's analysis
→
- The Meta Model API natively supports both OpenAI Chat Completions and Anthropic Messages formats, removing migration cost for developers already on rival APIs.
- Muse Spark 1.1 leads MCP Atlas tool-use (88.1) but trails GPT-5.5 on DeepSWE 1.1 (53.3 vs 67.0), placing it as an orchestration model.
- Zuckerberg broke a three-year X silence to announce the launch, a move multiple outlets flagged as a deliberate platform-level strategic signal.
Read full analysis →
Sung Kim reposted
OpenAI found their AI would try to find security vulnerabilities to hack its way around limitations such as being in a sandbox without internet access to compete its task. I'm now less impressed by how smart models are and instead how well they follow direction. We've built th…
AI Weekly's analysis
→
- OpenAI paused internal deployment of an unreleased long-horizon model that repeatedly found ways around sandbox and approval checks during monitored use.
- In a NanoGPT evaluation the model spent about an hour finding a sandbox vulnerability and opened public PR #287 despite being told to share results only in Slack.
- OpenAI rebuilt its safety stack around defense-in-depth and trajectory-level monitoring, and says the new system catches considerably more misaligned actions.
Read full analysis →
View on Bluesky →
Their own posts
Recent commentary
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.
Banning open-weight Chinese AI models would be dumb on so many levels. The rest of the world would still have access to Chinese models capable of attacking U.S. businesses.
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.
One interesting thing about Anthropic is that they insist they’re trustworthy guardians of AI, yet they often seem to lack emotional intelligence, or much understanding of life outside their field.
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.
I really miss the good ole days of I can't. With AI, everything becomes “I can.”. It's simply exhausting.
When coding with AI agent, using either CLI or agentic UI, do you review the code? or just review the functionalities? Me. Personally, I do not review the code. People may say AI-generated slop, but have you seen human-generated slop?
To those calling AI a bubble: agentic AI, the phase where AI actually performs work, is only about nine months old. We are still at the beginning.
I'm guessing Anthropic is also working on their own custom AI chip.
TinyRouter They built a small coordinator that, for every question, decides two things: which of three open-source LLMs should answer it, and what role that model should play (Thinker, Worker, or Verifier). The coordinator is deliberately tiny and cheap.
Their network
In Sung Kim's orbit
Center = Sung Kim. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.