This Week In OS AI Field Notes: 6 November 2026

Open-Source Models

• KIMI K3.1 PRO: Moonshot announced KIMI K3.1 Pro, the best open weights model for most coding, 3D modeling & CAD, and financial research tasks after Opus 5.5 and GPT-6 Astra at a much lower price. The new model is now capable of building entire worlds in game dev, creating interactive landing pages, performing heavy STEM simulations, and more. [ref] [blog] [23 OCT 2026]

• LFM3: Liquid AI introduced a new generation of specialized intelligence models capable of performing competitively with Gemma 4 12B, Gemini 3.5 Flash Lite, and GPT-5.6 Luna on less capable devices. Liquid AI models hold #1 on model intelligence and hardware resource costs, making it the best choice for on-device AI. [ref] [blog] [27 OCT 2026]

• NVIDIA NEMOTRON 3.5 TRANSCRIBE: NVIDIA released NEMOTRON 3.5 Transcribe, the #1 open-weight model on HuggingFace’s Open ASR English leaderboard. NVIDIA brought transcription in 40 languages, speaker attribution, translation, contextual prompting, and audio understanding with a 5.33% average word error rate for keeping track of who said what when everyone’s talking at once. [ref] [28 OCT 2026]

• PIKA H3 FLEX: Pika Labs released Pika H3 Flex model, which ranks #1 among open-source text-to-video alternatives on DesignArena’s t2v benchmark. It’s cheaper than Gemini and Seedance, features real time editing in 4K, and handles complex multi-shot camera choreography better than all video generation models on human anatomy and artistic styles. [ref] [blog] [29 OCT 2026]

Tools & Frameworks

• PROGRESS INDEX: Artificial Analysis released a new index to measure capability scale for every LM. Notably, open weights models generally trail proprietary models by about 4 months: since January 2025, each new highest open weights score came 2-8 months after a proprietary model first reached it. Currently the best open weights model, Kimi K3.1 Pro, lags 2 months behind the closed frontier. [index] [29 OCT 2026]

• STRIX X VERCEL: Strix, the open-source AI penetration testing tool, can now run on any of the 400+ models on Vercel AI Gateway. Security measures now reach models from Kimi, GLM, and Deepseek without the need to sign up for multiple providers. Anyone can use this full-stack security platform to secure code, APIs, web apps, infrastructure, and cloud. [ref] [github] [29 OCT 2026]

• DOOP: Doop is a multi-user design canvas created by Kevin Goedecke where people and AI agents can create side by side. Upload your design requests and references – from Meta ads to X article banners to company logos — and agents will draft multiple visuals, iterate on the ones you like, and guide you through the entire process. [github] [2 NOV 2026]

Policy & Security

• SINGAPORE AI RULING FOR BANKS: The Monetary Authority of Singapore finalized guidelines for how banks and other financial institutions manage AI risk, including generative AI and AI agents. Oversight scales with how risky each use is. The rules phase in from October 2027 and 2028. [news] [27 OCT 2026]

• U.S.A AI RESOLUTION: House Democrats introduced resolution to ban taxpayer-funded bailouts for AI companies if the industry faces a financial collapse. Failing firms should go through ordinary bankruptcy, and executives who broke the law should face consequences. Federal help should go to workers, retirees, and families instead because no one wants to repeat the 2008 crisis. [resolution] [29 OCT 2026]

• SUPER INTELLIGENCE VS ARTIFICIAL INTELLIGENCE SAGA: One month after Donald Trump ordered that “super intelligence” replace the term artificial intelligence, his leading tech allies, Elon Musk and Mark Zuckerberg, followed suit on earnings calls. Other tech CEOs from Microsoft, Amazon, and Alphabet were less eager to do the same and stuck to the term AI. [blog] [30 OCT 2026]

Podcasts & Media

• THE ALL-IN PODCAST, ANTHROPIC S-1 SHOCKER: Jason Calacanis, David Sacks, Chamath Palihapitiya, and David Friedberg discuss whether open-source models still threaten Anthropic after its S-1 revealed roughly 85% gross margins. They also cover why open source remains Anthropic’s biggest risk, since it could commoditize areas like coding if safety concerns slow Anthropic’s push toward recursive self-improvement. [youtube] [30 OCT 2026]

• 20VC PODCAST, HAS OPEN-SOURCE INFERENCE PEAKED?: Host and venture capitalist Harry Stebbings interviews Erik Bernhardsson, CEO of Modal, an AI infrastructure platform for training models, running inference, and powering AI agents. They discuss why open models remain vital for enterprises seeking tailored solutions, even as smaller models from leading AI labs become more affordable. [youtube] [2 NOV 2026]

What’s New at Sentient

• EVOSKILL V2 PAPER: The latest EvoSkill release automatically improves AI agent skills without changing the underlying model or runtime. It tests the agent, learns from its attempts, and rewrites its instructions to improve performance. The paper from the Sentient Labs research team compares v2 against other skill optimizers such as SkillOpt by Microsoft Research and SkillBoost by Alibaba Group on held-out tasks. [ref] [video] [paper] [github] [3 NOV 2026]

• OPEN AGI TEA HOUSE: We’re bringing together researchers, open-source contributors, and AI builders for an afternoon exploring the ideas shaping open-source AI—with tea, coffee, and board games. Researchers will walk through their work as we unpack their findings, examine limitations, and discuss what comes next. [RSVP] [14 NOV 2026]

• OPEN AGI BUILDER’S DAY IN SHANGHAI: Shanghai Jiao Tong University is co-hosting an event bringing together builders, researchers, and founders shaping the future of open-source AI. The agenda covers Sentient Labs’ EvoSkill v2 release, Sentient Foundation’s Fingerprinting Leaderboard, builder showcases, and panel discussions on China’s AI startup ecosystem. [RSVP] [15 NOV 2026]

Sentient Foundation is a non-profit organization dedicated to developing open-source AGI, free from single-entity control. We accelerate AGI research and development through Sentient Labs and a global community of open-source builders.

We're building intelligence that dreams. Dreaming machines, designed in the open. Advancing our shared future through open-source AGI.

Visit Sentient Foundation website, follow us on LinkedIn and Twitter/X.