# get merge: Monday, November 9, 2026

13 stories merged from 7 newsletters.

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

Web: https://getmerge.ai/2026-11-09

## TLDR

- PSI reports a Lean-checked AI proof of a quantum computing conjecture. Nvidia’s board opened an inquiry into reported foreign intelligence targeting of its CEO, while House chairs requested a briefing. Data-center votes favored limits as OpenAI pledged commitments for future campuses it owns and operates. Waymo will end Uber exclusivity in Austin and Atlanta in 2028.
- MIT’s new robot memory system compares views to correct uncertain object descriptions, while researchers report faster long-context attention with similar accuracy on tested tasks. A social-deduction experiment tests AI bluffing and coordination; Paul Boag suggests using stakeholder prototypes to clarify scope.
- OpenAI introduced a security model for authorized vulnerability testing, while Thinking Machines released an image-capable model through Tinker ahead of a planned open-weights release. Google’s video model now supports 15-second clips and conversational edits, and YouTube Playables added accessibility tags to game information.

## Top Stories

- **PSI reports AI-assisted proof of quantum computing conjecture**. PSI says its research system used a non-public Anthropic model to prove the Aaronson-Ambainis conjecture; the proof was checked by Lean. Sources: The Neuron, The Deep View. [1](https://psi.inc/research/aaronson-ambainis/)
- **Nvidia board opens investigation into reported intelligence targeting**. The board announced an independent inquiry involving CEO Jensen Huang. House committee chairs separately requested a closed-door briefing on November 18. Sources: The Rundown AI. [2](https://nvidianews.nvidia.com/news/statement-from-nvidia-board-of-directors) [3](https://foreignaffairs.house.gov/news/press-releases/chairmen-mast-moolenaar-and-crawford-request-jensen-huang-brief-congress-on-link-to-chinese-intelligence-asset)
- **Data center votes favor limits as OpenAI makes commitments**. Ballotpedia reports 46 of 47 decided measures favored greater limits. OpenAI pledged energy-cost and reporting commitments for future campuses it owns and operates. Sources: The Neuron, The Deep View. [4](https://ballotpedia.org/Data_center-related_ballot_measures,_2026) [5](https://openai.com/index/setting-new-standard-responsible-ai-infrastructure/)
- **Waymo to end Uber exclusivity in two cities**. Waymo plans to offer rides through its own app in Austin and Atlanta from January 2028. Uber says it plans to continue hosting Waymo rides. Sources: Superhuman. [6](https://techcrunch.com/2026/11/04/waymo-confirms-its-shaking-up-its-uber-relationship-in-january-2028/)

## Deeper Cuts

- **MIT system revisits uncertain robot memories**. Researchers compare object descriptions from multiple viewpoints, then select additional views to reduce uncertainty and update a robot's memory. Sources: Superhuman. [7](https://news.mit.edu/2026/clearing-up-robot-fuzzy-memory-1105)
- **Sparse attention paper targets ultra-long model contexts**. Researchers report per-layer speedups from clustering attention keys while maintaining similar accuracy on tested tasks. The paper and implementation are available. Sources: AlphaSignal. [8](https://arxiv.org/abs/2610.42147) [9](https://github.com/Prefrontal-AI/latent-sparse-attention)
- **Social deduction games probe model deception and coordination**. Dwarkesh Patel ran 386 games of Blood on the Clocktower between model teams, examining bluffing, coordination and strategic reasoning within the game. Sources: TLDR AI. [10](https://www.dwarkesh.com/p/clocktower)
- **Stakeholder AI prototypes can clarify expectations**. Paul Boag argues that stakeholder-built prototypes can serve as a starting specification, while their assumptions, accessibility and readiness still need review. Sources: TLDR Design. [11](https://boagworld.com/emails/ai-prototypes/)
- **Designer proposes governed AI skills for evolving design systems**. Drawing on earlier design-system work, Pekka Oskari Puhakka proposes governing an LLM skill as an alternative to relying on components and documentation alone. Sources: TLDR Design. [12](https://uxdesign.cc/state-of-a-system-5c8473db1e9?gi=dd7b91f06e04)

## Tools and Launches

- **OpenAI introduces GPT-6 Cyber**. OpenAI says its security model finds and validates vulnerabilities. Daybreak Red provides deeper authorized testing for verified defenders. Sources: TLDR AI. [13](https://openai.com/index/introducing-gpt-6-cyber/)
- **Inkling 2 Small launches with initially closed weights**. Thinking Machines offers the image-capable model through Tinker and plans an Apache 2.0 weights release after external safety testing. Sources: TLDR AI. [14](https://thinkingmachines.ai/news/inkling-2-small/)
- **Gemini Omni adds 15-second video and conversational editing**. Google is rolling out Omni 1.2 Flash through the Gemini API and Gemini for business. Sources: TLDR AI. [15](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/)
- **YouTube Playables adds game accessibility tags**. Developers can add tags describing supported controls, audio, visual and gameplay features on mobile and web. Sources: TLDR Design. [16](https://ux-news.com/youtube-playables-accessibility-tags/)

## Prompt of the day

One practical prompt for your preferred AI assistant from today's digest. [7](https://news.mit.edu/2026/clearing-up-robot-fuzzy-memory-1105)

Help me resolve an uncertain description or classification in my work. Ask for the item in question, the descriptions or observations I already have, and the decision that depends on it. Compare the accounts in a small table, showing where they agree, conflict or lack detail. Separate direct observations from interpretations, and do not treat repeated wording as independent confirmation. Propose the one or two additional checks or perspectives most likely to distinguish the plausible explanations, using material or tools I already have where possible. Then draft a provisional description that clearly marks what remains uncertain. If I provide new evidence, update the comparison rather than silently overwriting the old account. The result should make it clear what I can safely say now and what would change my conclusion.

## References

1. [psi.inc: A proof of the Aaronson-Ambainis conjecture](https://psi.inc/research/aaronson-ambainis/)
2. [NVIDIA: Statement From NVIDIA Board of Directors](https://nvidianews.nvidia.com/news/statement-from-nvidia-board-of-directors)
3. [Republican Foreign Affairs Committee: Chairmen Mast, Moolenaar, and Crawford Request Jensen Huang Brief Congress on Link to Chinese Intelligence Asset](https://foreignaffairs.house.gov/news/press-releases/chairmen-mast-moolenaar-and-crawford-request-jensen-huang-brief-congress-on-link-to-chinese-intelligence-asset)
4. [Ballotpedia: Data center-related ballot measures, 2026](https://ballotpedia.org/Data_center-related_ballot_measures,_2026)
5. [OpenAI: Setting a new standard for responsible AI infrastructure](https://openai.com/index/setting-new-standard-responsible-ai-infrastructure/)
6. [TechCrunch: Waymo confirms it's shaking up its Uber relationship in January 2028](https://techcrunch.com/2026/11/04/waymo-confirms-its-shaking-up-its-uber-relationship-in-january-2028/)
7. [MIT News: Clearing up a robot’s fuzzy memory](https://news.mit.edu/2026/clearing-up-robot-fuzzy-memory-1105)
8. [arXiv: Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention](https://arxiv.org/abs/2610.42147)
9. [GitHub: Prefrontal-AI/latent-sparse-attention: Code for Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention](https://github.com/Prefrontal-AI/latent-sparse-attention)
10. [dwarkesh.com: The benchmark where AIs have to lie to win](https://www.dwarkesh.com/p/clocktower)
11. [Paul Boag - User Experience Advice: Why Designers Should Welcome AI Prototypes](https://boagworld.com/emails/ai-prototypes/)
12. [UX Collective: State of a system](https://uxdesign.cc/state-of-a-system-5c8473db1e9?gi=dd7b91f06e04)
13. [OpenAI: GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents](https://openai.com/index/introducing-gpt-6-cyber/)
14. [Thinking Machines Lab: Introducing Inkling 2 Small](https://thinkingmachines.ai/news/inkling-2-small/)
15. [Google: Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips](https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/)
16. [UX News: YouTube Playables Adds Accessibility Tags to Game Info](https://ux-news.com/youtube-playables-accessibility-tags/)

Sources today: [The Neuron](https://www.theneurondaily.com/), [The Deep View](https://www.thedeepview.co/), [The Rundown AI](https://www.therundown.ai/), [Superhuman](https://www.superhuman.ai/), [AlphaSignal](https://alphasignal.ai/), [TLDR AI](https://tldr.tech/ai), [TLDR Design](https://tldr.tech/design).
