🌅 FTL Lab — AI Daily Digest · October 16, 2026
Today: 10 new articles, 5 trending models, 5 research papers
Data Pulse
- 10 news articles
- 5 trending models
- 5 research papers
- Cheapest GPU: Tesla V100 at $0.07/hr
Top AI Articles Today
NVIDIA had a day that stretched from the data center to the desktop, launching RTX Spark for Windows PCs while expanding its open Nemotron family with a 120-billion-parameter Super VL model built for always-on agents. Meanwhile, OpenAI pushed deeper into science and education with a sponsored GPT-Rosalind plan for academic labs and new teacher-focused tools, and Meta entered the open-weight frontier race with Muse Spark Open. If that weren't enough, DLSS 4.5 controls arrived on GeForce NOW, Europe's scientific community gained new NVIDIA resources, and Cal AI's influencer-driven rise ended in an acquisition by MyFitnessPal.
- Be There for the Beginning: Celebrate the Launch of NVIDIA RTX Spark — NVIDIA is launching RTX Spark tomorrow, billing it as a new beginning for Windows PCs built for creators, gamers, and AI agents. The company is inviting users to celebrate alongside NVIDIA and its partners at launch events around the world. As part of the festivities, NVIDIA is asking people to share the story of their first Windows PC.
- Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models — NVIDIA released Nemotron 3.5 Super VL, an open 120-billion-parameter model. The company says it delivers the fastest task completion and highest accuracy for specialized always-on agents. The release extends NVIDIA's open model strategy from data centers all the way to desktops.
- Dial Up the Cloud: NVIDIA DLSS 4.5 Controls Arrive on GeForce NOW — This GFN Thursday brings new NVIDIA DLSS 4.5 controls to GeForce NOW members. The update gives players more choice over how supported games look and perform in the cloud. Notably, the controls work even without the latest hardware.
- NVIDIA Expands Advanced Technology Access for Europe's Scientific Community Through Open Learning Resources, Collaborations and Compute — NVIDIA is broadening access to advanced technology for Europe's scientific community through open learning resources, collaborations, and compute. The company notes that turning accelerated computing's potential into scientific progress — from climate modeling to molecule design — depends on more than breakthrough technology alone. The pledge aims to close that gap for European researchers.
- The Nordics and NVIDIA Show What's Possible When a Region Commits to Advanced Computing — The Nordic countries have become Europe's leader in unicorn startup density, producing the most $1 billion-plus companies per capita on the continent. A new chapter is now unfolding as these nations race to build one of Europe's most consequential technology ecosystems. NVIDIA is highlighting the region as a model for what happens when a region commits to advanced computing.
- Bringing GPT-Rosalind to academic research labs — OpenAI introduced the Rosalind Academic Plan, offering eligible academic labs sponsored access to GPT-Rosalind. The program includes tools designed to advance life sciences research. It marks OpenAI's latest push to embed its models in scientific discovery workflows.
- Teaching students to think and learn through questions — OpenAI celebrated Teachers' Day with the launch of AI for Educators in Brazilian Portuguese. The announcement highlights how Study Mode can help teachers support active, step-by-step learning. The focus is on teaching students to think and learn through questions rather than passive answers.
- Sullivan & Cromwell brings legal expertise into agreement review — Law firm Sullivan & Cromwell collaborated with OpenAI to develop Agreement Analyzer. The AI tool is designed to identify issues in a draft agreement. It represents a notable example of a major legal institution building domain-specific AI on top of OpenAI's models.
- Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta — Meta released Muse Spark Open, an open-weight frontier multimodal mixture-of-experts model. The release lands on Hugging Face and adds another major entrant to the open-weight frontier race. It signals Meta's continued commitment to openly available multimodal AI.
- When it came to Cal AI's viral success, a small army of influencers paid off — Cal AI generated over $30 million in revenue in just two years, fueled by a small army of influencers. The company was acquired by MyFitnessPal in March for an undisclosed amount. Its trajectory offers a case study in how influencer-driven distribution can supercharge an AI app's growth.
Desk note — NVIDIA's simultaneous push across RTX Spark, Nemotron VL/Bolt, DLSS 4.5 on GeForce NOW, and European compute partnerships means the company is now the single upstream dependency for a developer's training stack, inference runtime, and edge deployment target. That vertical bundling raises a falsifiable risk: any team shipping vision-language features on Nemotron weights inherits NVIDIA's model licence terms and CUDA-only optimisations, so switching costs after integration exceed the 2–4 weeks typical of swapping between permissively licensed checkpoints. Budget for that lock-in explicitly, or keep at least one non-NVIDIA inference path in the architecture.
— FTL Lab
Model Moves
Models whose download count moved since 2026-10-09. Not a ranking.
| Model | Task | Likes |
|---|---|---|
| Qwen/Qwen3.8-27B | image-text-to-text | 17789 |
| ibm-granite/granite-timeseries-tspulse-r1 | unknown | 43 |
| unsloth/Qwen-Image-2.1-GGUF | text-to-image | 405 |
| Snowflake/snowflake-arctic-embed-l-v2.0 | sentence-similarity | 262 |
| Qwen/Qwen3.6-35B-A3B | image-text-to-text | 2990 |
Research
- Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion — Huixuan Cheng, Bike Zhang, Shankar Sastry. Recent advances in robot learning have driven rapid progress in humanoid control, enabling increasingly agile behaviors in the real world.
- Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sp — Zhefei Gong, Hefeifei Jiang, Zehao Wang. Mobile Manipulation (MoMa) is a cornerstone of autonomous robotic systems, yet real-world fine-tuning remains a formidable challenge because of high-dimensional action spaces and intricate coupling be...
- Co-Evolving Acting and Forecasting for Language Agents — Haofei Yu, Yixing Chen, Yining Zhao. Training language agents is bottlenecked by the high cost of long-horizon rollouts, yet standard agentic reinforcement learning (RL) learns only from a terminal scalar reward, leaving the supervision...
- Your Transformer Is Secretly Elastic — Walter Simoncini, Dheeraj Varghese, Gertjan J. Burghouts. Scaling up transformers has produced models with broad capabilities, deployed across heterogeneous hardware and fluctuating workloads.
- Price of Anarchy of First-Price Auctions in Agentic Markets — Shubhendu Trivedi, Jacob Eisenstein, Alekh Agarwal. An agentic market is a decentralized serving-time market where AI agents bid to claim each user prompt.
GPU Deals
| GPU | Price | Provider |
|---|---|---|
| Tesla V100 | $0.07/hr | Vast.ai |
| RTX 5070 | $0.10/hr | Vast.ai |
| RTX 4090D | $0.11/hr | Vast.ai |
View full GPU pricing dashboard
Community Events
New this week:
- Papers We Love: AI Edition (Online)
- MLOps Community Weekly Meetup (Online)
- NeurIPS 2026 (Online)
- Paris Machine Learning Meetup (Paris, France)
Was this article helpful?
Let us know to improve our AI generation.