🌅 FTL Lab — AI Daily Digest · September 21, 2026

Data Pulse

10 news articles
5 trending models
5 research papers
Cheapest GPU: Tesla V100 at $0.03/hr

Top AI Articles Today

The AI industry’s biggest players spent today projecting confidence while quietly guarding their secrets, with Nvidia’s Jensen Huang dismissing existential risk and President Trump unveiling a new “AI Force” even as his own base revolts over data centers. Meanwhile, world-model startups are hoarding cash and information in equal measure, and researchers keep reminding us that humans — not rogue machines — remain the sharpest threat to critical infrastructure. From Capitol Hill to your cat’s dinner bowl, here’s what mattered in AI today.

World model companies are keeping a lot of secrets — Everyone in the world-models space is sitting on piles of cash and buzz, yet founders, investors, and even their own data suppliers refuse to say what they’re actually building. The secrecy spans the entire supply chain, leaving outsiders to guess at the technology behind the hype. It’s a striking contrast to the usual startup playbook of loud, early demos.
Is the AI industry really ready to slow down? — On the Equity podcast, the team debated whether AI executives genuinely mean it when they call for slowing down development. The conversation probed the gap between public caution and the breakneck pace of private investment. So far, the incentives suggest the brakes aren’t coming anytime soon.
TechCrunch Mobility: How do we know when an AV is safe enough? — TechCrunch Mobility’s latest edition tackles the thorny question of measuring autonomous vehicle safety. As AI takes on a bigger role in transportation, regulators and companies still lack a shared standard for “safe enough.” The newsletter remains a free hub for tracking the future of transportation.
No one is surprised that Nvidia’s Jensen Huang thinks AI fears are overblown — In an interview with CBS Sunday Morning, Huang claimed there is a “0% chance” of AI ending humanity. The man positioned to profit most from the AI boom dismissed concerns that veteran researchers have studied for decades. His optimism lands as Nvidia’s chips power nearly every major model in development.
Trump now says he wants to form an ‘AI Force’ — President Trump posted on Truth Social that he wants to appoint an “AI czar” to lead a new “AI force.” The announcement came amid growing calls across the political spectrum, and even inside the industry, to pump the brakes on AI development. Trump added that his administration “will not in any way hinder” the technology’s progress.
Humans, not rogue AI, are still the biggest cybersecurity risk to energy systems — Long before recent high-profile hacks raised the specter of AI “killing all humans,” energy systems were already disturbingly vulnerable to cyberattack. The risk keeps growing, and the primary threat remains human actors rather than autonomous machines. As one expert put it: “We were always prey.”
Meta's Muse Is Better at Surveilling Than Helping Me — Meta’s new Muse app continues the company’s pattern of opting users into data collection for AI training. It also nudges users to hand over bank account, email, and passport information. The result, per WIRED’s reviewer, feels less like assistance and more like surveillance.
It’s Donald Trump Versus MAGA on Data Centers — President Trump has doubled down on data centers and AI, framing them as central to American competitiveness. His base, however, is running in the opposite direction, with local opposition mounting over energy use and land impact. The rift exposes a growing fault line within the coalition over AI’s physical footprint.
What ChatGPT Thinks It Knows About You Is Affecting Its Answers. Here’s How to Change That — ChatGPT’s latest memory upgrade shapes its responses based on what it has stored about you. Understanding how that memory works can make OpenAI’s productivity tool considerably more useful. WIRED walks through the settings you can adjust to control what the model remembers.
[Petlibro’s new AI-powered feeder is a game changer for multi-cat homes](https://techcrunch.com/2026/09/19/petlibros-new-ai-powered-feeder-is-a-significant development-for-multi-cat-homes/) — Petlibro’s Granary 2 smart feeders use a built-in scale and, on pricier models, an AI camera to track exactly how much each cat eats and when. The setup is a genuine upgrade for households juggling multiple felines and their diets. The fanciest health-monitoring features, however, require an extra subscription.

Desk note — World-model vendors are converging on a common posture: publish the demo, withhold the weights, and let benchmark scores stand in for reproducibility. That shifts the build decision from "which model is best" to "which vendor's API contract and data-retention terms can survive a procurement review," because a model that cannot be self-hosted becomes a recurring line item and a single point of failure. The falsifiable bet: within two quarters, at least one major world-model lab will ship a licensed on-prem tier priced above its hosted API, and teams without an abstraction layer over these endpoints will absorb the migration cost.

— FTL Lab

Model Moves

Models whose download count moved since 2026-09-20. Not a ranking.

Model	Task	Likes
Qwen/Qwen3-0.6B	text-generation	1649
mudler/locate-anything.cpp-gguf	object-detection	14
cdiamond/Qwen3.8-27B-iMatrix-NVFP4-MTP-GGUF	image-text-to-text	11
prism-ml/Ternary-Bonsai-2-27B-gguf	text-generation	1500
timm/mobilenetv3_small_100.lamb_in1k	image-classification	122

Research

AutoRecLab: Describe the Experiment, Get the Code! — Moritz Baumgart, Philipp Meister, Justus Krell. Empirical evaluation is central to recommender-systems (RecSys) research, but turning experimental designs into executable code remains a manual and error-prone task.
Do Personality-Tuned LLMs Make Better Social Agents? — Tim Krabbe, Xiaodan Shi. LLMs are increasingly used in social simulations for socially interactive agents and robots, offering more flexibility than rule-based systems.
EnterpriseVal: Quantifying the Efficacy, Reliability and Value of Generative AI — Abbas Raza Ali, Muhammad Ajmal Siddiqui, Moona Zahid. Frontier language models now produce professional deliverables that expert graders judge to match human work on a substantial share of economically valuable tasks, yet most enterprise GenAI initiative...
Federated Deep Clustering Networks for High-Dimensional and Heterogeneous Data — Morris Stallmann, Charalampos S. Kouzinopoulos, Marcin Pietrasik. Clustering high-dimensional data is a fundamental task in unsupervised machine learning with applications to a variety of domains.
Touvigation: Embodied Adaptive Object Acquisition for Blind and Low-Vision Users — George Xi Wang, Xiangyu Li, Shaoyue Wen. Blind and low-vision users often face challenges when locating and physically acquiring objects in unfamiliar indoor environments.

GPU Deals

GPU	Price	Provider
Tesla V100	$0.03/hr	Vast.ai
RTX 4070S Ti	$0.07/hr	Vast.ai
RTX 5070	$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)

View all events