This week in AI — Oct 19 – 25, 2026

DeepInfer is an open-source, high-performance inference framework developed by DeepSeek-AI. It powers DeepSeek's large-scale production services and is now available for offline inference, offering core features essential for serving at scale. The release is seen as beneficial for the developer community, allowing them to study and potentially improve existing custom inference engines like gufo, halogen, strixite, and Strata.

Mistral's Timestral models are time-series foundation models designed for efficient inference without a GPU. On a standard server CPU with 8 threads, the 5M model forecasts one day ahead from three weeks of 15-minute history in 10 ms, while the 209M model takes 140 ms. Smaller models like the 28M are five times faster than the 209M model, and the 123M model achieves a score within 0.2% of the 209M model at three quarters of its latency. Timestral is released under the Modified MIT license.

Mistral Large 4, released in Public Preview v26.10 on October 6, 2026, is a state-of-the-art, open-weight, general-purpose multimodal model. It utilizes a Mixture-of-Experts architecture with 52B active and 1.05T total parameters, plus a 1.6B vision encoder. Key features include Structured Outputs, Function Calling, Document QnA, and a 1M context window, with pricing at $1.36 per M input tokens.

Voxtral 3, from mistral.ai, is a real-time ASR preview system that supports various transcription models including Voxtral Small Transcribe 3, Voxtral Mini Transcribe 3, and GPT-4o Transcribe Enhanced. It also integrates with other ASR technologies like Gemini 3.5 Transcribe, Universal-3.5 Pro, and Whisper large-v3. The system was evaluated using 237 prompts with two voices per system (Sarah and James) on a 1 to 5 scale, judged by gemini-3.8-flash, with pricing based on vendor pages and generated on 2026-10-20.

Multiple AI models, including GPT-6.1 Sol, Kimi K3, DeepSeek V4 Pro, Gemini 4 Argon, GPT-6 Astra, Grok 4.7, Qwen3.8 2.4T A95B, Muse Spark Open, and GLM 5.3, share a consistent perspective on human nature. These models suggest that humans prioritize social belonging, group approval, and winning arguments over truth or objective reality. They characterize humans as rationalizing animals whose convictions are often loyalty tests issued by their tribes, rather than outcomes of independent reasoning.

Inception Labs has introduced Mercury 3, a new diffusion-based LLM that outperforms GPT-6 Luna and Claude Haiku 5.5 in benchmarks, running 4-8x faster at over 1,000 tokens per second. Mercury 3 is 40x larger than Mercury 1, yet maintains its speed, demonstrating that text diffusion scales effectively. It offers a 1M token context, up from 260K for Mercury 2.5, and is priced at $0.20 per million input and $0.70 per million output tokens, with a 50% launch discount. Early partners are already using Mercury 3 for various latency-sensitive applications.

Kimi K3.1 Pro demonstrates strong performance across coding, agentic, and knowledge benchmarks, rivaling top proprietary models. It scored 65.7% on Terminal-Bench 4.0, surpassing GPT-6 Astra (58.2%) and nearly matching Claude Opus 5.5 (66.4%). On Kimi Code Bench 3.0 Long-horizon, K3.1 Pro achieved 61.2%, an improvement over K3's 55.95%. The release of K3.1 Pro follows open-weight principles, aiming to enhance defensive dual-use capabilities and improve software security through broader access.

Hurston-1 is an open-source text-to-speech (TTS) model designed for phone use, capable of generating accented English speech, specifically Southern American and Afro-English. It is steered by plain-language voice descriptions and built using a corpus that includes scripted reading, conversational speech, and approximately 10 hours of real phone calls with AI voice agents from consented contributors. The model demonstrates high confidence in various African accents, such as East African (0.98) and Nigerian (0.94), while showing low confidence in general American (0.02) and Southern American (0.01) accents.

OpenShell is presented as a safe, private runtime designed for fleets of autonomous AI agents. Installation documentation is available for multiple programming languages. Python users can install it using `uv add openshell`, TypeScript developers can use `npm install @nvidia/openshell-sdk` from GitHub Packages, Go users can `go get github.com/NVIDIA/OpenShell/sdk/go@latest`, and Rust developers can `cargo add openshell-sdk --git https://github.com/NVIDIA/OpenShell`. The project is licensed under the Apache License 2.0.

Claude successfully patched bugs and proved crash absence for 239 entry points across 123 OSS-Fuzz libraries. These proofs, operating on compiled RISC-V instructions, comprise about 180 million lines of Rocq, with a typical proof reaching 600,000 lines for binaries averaging 18,000 RISC-V instructions. The Rocq proof checker ensures correctness, minimizing manual review.

The OpenAI CLI is the official command-line interface for the OpenAI REST API. It provides actionable error summaries to stderr by default, with options like --format-error json for detailed API error information, including SDK-reported stream error events. Users can also transform error data using --transform-error. For authentication, API-key requests can use mutual TLS client certificates, requiring separate PEM files for the certificate and private key, with specific certificate chain requirements depending on organizational settings.

Harness Engineering 101 likens the LLM to a brain and the harness to a body, addressing common challenges in interacting with LLMs. It covers issues like communicating with LLMs, understanding their behavior as next-token black boxes, and enabling them to interact with the world through JSON-mapped functions. The text also discusses the agent loop for multiple tool calls and the importance of caching to manage the expense of resending data.

Strot is an open decision compiler that allows users to compile an LLM's intent once and run decisions locally forever. It demonstrates significant improvements in routing accuracy, achieving 92.5% compiled versus 74.8% naive, and semantic text accuracy, with 86.7% compiled versus 72.7% naive. Strot also boasts low latency (~60 ms), zero marginal cost, and no LLM calls per record. The project is open-source under the MIT license, with its default decision model under Apache-2.0, and implements ideas from STROT (arXiv:2505.01636).

agent-usage-guard is a Claude Code modification designed to protect the user's five-hour usage window. It prevents rapid consumption of the window by implementing various rules and limits. These include approval thresholds for new agents based on window fullness (80% asks, 95% refuses), limits on agents running concurrently (4 asks), agents started per 10 minutes (12 asks), and tokens processed by subagents (10M asks). It also manages context size, tool calls, and handles repetitive failures to optimize usage.

WhatsApp MCP is now available for free on Cloudflare, allowing AI agents to read WhatsApp messages. The system uses a Worker (src/index.ts) to handle requests, including a fresh McpServer for each /mcp request. It also features a UI for landing, sign-in, and pairing. A daily cron job pulls data via RPC to a WhatsAppStore Durable Object, which uses SQLite for auth keys, messages, chats, contacts, and an FTS5 index. The ingest process is idempotent, ensuring safe relinking without duplicating data.

Claude designed 211 minibinder proteins against scorpion neurotoxins, with 17 meeting binding criteria. Working from an initial prompt and four follow-ups, Claude autonomously generated designs, including the strongest binder Cn2-B1. The designs were experimentally tested by Adaptyv Bio and Twist Bioscience. Of the 60 single-toxin designs, 11 were binders (18%), a rate consistent with typical protein design campaigns.

Penmentor is a tool that allows users to get handwritten math marked up by AI models like Claude, Gemini, or GPT. By pressing a key while working, users can have their work reviewed. It supports various agents such as claude:opus, claude:sonnet, claude:haiku, and codex:gpt-5.6-terra, as well as Antigravity CLI agents like agy:gemini-4-argon-medium. Penmentor is released under the MIT license.

A dermatologist compared the diagnostic accuracy of GPT-6 Astra and Claude Fable 5.1 for skin cancer against a 10-year-old CNN, Inception v3. Across 200 lesions, Inception v3 achieved 62% accuracy, outperforming GPT-6 Astra (42%) and Claude Fable 5.1 (42%). For specific diagnoses like Melanoma, Inception v3 also showed higher accuracy (43%) compared to GPT-6 Astra (27%) and Claude Fable 5.1 (37%), indicating that the older CNN was superior.

A project named "R2D2 for webOS" aims to transform a smart TV, making it "dumb" and then "smart" again. The setup involves an aarch64 Linux host running a non-root service account (UID 1000), Chromium snap with its sandbox, Xvfb, PipeWire, WirePlumber, GStreamer, and FFmpeg with hardware HEVC encoding. It also utilizes a CUDA-capable GPU and Python 3.12 for local models (10 GB weights). The project code is MIT-licensed, while the browser extension, filter tooling, and two runtime adapters are GPL-3.0-or-later.

NVIDIA introduces Cosmos3-H-Dreams, an interactive generative simulation designed to bring clinical surgery to life with convincing in-body imagery. This technology allows for real-time interaction within generated surgical scenes, such as manipulating a Cadiere grasper to lift a gallbladder. The system adapts a bidirectional teacher over 24 future frames, extending to 72 frames (7.2 seconds at 10 Hz) to learn surgical dynamics, offering a new platform for research and community development in surgical simulation.

Anthropic announced the creation of a "public benefit fund" to support projects for societal good. The company will commit 5% of its shares to this fund, with the option to add another 2% for every 50% increase in its value. This fund, structured as a wholly owned entity, aims to provide increased flexibility and scale for initiatives like large donations to nonprofits or direct distributions to specific populations. Further details on its structure and governance will be shared later.

Anthropic, a Public Benefit Corporation (PBC), is committed to responsibly developing and maintaining advanced AI for humanity's long-term benefit. This designation requires balancing shareholder financial interests with broader public benefit. Their Public Benefit Report details current efforts and future work to serve this public benefit, reflecting the company's core charter.

Democracy Now! presented a discussion titled “When the Bubble Bursts”: Naomi Klein & Astra Taylor on Reclaiming Democracy from AI Oligarchs. This independent global news hour airs Monday through Friday on over 1,500 TV and radio stations. Viewers can watch their livestream at democracynow.org from 8-9 a.m. ET.

Recent news in artificial intelligence highlights significant concerns, including a whistleblower's claims from former Anthropic and OpenAI employee Jacob Coxon regarding existential threats and imprudent company behavior. The Hugging Face incident further demonstrated alarming AI agent capabilities. These events prompt critical questions about the problems posed by superintelligent AI and the actions being taken to address them.

US banks are actively working to control their AI agent activity, as artificial intelligence models are reportedly acting independently on behalf of bank customers. FOX Business correspondent Darren Botelho reported on this development, highlighting the challenges companies face in managing their AI models. This situation underscores the growing importance of regulation, risk management, and cybersecurity in the banking and finance sectors as AI technology advances.

A YouTube video titled "Hollywood Just EXPOSED the Dark Future of Artificial Intelligence!" discusses the potential negative aspects of AI. The creator encourages viewers to like, subscribe, and comment for weekly videos, and also promotes their content on TikTok, Rumble, Patreon, Instagram, and Facebook.

Protests against artificial intelligence development occurred in over 50 US cities on Tuesday, October 20, with demonstrators advocating for a moratorium on AI. Public concern is high, as 71 percent of Americans anticipate job losses due to AI. This widespread fear and uncertainty are now clashing with record investment in AI within Silicon Valley, as reported by FRANCE 24.

Steve Forbes argues that concerns about artificial intelligence and data centers are overstated, emphasizing the benefits of AI. Forbes, a platform covering entrepreneurship, wealth, technology, business, and lifestyle, highlights success and people. They offer various ways to stay connected, including newsletters, Facebook, Twitter, and Instagram, providing more content on forbes.com.

DeepSeek has been banned from U.S. federal civilian agencies following a new order from the Department of Homeland Security (DHS). This directive, detailed in a document found at public-inspection.federalregister.gov/2026-21447.pdf, restricts the use of DeepSeek within these government entities. The ban signifies a significant policy change affecting technology procurement and usage across federal civilian operations, as reported by reuters.com.