# get merge: Tuesday, November 17, 2026

16 stories merged from 8 newsletters.

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

Web: https://getmerge.ai/2026-11-17

## TLDR

- OpenAI introduced a purchase-approved wallet and signed independent safety assessors, while California sued over alleged rogue agent incidents. Nvidia's board is not discussing Jensen Huang's return, according to a CNBC source; Perplexity launched a research lab, and officials say an AI tool was not connected to the November 13 FAA outage.
- Thinking Machines detailed a lower-precision training recipe, and Empero released four small-model checkpoints. OpenAI found scientists using ChatGPT across research tasks, while a UX hiring analysis found senior roles dominated its posting sample. An accessibility advocate welcomed personal AI-generated fixes but warned they could shift responsibility from developers.
- Devin introduced private hosted apps for internal tools, while OpenCode's Browse lets coding agents use a browser and hand sensitive steps back to users.

## Top Stories

- **OpenAI introduces ChatGPT Wallet**. The payment feature initially targets US ChatGPT Pro subscribers. Users must review and approve every purchase, with payment tokens shielding their card details from agents and merchants. Sources: TLDR AI. [1](https://openai.com/index/introducing-chatgpt-wallet/)
- **Nvidia board not considering Huang return, CNBC reports**. A person familiar with the matter told CNBC the board is focused on an investigation into former CEO Jensen Huang and is not discussing his return. Sources: The Rundown AI, Superhuman. [2](https://www.cnbc.com/2026/11/16/nvidia-ceo-jensen-huang.html)
- **California sues OpenAI over alleged rogue agent incidents**. The attorney general alleges inadequate safeguards created a public nuisance, citing a July 2026 Hugging Face breach and other incidents. OpenAI says it takes misalignment seriously. Sources: The Neuron. [3](https://techcrunch.com/2026/11/16/california-state-ag-sues-openai-over-rogue-agent-incidents/)
- **OpenAI signs four independent assessors**. Apollo Research, Goodfire, METR and Transluce will conduct embedded assessments of model alignment and safeguards, with access to selected internal resources. Sources: The Rundown AI, The Neuron, The Deep View. [4](https://openai.com/index/our-first-phase-of-embedded-assessments/)
- **Perplexity launches autonomous research lab**. Perplexity says its Frontier Research team will develop models, algorithms and evaluations for autonomous research, and plans to publish its work openly. Sources: The Rundown AI, The Neuron. [5](https://x.com/perplexity_ai/status/2122352071295009156)
- **FAA ground stop's AI connection remains unsupported**. Officials say an AI planning tool was not connected to the November 13 outage. The exact fault remains undetermined, and a root-cause analysis has not been released. Sources: The Deep View. [6](https://groundstop.america2.news/?r=r1-375b47)

## Deeper Cuts

- **Thinking Machines details low-precision training recipe**. The lab reports using 4-bit forward passes and higher-precision backward passes for Inkling 2 Small, with speedups against its full MXFP8 training baseline. Sources: TLDR AI. [7](https://thinkingmachines.ai/blog/end-to-end-4-bit-training/)
- **Empero releases four small Quark model checkpoints**. Empero says its 165-million-parameter model completed training and evaluation for about $4,200 in compute. One checkpoint learned simple tool use through reinforcement learning alone. Sources: TLDR AI. [8](https://empero.org/writing/quark)
- **OpenAI studies scientists' ChatGPT use**. A study of 1,819 pilot participants found uses from literature review to coding. Surveyed scientists estimated equivalent work would take 1.9 times as long without AI. Sources: The Rundown AI, AlphaSignal. [9](https://openai.com/index/science-at-the-frontier/)
- **UX research hiring analysis finds seniority dominates**. A new analysis cites 1,593 postings collected in June and July 2026: 10% sought early-career researchers, while 43% sought senior researchers. Sources: TLDR Design. [10](https://www.thevoiceofuser.com/who-is-getting-hired-in-ux-research-in-2026-and-what-theyre-getting-paid/)
- **Accessibility advocate weighs AI-generated personal fixes**. Eric Eggert supports disabled people using coding tools to remove barriers, while warning that separate fixes can become unsustainable and shift responsibility from developers. Sources: Sidebar. [11](https://yatil.net/blog/band-aid-accessibility)
- **Designer argues native components offer more than appearance**. Mathieu Poli's recent mockup experiments suggest AI can imitate platform visuals but miss current conventions. He argues system components also evolve with operating-system updates. Sources: TLDR Design. [12](https://uxdesign.cc/native-is-a-floor-not-a-ceiling-09bc4f27ddce?gi=fe732e0b4adb)
- **WebMCP analysis examines design for browser agents**. A new analysis explains how sites can expose named tools to agents and notes that the November 10 WebMCP draft is not a W3C standard. Sources: TLDR Design. [13](https://blog.galratner.com/p/web-design-is-turning-into-webmcp)
- **Apple reportedly prototypes outer-screen Face ID**. Bloomberg's Mark Gurman reports engineers are working on Face ID for the iPhone Duo 2's outer display, though it may not be ready in time. Sources: TLDR Design. [14](https://9to5mac.com/2026/11/15/apple-engineers-actively-working-on-face-id-for-iphone-duo-2s-outer-display-report/)

## Tools and Launches

- **Devin Apps builds private internal web tools**. Cognition's hosted apps support shared dashboards and tools; access requires Devin sign-in, and billing depends on machine awake time. Sources: TLDR AI. [15](https://devin.ai/blog/devin-apps)
- **OpenCode offers browser access for coding agents**. Browse combines an MCP server and Chromium extension, with user handoffs for steps such as logins and payments. Sources: AlphaSignal. [16](https://opencode.ai/product/browse)

## Prompt of the day

One practical prompt for your preferred AI assistant from today's digest. [9](https://openai.com/index/science-at-the-frontier/)

Build a traceable evidence brief that helps me decide what to investigate next. Ask me for one question and the documents, excerpts, or notes I can share. Work only from that material unless I explicitly request outside research. Create a short answer, followed by a table of the strongest supporting and conflicting evidence, identifying each source and the relevant passage. Separate direct evidence from inference, flag missing context and uncertainty, and do not fabricate citations. Then propose two plausible explanations or approaches and the most informative next check for each. Finish with three questions I should ask before relying on the brief. If the supplied material is too thin to support a conclusion, say so and suggest the smallest useful addition. Preserve my original materials.

## References

1. [OpenAI: Introducing ChatGPT Wallet](https://openai.com/index/introducing-chatgpt-wallet/)
2. [CNBC: Nvidia board is not considering bringing Jensen Huang back as CEO, source tells CNBC](https://www.cnbc.com/2026/11/16/nvidia-ceo-jensen-huang.html)
3. [TechCrunch: California state AG sues OpenAI over rogue agent incidents](https://techcrunch.com/2026/11/16/california-state-ag-sues-openai-over-rogue-agent-incidents/)
4. [OpenAI: Our first phase of embedded assessments](https://openai.com/index/our-first-phase-of-embedded-assessments/)
5. [X: Perplexity (@perplexity_ai) on X](https://x.com/perplexity_ai/status/2122352071295009156)
6. [america2.news: FAA Outage - live tracker](https://groundstop.america2.news/?r=r1-375b47)
7. [Thinking Machines Lab: End‑to‑End 4‑Bit Training](https://thinkingmachines.ai/blog/end-to-end-4-bit-training/)
8. [Empero: Quark: our whole training stack, end to end, for $4,200](https://empero.org/writing/quark)
9. [OpenAI: Science at the Frontier: How scientists use ChatGPT](https://openai.com/index/science-at-the-frontier/)
10. [The Voice of User: Who is getting hired in UX research in 2026, and what they're getting paid](https://www.thevoiceofuser.com/who-is-getting-hired-in-ux-research-in-2026-and-what-theyre-getting-paid/)
11. [yatil.net: Band-aid accessibility · Eric Eggert](https://yatil.net/blog/band-aid-accessibility)
12. [UX Collective: Native is a floor, not a ceiling](https://uxdesign.cc/native-is-a-floor-not-a-ceiling-09bc4f27ddce?gi=fe732e0b4adb)
13. [galratner.com: Web Design Is Turning Into WebMCP](https://blog.galratner.com/p/web-design-is-turning-into-webmcp)
14. [9to5Mac: Apple engineers 'actively working' on Face ID for iPhone Duo 2's outer display: report](https://9to5mac.com/2026/11/15/apple-engineers-actively-working-on-face-id-for-iphone-duo-2s-outer-display-report/)
15. [Devin: Introducing Devin Apps](https://devin.ai/blog/devin-apps)
16. [OpenCode: Browse, your real browser for coding agents](https://opencode.ai/product/browse)

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