# get merge: Tuesday, July 28, 2026

21 stories merged from 4 newsletters.

AI & design news, merged daily. One briefing that merges and deduplicates stories from leading AI and design newsletters.

Web: https://getmerge.ai/2026-07-28

## TLDR

- Moonshot released Kimi K3 (2.8T params) as the largest open-weight model ever, while SSI secured a $5B NVIDIA deal and the Open Secure AI Alliance launched without OpenAI, Google, or Anthropic.
- Pydantic's "human-in-the-loop is tired" essay explores the psychology of AI coding, Microsoft's MAI-Cyber-1-Flash redefines security economics, and shared Claude chats were exposed in Google search results.
- Kimi K3 weights, FlashKDA attention kernel, OptMem infinite memory, MAI-Cyber-1-Flash, and Meta AI in Threads DMs lead this week's launches.

## Top Stories

- **Moonshot drops Kimi K3 weights**. The 2.8T-parameter open-weight model is the largest ever released, competing with Claude Fable 5 and GPT-5.6 on benchmarks while carrying a restrictive license for companies over $20M revenue. Moonshot also open-sourced FlashKDA attention kernels and AgentEnv training infrastructure. Sources: AlphaSignal, The Neuron, The Rundown AI. [1](https://huggingface.co/moonshotai/Kimi-K3)
- **SSI breaks silence with $5B NVIDIA deal**. Ilya Sutskever's Safe Superintelligence emerged from two years of quiet research, securing a multi-billion NVIDIA investment and access to the Vera Rubin GPU platform for a 10x compute upgrade within 12 months. Sources: AlphaSignal, The Rundown AI. [2](https://x.com/ssi/status/2081732119194394763?s=20)
- **NVIDIA launches Open Secure AI Alliance**. A coalition of 36 organizations (Microsoft, IBM, Cisco, CrowdStrike, Hugging Face, Palantir) to build open security tools for AI agents. The notable absences: OpenAI, Google, Anthropic, Amazon, and Meta. Sources: The Neuron. [3](https://blogs.nvidia.com/blog/open-secure-ai-alliance/)
- **Anthropic clarifies open-weights stance**. Dario Amodei published that Anthropic has never advocated for a ban on open-weights models, instead pushing for chip controls and distillation crackdowns, while declining to sign the Nvidia-led open-weights letter. Sources: The Rundown AI. [4](https://www.anthropic.com/news/position-open-weights-models)

## Deeper Cuts

- **The human-in-the-loop is tired**. Pydantic's essay on reward functions, dopamine, and what it feels like when the code starts writing itself. A meditation on the psychological shift from writing code to guiding code. Sources: Sidebar. [5](https://pydantic.dev/articles/the-human-in-the-loop-is-tired)
- **Claude shared chats leaked in Google**. A Reddit user discovered shared Claude conversations and Artifacts appearing in search results, exposing legal questions, personal info, and vibe-coded app data. Sources: The Neuron, The Rundown AI, AlphaSignal. [6](https://www.404media.co/tons-of-peoples-claude-chats-and-creations-are-exposed-on-google/)
- **Microsoft launches MAI-Cyber-1-Flash**. A cybersecurity model hitting 96% on CyberGym benchmark at half the cost of top rivals, integrated into the MDASH agent system for automated vulnerability hunting. Sources: The Rundown AI. [7](https://microsoft.ai/news/introducing-mai-cyber-1-flash-inside-mdash/)
- **OpenAI study: AI is dissolving job titles**. 43.5% of occupation-specific ChatGPT messages involved work normally associated with a different occupation. Marketers analyze data, operators write code, managers moonlight as researchers. Sources: The Neuron. [8](https://openai.com/index/how-ai-is-expanding-what-people-do-at-work/)
- **Cursor's planner-worker swarm pattern**. A frontier model plans the work, cheaper models execute subtasks, and a shared field guide prevents repeated mistakes. The SQLite-in-Rust experiment showed significant cost savings. Sources: The Neuron, The Rundown AI. [9](https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work/)
- **Drawing the Mona Lisa with four frontier models**. A side-by-side tracking of every stroke, dollar, and output as GPT-5.6, Claude, Gemini, and Grok tried to draw the Mona Lisa with colored pencils. Sources: Sidebar. [10](https://www.tryai.dev/blog/ai-drawing-arena-colored-pencils-claude-gpt-grok)
- **EU AI Omnibus enters force**. Extending compliance timelines and expanding regulatory sandboxes across the bloc. Sources: The Neuron. [11](https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force)
- **METR's expenditure-horizon metric**. A new framework for judging when AI agents actually become cheaper than humans, factoring in supervision and retry costs. Sources: The Neuron. [12](https://the-decoder.com/metr-introduces-a-new-metric-to-calculate-exactly-when-ai-agents-become-more-expensive-than-humans/)

## Tools and Launches

- **Kimi K3**. 2.8T-parameter open-weight model, native multimodal, 1M token context. Sources: AlphaSignal, The Neuron, The Rundown AI. [1](https://huggingface.co/moonshotai/Kimi-K3)
- **FlashKDA**. Open-source attention kernel cutting prefill time by up to 2.2x. Sources: AlphaSignal. [13](https://github.com/MoonshotAI/FlashKDA)
- **OptMem**. Plug-and-play infinite memory prompt for any AI agent, using compressed summary trees. Sources: AlphaSignal. [14](https://github.com/VictorTaelin/OptMem)
- **MAI-Cyber-1-Flash**. Microsoft's cybersecurity AI for automated bug hunting at half the cost. Sources: The Rundown AI. [7](https://microsoft.ai/news/introducing-mai-cyber-1-flash-inside-mdash/)
- **AgentEnv**. The distributed agent training system behind Kimi K3, open-sourced. Sources: AlphaSignal. [15](https://github.com/kvcache-ai/AgentEnv)
- **Meta AI in Threads DMs**. Chat with Meta AI about posts, images, and links without leaving the app. Sources: The Neuron. [16](https://techcrunch.com/2026/07/27/threads-users-can-now-chat-with-meta-ai-in-their-dms/)
- **jcodemunch MCP**. Open-source MCP server cutting AI coding token costs by 95%. Sources: AlphaSignal. [17](https://github.com/jgravelle/jcodemunch-mcp)
- **FLUX 3**. Black Forest Labs' multimodal AI with 20-second generations. Sources: The Rundown AI. [18](http://www.rundown.ai/tools/flux-3)
- **Buzz**. Open-source workspace where AI agents join team chats. Sources: The Rundown AI. [19](https://www.rundown.ai/tools/buzz)

## References

1. [Hugging Face: Moonshot AI / Kimi-K3](https://huggingface.co/moonshotai/Kimi-K3)
2. [SSI on X: Partnership with NVIDIA](https://x.com/ssi/status/2081732119194394763?s=20)
3. [NVIDIA Blog: Open Secure AI Alliance](https://blogs.nvidia.com/blog/open-secure-ai-alliance/)
4. [Anthropic: Position on Open-Weight Models](https://www.anthropic.com/news/position-open-weights-models)
5. [Pydantic: The Human-in-the-Loop is Tired](https://pydantic.dev/articles/the-human-in-the-loop-is-tired)
6. [404 Media: Claude Chats Exposed on Google](https://www.404media.co/tons-of-peoples-claude-chats-and-creations-are-exposed-on-google/)
7. [Microsoft AI: MAI-Cyber-1-Flash](https://microsoft.ai/news/introducing-mai-cyber-1-flash-inside-mdash/)
8. [OpenAI: How AI is Expanding What People Do at Work](https://openai.com/index/how-ai-is-expanding-what-people-do-at-work/)
9. [The Decoder: Cursor's Agent Swarm](https://the-decoder.com/cursors-agent-swarm-suggests-cheaper-models-can-handle-most-coding-when-frontier-models-plan-the-work/)
10. [TryAI: Drawing the Mona Lisa with AI](https://www.tryai.dev/blog/ai-drawing-arena-colored-pencils-claude-gpt-grok)
11. [European Commission: AI Omnibus Enters Force](https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force)
12. [The Decoder: METR Expenditure-Horizon Metric](https://the-decoder.com/metr-introduces-a-new-metric-to-calculate-exactly-when-ai-agents-become-more-expensive-than-humans/)
13. [GitHub: MoonshotAI / FlashKDA](https://github.com/MoonshotAI/FlashKDA)
14. [GitHub: VictorTaelin / OptMem](https://github.com/VictorTaelin/OptMem)
15. [GitHub: kvcache-ai / AgentEnv](https://github.com/kvcache-ai/AgentEnv)
16. [TechCrunch: Meta AI in Threads DMs](https://techcrunch.com/2026/07/27/threads-users-can-now-chat-with-meta-ai-in-their-dms/)
17. [GitHub: jgravelle / jcodemunch-mcp](https://github.com/jgravelle/jcodemunch-mcp)
18. [Rundown: FLUX 3](http://www.rundown.ai/tools/flux-3)
19. [Rundown: Buzz](https://www.rundown.ai/tools/buzz)

Sources today: [AlphaSignal](https://alphasignal.ai/), [The Neuron](https://www.theneurondaily.com/), [The Rundown AI](https://www.therundown.ai/), [Sidebar](https://sidebar.io/).