Meet Daon, the cat that became a global star🐱
Our newest episode of Creator Off Script features Daon_box, the Korean creator behind Daon's stories. Since the beginning of this year, his AI-generated cat adventures have captured hearts around the world, reaching over 20M views.

With Kling AI, Daon Box keeps Daon's character consistent across different scenes, preserves every little detail in native 4K, and shapes each adventure with greater creative control.

Dive into the episode to discover how Daon Box transformed everyday moments with his cat into stories loved by millions.

Today, we are announcing a series of updates that give customers frontier-grade security at half the cost.

MAI-Cyber-1-Flash is our first cybersecurity model, built ground up to find the most challenging vulnerabilities in complex code bases. When combined with MDASH, it delivers world-class performance at 50 percent of the cost of leading models.

We are bringing this capability to market through Project Perception, a complete agentic security offering grounded in real-world signals and security workflows. Teams of specialized agents work together to simulate attacks, detect and triage/investigate, and fix and remediate.

This is the benefit of building the harness, context/signals, and action space separate from one model family. By combining specialized models and data with the right agents, tools, security context, and harness, we can advance the frontier of cost to outcome.

The Kimi K3 architecture figure for yesterday's big open-weight model release, along with some observations and thoughts.

1. Yes, it looks relatively complicated, but it's essentially a scaled-up production version of their Kimi Linear model they released last year (scaled up from 48B -> 2.8T; K3 is by far the biggest open-weight model right now)

2. The one new component compared to Kimi Linear is the LatentMoE. I omitted it in the figure below since it's already very crowded, but that's essentially the same LatentMoE as in Nemotron 3 Ultra (you can find it in my LLM Architecture Gallery if you are curious). The idea here is to compress (down-project) large linear layers similar to multi-head latent attention.

3. Kimi K3's overall trend (similar to Nemotron 3, DeepSeek V4, and others) is also towards better inference efficiency. That is, there are many components that replace existing components with efficiency-tweaked versions. I.e., MoE -> LatentMoE, regular attention -> multi-head latent attention and Kimi Delta Attention. (I also have short tutorials and write-ups in my gallery if you are curious about additional details).

4. The one component change that is not an efficiency tweak is attention residuals. Like DeepSeek V4 improved the residual path with mHC (manifold-constrained Hyper-Connections), attention residuals are a way to improve the residual path, but it works a bit differently. I.e., mHC made the residual path wider. Attention residuals (also already part of Kimi Linear) connect the residuals across layers; the connection itself uses an attention score for an important/contribution weight. According to the report, it improves the validation loss and downstream performance (a bit) consistently and adds about 4% in training cost and 2% in inference cost.

5. Interestingly, Kimi K3 got rid of all RoPE layers and uses NoPE (No Positional Embeddings) everywhere instead. (Again, this is inherited from Kimi Linear). In other architectures, the recent trend was towards RoPE in local attention layers (like sliding window attention) and NoPE in the global layers. There were a few architectures that only used NoPE everywhere, but this is the first frontier-level one as far as I know.

6. Kimi K3 now also has native multimodal support, which is great!

There are several other interesting training tidbits in the technical report, but that's it from the architecture front so far. A really great release overall.

🎙️ Voice AI Night is coming to Mountain View on July 30.

StepFun is teaming up with SEAMATE, PLAUD, Firework, and Cresta to bring together builders across foundation models, AI hardware, video commerce, and enterprise AI.

We’ll dive into what it really takes to build Voice AI—from architecture, latency, barge-in, and memory to the challenges of turning great technology into products people actually use.

📍 Mountain View
📅 July 30
🔗 RSVP Link:
luma.com/seamate-voiceai730

AI21 joined
@nvidia
,
@Microsoft
,
@a16z
, and dozens of others in signing the Open Weights and American AI Leadership letter.

We build agentic systems and agent optimization products on top of models, open and closed alike. The competition between them is what keeps the field honest - on quality, on cost, on who gets access.

Both kinds of models will exist, and both should. Closing off the open path would be a mistake the whole ecosystem pays for.