AI Newsletters › TLDR AI › Issue

ChatGPT Wallet, Devin Apps, always-on agents | TLDR AI. TLDR AI 11/16

Monday, November 16, 2026 · 5 min read

See what's next for Google Cloud AI infrastructure (Sponsor)

Reasoning agents that run 24/7 place new demands on infrastructure. On November 9, Google Chief Technologist Amin Vahdat and fellow product leaders showed how Google is rising to the challenge. The exclusive broadcast covers:
  • The next horizon of AI compute: The latest advancements in TPUs supporting agentic workloads with high throughput and ultra-low latency.
  • AI Hypercomputer architecture: How co-designed compute, storage, networking, and open software drive efficiency across every layer.
  • Production scaling strategies: Approaches to shortening deployment cycles, cutting infrastructure costs, and scaling open-weight and proprietary models.
  • Project Suncatcher preview: A first look at Google's moonshot - taking compute into space.

Watch the replay now.

Devin Apps (3 minute read)

Devin Apps are private web apps that Devin builds and hosts from a description, such as dashboards and approval queues. Apps can use a team's existing integrations and codebase, and viewers must sign in to Devin.

ChatGPT Wallet (5 minute read)

OpenAI introduced ChatGPT Wallet, allowing AI agents to make purchases without accessing users' actual card details while requiring a security check for every transaction.

Quark: our whole training stack, end to end, for $4,200 (7 minute read)

Quark is a language model with about 165 million parameters trained for about $4,200 of compute (about 1,300 H-100-hours). It has the best benchmark average among open models of its size trained on similar token budgets. The model was created to explore every part of the training stack the team was using before spending serious money on a large model. This post looks at the model's data, architecture, pretraining at scale, context extension, reinforcement learning, evaluation, and release.

End-to-End 4-Bit AI Training (9 minute read)

Thinking Machines explored the technical challenges of training language models using 4-bit precision, including quantization errors and their effects on model accuracy. The deep dive examines BF16, MXFP8, and NVFP4 formats.

On the Brutality of Amdahl's Law (12 minute read)

The world is far too slow to see the high double-digit or more economic growth and seismic transformations in the economy that AI is predicted to bring. Intelligence isn't the only technology curve to go quasi-infinite in the last century and a half. Electricity, communication, computers, transportation, and agriculture have all experienced extreme growth and transformation. Despite this, we've only seen 2% GDP per capita growth for nearly 150 years.

Embla 1 (5 minute read)

Embla 1 is a reasoning model built for phones. It is designed to be fast and to use few tokens per answer. The model can run on a phone, privately, offline, for free. At 4 bits, Embla 1 averages 57.4 on BFCL, IFBench, AA-Omniscience, GPQA, and MATH-500 and answers in 1.77 seconds end-to-end on an iPhone 18 Pro.

Qwen4-Max: Your New Teammate Has Arrived (16 minute read)

Qwen4-Max can take on a wide range of creative and professional work in the fields of 3D modeling, game development, precision manufacturing, finance, law, healthcare, and frontier scientific discovery. It demonstrates strong 3D creation capabilities and can use Blender to carry out geometry modeling, scene assembly, material setup, lighting, and rendering. The model works fluently with leading game engines and can handle scene construction, gameplay logic, integration of external assets, material setup, HUD design, and visual effects, and deliver a playable game. Smaller models in the family are coming to open source.

Releasing Personal Agent Protocol 1.0 (5 minute read)

The Personal Agent Protocol enables personal agents to interact with companies through business agents, websites, and APIs. v.0 of the protocol is now available. The protocol defines how personal agents connect with companies that implement it. Customers control what personal agents can and can't do, and can manage their access over time.

GPT-6 Astra Tackles a Fundamental Physics Problem (15 minute read)

Physicist Nathaniel Craig worked with GPT-6 Astra to develop a candidate lattice formulation of the Standard Model, addressing a longstanding challenge in defining the theory without relying on perturbative approximations. Astra helped construct the mathematical framework and develop a detailed proof supporting the proposed approach.

Building always-on agents (13 minute read)

Always-on agents, revived by recent advancements, now manage context and act without direct supervision, running on a core loop enhanced with event inboxes, timers, memory, permissions, and outboxes.

More from TLDR AI

All issues →
Nov 13, 2026 · 5 min read

Google introduced Gemini 4 Hydrogen, a fast, efficient model optimized for coding and agentic workflows. Self-hosted Claude Managed Agents offers private…

Nov 12, 2026 · 6 min read

Grok Bot 1.0 adds Team Bots you can publish and share, Slack and Microsoft Teams, meetings on Meet and Zoom, shared connectors, and reviewed learnings.

Nov 11, 2026 · 5 min read

Google's strongest model is now available to all paying developers and enterprises through the API in AI Studio and Gemini for business, and is rolling…

Nov 10, 2026 · 5 min read

Anthropic released Claude Fable 5.2 and Mythos 5.2, featuring improvements in coding, reasoning, and vision, with up to 30% faster output generation and…

Nov 9, 2026 · 5 min read

OpenAI reportedly secured $11 billion in a new funding round at a $1.25 trillion valuation, with the round potentially reaching $20 billion before closing.