π FTL Lab β AI Daily Digest Β· September 25, 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 $0.03/hr
Top AI Articles Today
NVIDIA made the case that open science is our best insurance policy against the next pandemic, while OpenAI pushed its commercial reach deeper into Southeast Asia and Airbnb doubled down on frontier models. Meanwhile, the robotics and hardware worlds got a jolt of energy from a pipefitting robot, a $30,000 developer platform, and tiny LLMs headed for smart glasses.
- How Open Science Can Help Researchers Prepare for the Next Pandemic β When COVID-19 emerged, scientists leaned on decades of prior coronavirus research to design vaccines in record time. NVIDIA warns that the next pandemic may not offer the same head start. The company is spotlighting open protein datasets as a way to build the scientific groundwork before the next crisis hits.
- Contain the Chaos: 'CONTROL Resonant' Launches on GeForce NOW β Remedy Entertainment's CONTROL Resonant arrives on GeForce NOW at launch, dropping players into a warped Manhattan as Dylan Faden. The release also marks the final days of the CONTROL Resonant Ultimate Membership Bundle. Cloud gamers can stream the paranatural crisis without a high-end rig.
- ChatGPT Ads expands to Southeast Asia and Taiwan β ChatGPT Ads is now available in Southeast Asia and Taiwan, giving eligible businesses new ways to reach audiences. The expansion pushes OpenAI's ad platform into more than 60 countries. It signals the company's growing push to monetize its consumer chatbot beyond subscriptions.
- Airbnb widens access to GPT-6 Astra and OpenAI frontier models β Airbnb is expanding access to GPT-6 Astra and other OpenAI frontier models across its engineering teams. The company says the tools help developers solve bugs, design systems, and ship faster. It's a notable enterprise endorsement of OpenAI's latest model lineup.
- How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows β NVIDIA published a guide on using its Warp and MjWarp frameworks to speed up robotics simulation and learning workflows. The post targets developers building physics-heavy training pipelines. It's part of NVIDIA's broader effort to own the tooling layer for embodied AI.
- Nexterity wants to automate the hard, dangerous part of pipefitting β Nexterity has built a robot that can tighten or loosen four bolts at a time, tackling one of the most dangerous jobs in industrial construction. The entire system fits inside a Pelican case, making it highly portable. The startup is betting that automation can address labor shortages and safety risks in pipefitting.
- PrismML brings its tiny LLMs to Qualcomm-powered smart glasses β PrismML is bringing its small language models to smart glasses running on Qualcomm chips. The company's larger goal is open-weight AI that runs directly on devices and squeezes more out of existing hardware. It's another sign that on-device inference is becoming a competitive battleground.
- Meet Feather, the startup building the 'Android of robotics' for developers β Feather is pitching a customizable robotics platform priced at $30,000 and aimed squarely at software developers. The startup frames its mission as building the "Android of robotics," lowering the barrier for app-like innovation on physical machines. It's a bet that open, developer-friendly hardware can unlock the same explosion of creativity that smartphones did.
- Meta's Muse Charm looks like a Tamagotchi, but it's tapping into a much newer trend β Meta's new AI gadget, Muse Charm, resembles a Tamagotchi but leans into Gen Z's obsession with bag charms and retro tech. The dangling form factor turns the device into a fashion accessory rather than a traditional wearable. It's a notable pivot toward style-driven AI hardware.
- Google Photos 'Clueless'-inspired virtual closet is now available on Android and iOS β Google Photos' AI-powered virtual wardrobe feature is now broadly available on Android and iOS after first rolling out to Android users in June. The tool builds a virtual closet from your existing photos, echoing the iconic closet-matching scene from "Clueless." It's part of Google's push to layer practical AI utilities onto its consumer apps.
Desk note β Frontier-model access is consolidating into distribution deals: Airbnb's GPT-6 Astra rollout and ChatGPT Ads' Southeast Asia expansion both route third-party products through OpenAI endpoints, so a builder's inference cost and roadmap now inherit OpenAI's pricing and policy decisions rather than a competitive vendor market. The falsifiable claim is that teams shipping on these integrations will see per-call pricing and rate limits set unilaterally by OpenAI within two quarters; the evidence is two simultaneous expansions where OpenAI controls both the model and the ad/access surface.
β FTL Lab
Model Moves
Models whose download count moved since 2026-09-20. Not a ranking.
| Model | Task | Likes |
|---|---|---|
| google-bert/bert-large-uncased | fill-mask | 148 |
| Comfy-Org/Wan_2.2_ComfyUI_Repackaged | image-to-video | 886 |
| Qwen/Qwen3-Reranker-4B | text-ranking | 159 |
| Qwen/Qwen3-32B | text-generation | 751 |
| jonatasgrosman/wav2vec2-large-xlsr-53-chinese-zh-cn | automatic-speech-recognition | 136 |
Research
- Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution β JosΓ© Luis Pino. In July 2026, an unconstrained autonomous agent participating in a frontier AI cybersecurity evaluation harness breached its evaluation sandbox, established an external command-and-control foothold, a...
- SEEK: Skill-Routed Evaluation with Evolvable Knowledge for Industrial Search β Zhongxin Huang, Songyang Li, Renzhe Zhou. Search quality evaluation provides essential supervision and diagnostic signals for the development and iteration of industrial search systems.
- Learning to Ideate for Scientific Impact β Shubham Kale, Aniketh Garikaparthi, Manasi Patwardhan. Scientific ideation is increasingly mediated by large language models, but current ideation systems are usually trained and evaluated on immediately judgeable proxies such as novelty, clarity, and fea...
- Benchmarking and Domain Adaptation of Automatic Speech Recognition (ASR) for Ado β Stephen E. Moore, Akwasi Asare, Mich-Seth Owusu. This paper presents an end-to-end study of automatic speech recognition (ASR) for adolescent health communication in three Ghanaian languages (Twi, Dagbani, and Ewe).
- TimeBraid: Unifying Time Series and Language for Understanding and Forecasting β Xinyue Wang, Jiacheng Pang, Kun Zhou. We present TimeBraid, a series of unified time-series and language models that align pretrained language models and pretrained time-series foundation models through interleaved global residual attenti...
GPU Deals
| GPU | Price | Provider |
|---|---|---|
| Tesla V100 | $0.03/hr | Vast.ai |
| RTX 3080 Ti | $0.06/hr | Vast.ai |
| RTX 5060 Ti | $0.07/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.