D.A.D.: A Look At The Democrat-Republican Gap In AI Use — 2/15
The Daily AI Digest
Your daily briefing on AI
February 15, 2026 · 13 items · ~6 min read
From: Hacker News, Hugging Face Models, Hugging Face Spaces, NBER
D.A.D. Joke of the Day
My AI gave me five different answers to the same question. Finally, something at work that understands the value of a second opinion.
What's New
AI developments from the last 24 hours
Browser Tools Let Users Hide YouTube Shorts Entirely
A uBlock Origin filter list that hides all YouTube Shorts from the platform is circulating among users frustrated with the short-form video format. The list works with the popular ad-blocking browser extension. Community discussion surfaced several alternatives: browser extensions like RYS and Unhook, or simply pausing YouTube watch history to reduce algorithmic recommendations. Some pushed back on the premise—noting Shorts can actually be more efficient than long-form videos padded with filler and creator ad reads.
Why it matters: Reflects ongoing tension between platforms pushing algorithmic short-form content and users who want control over what they see—a dynamic playing out across every major social platform.
Major Publishers Block Internet Archive Over AI Training Fears
Major publishers including The Guardian, The New York Times, and Financial Times are blocking or limiting the Internet Archive's web crawlers. The concern: AI companies could use the Archive's structured databases and APIs as a convenient backdoor to scrape publisher content for training data. The Guardian acknowledged it hasn't documented specific instances of this happening via the Wayback Machine—the moves are preemptive. The crackdown signals how broadly publishers are now treating any large-scale content repository as a potential AI training pipeline.
Why it matters: This could reshape access to the web's historical record, as publishers increasingly view digital preservation infrastructure through the lens of AI liability rather than public benefit.
Smart Sleep Mask Allegedly Exposes Users' Brainwave Data to Open Server
A security researcher reportedly discovered that a smart sleep mask broadcasts users' brainwave data to an unsecured server, with all devices sharing identical credentials. The alleged flaw means anyone who accesses the system could potentially read brainwave data from any user—and, more alarmingly, send electrical impulses to other users' masks. The company wasn't named in the original post, though community members investigating the claims suspect it may be Luuna. The researcher provided no technical evidence in the public disclosure.
Why it matters: If verified, this highlights how consumer neurotechnology devices may be shipping with security practices that lag far behind their physical access to users—a regulatory and liability issue as brain-computer interfaces move mainstream.
Speculation Swirls: Should OpenAI Build a Slack Competitor?
An AI newsletter is floating the idea that OpenAI should build a Slack competitor, citing Sam Altman's recent ask for product suggestions. The argument: workplace messaging would be a natural extension of ChatGPT Enterprise, create customer lock-in through network effects, and target a market where Slack faces criticism over API pricing. The piece notes OpenAI hired former Slack CEO Denise Dresser in December and launched group chats three months ago. This is opinion, not reporting—no indication OpenAI is actually pursuing this.
Why it matters: The speculation reflects real anxiety among enterprise software vendors about whether AI labs will vertically integrate into their territory, a pattern worth watching as OpenAI expands beyond chat interfaces.
Babylon 5 Now Streaming Free on YouTube in Distribution Experiment
Warner Bros. Discovery is releasing full episodes of Babylon 5 on YouTube for free, starting with the pilot and adding episodes weekly. The move comes as the 1990s sci-fi series leaves Tubi's ad-supported catalog in February 2026. It's a notable distribution experiment: a major studio using YouTube as a primary viewing platform for legacy content rather than driving audiences to proprietary streaming services.
Why it matters: This signals studios are testing YouTube as a legitimate home for catalog content—worth watching if you're tracking how media companies are rethinking streaming economics and where audiences actually spend time.
What's Innovative
Clever new use cases for AI
Open-Source App Runs AI Text and Image Generation Entirely on Your Phone
Off Grid, a new open-source app, runs AI text generation, image creation, vision analysis, and voice transcription entirely on your phone with no internet connection required. All processing stays local—nothing uploads to external servers. The MIT-licensed app is available as an Android APK on GitHub, with iOS buildable from source. Performance claims include 15-30 tokens per second for text and 5-10 seconds for image generation on supported hardware. Early reactions on Hacker News praised the project's design, though users flagged a broken download link in the initial release.
Why it matters: For professionals handling sensitive data—legal, medical, financial—offline AI tools eliminate the privacy trade-off that comes with cloud-based assistants, though this is a hobbyist project requiring technical comfort to set up.
Multimodal AI Model Claims to Run on Less Powerful Hardware
AIDC-AI released Ovis2.6-30B-A3B, an open-source multimodal model that can process images and text together. It uses a mixture-of-experts architecture—a design where only a fraction of the model activates for each task, reducing computational costs. Despite having 30 billion total parameters, only 3 billion are active at once, meaning it could potentially run on less powerful hardware than its size suggests. The model is available on Hugging Face for developers to test and integrate.
Why it matters: This is developer infrastructure—another entrant in the crowded open-source vision-language space, potentially useful for teams building custom image-analysis tools but not something that changes off-the-shelf AI options for most business users.
Chinese Lab Drops GLM-5 Model on Hugging Face
A model called GLM-5-FP8 appeared on Hugging Face from zai-org, a text-generation model using a Mixture of Experts architecture. The FP8 format refers to 8-bit floating point precision—a compression technique that reduces memory requirements while aiming to preserve performance. No benchmarks, documentation, or performance claims accompanied the release, making it difficult to assess where this fits in the current landscape.
Why it matters: This is developer plumbing for now—without benchmarks or documentation, there's no way to know if GLM-5 competes with leading models or is a niche experiment, so it's not actionable for most readers yet.
"OpenResearcher" Tool Appears on Hugging Face With Few Details
A tool called OpenResearcher has appeared on Hugging Face, built with Gradio. Beyond the name and basic technical setup, no details are available about what it actually does or what research capabilities it offers.
Why it matters: This is too thin to act on—worth watching if you're tracking AI research tools, but there's nothing concrete here yet.
Free Virtual Try-On Demo Lets Retailers Test AI Fitting Room Tech
A virtual try-on tool called FASHN VTON 1.5 is now available on Hugging Face, using diffusion models to show how clothing items would look on a person's photo. The free demo lets users upload images and generate try-on results. This is developer/demo-stage tooling—retail teams exploring AI-powered virtual fitting rooms may want to test it, but there's no published accuracy data or integration pathway yet.
Why it matters: Virtual try-on tools could reduce return rates for online retailers, but this appears to be an early-stage demo rather than an enterprise-ready product.
What's Controversial
Stories sparking genuine backlash, policy fights, or heated disagreement in the AI community
Quiet day in what's controversial.
What's in the Lab
New announcements from major AI labs
Quiet day in what's in the lab.
What's in Academe
New papers on AI and its effects from researchers
AI Financial Analysis Helps or Hurts Depending on Platform Culture
A research paper finds AI-generated financial content produces opposite effects depending on where it's posted. On Seeking Alpha, AI-assisted analysis appears to improve content quality—sentiment in AI-enhanced posts positively predicts future stock returns, and trading around these posts shows tighter spreads and more informed order flow. On WallStreetBets, the pattern reverses: AI adoption follows retail buying surges, and the resulting content precedes higher volatility, increased trading volume, and more lottery-like return distributions. The researchers attribute the divergence to platform culture—one rewards analytical rigor, the other amplifies emotional narratives.
Why it matters: As AI writing tools become ubiquitous, this suggests the same technology can either sharpen financial analysis or accelerate speculation, depending entirely on the community norms where it's deployed—a finding with implications for anyone using AI to draft market commentary or consuming AI-enhanced investment content.
Democrat-Republican Gap in AI Use Reflects Job Types, Not Politics
An NBER working paper finds Democrats report using AI at work more frequently than Republicans—27.8% vs. 22.5% weekly or daily in late 2024—but the partisan gap largely disappears once researchers control for education, industry, and occupation. Democrats tend to work in jobs with higher AI exposure (tech, professional services) and report greater concern about AI-driven job displacement. The study, by Stanford's Nicholas Bloom and Christos Makridis, suggests the apparent political divide reflects where people work, not political identity itself.
Why it matters: For executives reading culture-war framing around AI adoption, this offers a corrective: workforce composition—not ideology—drives the gap, which matters for realistic planning around AI rollouts across diverse teams.
Study Claims AI Can Read Personality From LinkedIn Photos, Predict Career Success
An NBER working paper claims AI can extract Big Five personality traits from facial photos alone—and that these extracted traits predict career outcomes. Researchers analyzed LinkedIn photos of 96,000 MBA graduates and found the AI-derived personality scores correlated with school prestige, compensation, job transitions, and career advancement. The predictive power reportedly matched that of race, attractiveness, and educational background. The study also found people tend to sort into jobs where their apparent personality traits are valued, earning more when traits align with occupational demands.
Why it matters: If validated, this research raises serious questions about algorithmic discrimination in hiring—suggesting AI systems could encode personality judgments from photos into employment decisions, whether intentionally deployed or not.
What's On The Pod
Some new podcast episodes
The Cognitive Revolution — Approaching the AI Event Horizon? Part 2, w/ Abhi Mahajan, Helen Toner, Jeremie Harris, @8teAPi
AI in Business — In a Sea of Complexity, Does a "Successor" Exist? - with Stephen Wolfram of Wolfram Research
The Cognitive Revolution — Approaching the AI Event Horizon? Part 1, w/ James Zou, Sam Hammond, Shoshannah Tekofsky, @8teAPi