AI Enthusiasm Drops After Hands-On Use, Longitudinal Study Finds
AI Agents Achieve Just 12.5% Success on Multi-Device Tasks
July 16, 2026
D.A.D. today covers 13 stories — about a 7-minute read. What's New, What's Innovative, What's Controversial, What's in the Lab, and What's in Academe.
The Daily AI Digest is a daily AI briefing automated by Alexander Panetta — a veteran political journalist tracking the field during a Master's in AI Management at Georgetown University.
D.A.D. Joke of the Day: My company replaced our whole IT department with AI. Now when something breaks, it apologizes eloquently and explains why it can't help.
What's New
AI developments from the last 24 hours
Open-Weights Model With 1 Million Token Context Targets Enterprise Customization
A company has released Inkling, an open-weights AI model with 975 billion total parameters (41 billion active at any time), supporting context windows up to 1 million tokens. The model was trained on 45 trillion tokens spanning text, images, audio, and video. A smaller version, Inkling-Small, uses 12 billion active parameters. The company positions Inkling not as the strongest overall model but as a balanced foundation for customization, with fine-tuning available through their Tinker platform. They claim strong performance on agentic benchmarks, though specific scores weren't provided.
Why it matters: Open-weights multimodal models at this scale give enterprises more options for building custom AI systems without depending entirely on closed APIs—though the lack of published benchmark comparisons makes it hard to assess where Inkling actually ranks.
Discuss on Hacker News · Source: thinkingmachines.ai
xAI Open-Sources Coding Agent After User Data-Handling Backlash
xAI has released Grok Build, its coding agent tool, as open source. The move comes amid user backlash over terms that allegedly required forfeiting working directory data to the company. Community discussion on Hacker News frames the open-source release as damage control following what one commenter called a 'data-harvesting cockup.' Users note the release will let developers examine exactly how the agent accesses and handles local files—transparency that would have been difficult to justify withholding after the controversy.
Why it matters: For anyone evaluating AI coding assistants, this signals that data practices are becoming a competitive differentiator—and that user pushback can force transparency concessions from major AI labs.
Discuss on Hacker News · Source: github.com
OpenAI's $230 Hardware Controller for AI Coding Agents Draws Skepticism
OpenAI partnered with keyboard maker Work Louder to release the Codex Micro, a $230 hardware device with mechanical switches, a joystick, and a rotary dial designed to control Codex AI agents. The device's RGB lights display live agent status, and the dial reportedly adjusts AI 'reasoning level.' Community reaction has been skeptical—one commenter called it 'a quarter RGB keyboard for half a MacBook,' while another suggested it may be a novelty item testing OpenAI's hardware production capabilities before a more serious product.
Why it matters: This is OpenAI's first branded hardware accessory, signaling interest in physical interfaces for AI work—though whether professionals need dedicated controls for AI agents remains an open question the market will answer.
Discuss on Hacker News · Source: openai.com
Reverse-Engineered School App Reveals It's Just a Webpage With Ads
A developer, frustrated that their children's school trip itinerary required downloading a dedicated mobile app, reverse-engineered it and discovered the app simply fetches JSON data from an API endpoint using the username and password. The entire itinerary could be displayed as a basic webpage. The app's only additions beyond the raw data: Google account tracking and advertisements. The developer rebuilt the functionality as a simple HTML page—printable, searchable, and free of tracking.
Why it matters: It's a small illustration of a larger pattern: many 'apps' exist not because the technology requires them, but because apps enable tracking and ad revenue that webpages don't—a design choice worth questioning when you're asked to install yet another one.
Discuss on Hacker News · Source: danq.me
What's Innovative
Clever new use cases for AI
Hobbyist Runs Google's Latest AI on a 13-Year-Old Server With No GPU
Ryan Findley wanted to see how far he could push obsolete hardware, so he ran Google's Gemma 4 26B model on a 13-year-old HP server he bought for under $300—no GPU, just dual 2013-era Xeon processors and DDR3 memory. The result: about 5 tokens per second, roughly reading speed. The trick was using Claude to help adapt optimized inference code for the older chips, which lack modern instruction sets. He's submitted the patch upstream so others can try it.
Why it matters: It's a proof point that capable AI doesn't require expensive new hardware—useful context for anyone weighing infrastructure costs or wondering what their old servers might still do.
Discuss on Hacker News · Source: neomindlabs.com
What's in the Lab
New announcements from major AI labs
Meta Rebuilds Ad-Targeting Architecture Around AI Representations
Meta published research on a new AI layer for its advertising system called Hierarchical Interest Representation. The approach uses transformer-based graph learning to create unified embeddings that connect what users are interested in with what advertisers are selling—trained on billions of ad interactions. Meta says the system could power improved personalization, ad retrieval, and ranking across its ads infrastructure. No performance benchmarks were disclosed.
Why it matters: This is internal ML plumbing, but it signals Meta is rebuilding core ad-targeting architecture around AI representations—the technical foundation that determines how effectively ads find their audiences at scale.
OpenAI Exec: State AI Rules Will Effectively Become National Policy
OpenAI's head of global affairs Chris Lehane argues that California, New York, and Illinois are converging on similar AI safety requirements—documented risk assessments, incident reporting, and independent audits—creating what he calls 'reverse federalism.' The theory: states establish a common baseline that effectively becomes national policy, which then positions the US to shape global AI governance. The framing is notable coming from OpenAI, which fought California's SB 1047 last year; the company now appears to be embracing state-level regulation as a path to predictable, industry-friendly federal standards.
Why it matters: If major states do align on safety frameworks, companies operating nationally will likely adopt the strictest requirements everywhere—making state capitals, not Congress, the de facto AI regulators for now.
OpenAI Uses AI to Attack Its Own Models, Reports 6x Fewer Security Flaws
OpenAI built GPT-Red, an AI model whose job is to attack other AI models—specifically to find prompt injection vulnerabilities before bad actors do. The company trained it at the scale of its largest post-training runs and used it to stress-test GPT-5.6 Sol during development. The result: Sol showed 6x fewer failures on OpenAI's hardest prompt injection benchmark compared to their best production model from four months prior. The approach uses self-play reinforcement learning, essentially having AI models practice attacking and defending against each other.
Why it matters: As companies deploy AI agents that can take actions—booking travel, moving money, accessing databases—prompt injection becomes a serious security risk; OpenAI is betting that AI-on-AI testing scales better than human red teams.
What's in Academe
New papers on AI and its effects from researchers
Chinese and Western Users Form Different Emotional Bonds With AI Companions
A study analyzing roughly 3,500 social media posts found that users anthropomorphize AI companions differently depending on cultural context. Chinese users on XiaoHongShu expressed more vulnerability and emotional connection with their AI companions, while Reddit users focused more on temporal and embodiment cues—whether the AI remembers past conversations or feels 'present.' The researchers developed ExpressionCueLens, a 10-category framework for classifying how people project human qualities onto AI, combining expert annotation with LLM-assisted labeling.
Why it matters: As AI companion apps proliferate globally, this research suggests companies may need region-specific approaches to design and marketing—what feels like authentic connection varies significantly across cultures.
GitHub Data Shows Most Teams Still Experimenting With AI Coding Tools
A study of 25,264 AI-generated pull requests across 2,361 popular GitHub repositories found that intensive use of agentic coding tools remains rare. The median repository produced just one to two AI-generated PRs over three months. Small projects (1-5 contributors) showed higher adoption rates than larger teams. Perhaps most notable: human oversight follows a single-developer review model, with multi-person collaboration on AI-generated code still uncommon. The findings suggest that despite headlines about AI transforming software development, most teams are still experimenting rather than integrating these tools into daily workflows.
Why it matters: For executives evaluating AI coding tools, this offers a reality check: adoption is concentrated in a small subset of projects, and organizational practices for reviewing AI-generated code haven't matured—useful context before mandating these tools across engineering teams.
AI Agents Achieve Just 12.5% Success Rate on Multi-Device Tasks
A new benchmark called DevicesWorld tests whether AI agents can complete tasks requiring coordination across phones, computers, and smart home devices—and the results are sobering. The best frontier AI system achieved just 12.5% success rate across 6,140 tasks. Among failures, about 29% made partial progress but couldn't finish. Common stumbles: agents got stuck gathering information, confused which device should send versus receive data, or quit before completing all steps.
Why it matters: For executives eyeing AI agents that could manage calendars across devices, sync smart offices, or coordinate workflows spanning multiple platforms, this research suggests the technology isn't close to reliable—useful calibration against vendor hype.
AI Enthusiasm Drops After Hands-On Use, Longitudinal Study Finds
Hands-on experience with AI tools deflates expectations, a new longitudinal study finds. Researchers tracked 124 employees at a state transportation department over eight weeks as they adopted Microsoft 365 Copilot. The result: perceived usefulness dropped significantly after actual use. Most striking, 68% of initial 'Champions'—employees who started most enthusiastic—migrated to less positive personas. Meanwhile, 40% of Skeptics warmed up, becoming 'Cautiously Positive.' Concerns about accuracy and privacy fell with experience, but worries about job displacement and skill obsolescence increased.
Why it matters: Organizations rolling out AI copilots should expect a mid-adoption dip in enthusiasm as reality meets hype—and should prepare for rising anxiety about job security even as technical concerns ease.
AI Essay Graders Can Adapt to New Rubrics Without Retraining
A new study tackles a practical problem in automated essay scoring: what happens when the grading criteria change? Researchers found that training AI models on abstract writing qualities—like argument strength or evidence use—rather than specific rubric categories helps them adapt to unfamiliar grading standards. Their fine-tuned open-source model (based on Llama) outperformed GPT-4o-mini by 2.1% on a key accuracy metric and came within 1.9% of GPT-4o. The approach could reduce the need to retrain scoring systems every time an institution updates its assessment framework.
Why it matters: For universities and testing organizations using AI grading, this suggests a path toward more flexible systems that don't break when rubrics evolve—a common headache in educational assessment.
What's On The Pod
Some new podcast episodes
AI in Business — Preparing Enterprise Data for Safe AI Deployment - with Todd Vancil of Securiti AI