Our 2026 Tech Resolutions + We Answer Your Questions

Episode Summary

Executive Summary: Hard Fork opens with New Year banter and then revisits the hosts’ prior resolutions: Casey’s meditation goal largely failed, while Kevin’s effort to post more online partly succeeded. They set new 2026 resolutions around short-form video and stable productivity systems, then spend most of the episode answering listener questions about AI ethics, chatbot liability, data centers in space, model poisoning fears, robot caregivers, model selection, the Turing test, and game strategy.

Main Topics: New Year's tech resolutions (Priority: 5/5): The hosts review last year’s goals and set new ones: Kevin wants to get good at short-form video and be more present; Casey wants to avoid changing his productivity system for a year after finally finding one that works. Meditation, burnout, and purpose (Priority: 4/5): Casey reflects on failing to build a meditation habit with AI help, saying the deeper fix for burnout was rest, reconnection with purpose, and better life balance rather than any single tool. Short-form video and journalism (Priority: 5/5): Kevin argues that video is becoming the dominant format online and that journalists need to experiment with authentic, high-value short-form content without chasing gimmicks or virality. AI ethics, trust, and family life (Priority: 5/5): Listener questions probe whether it is ethical to add AI Santa to security footage, whether robots should help with childcare, and how much trust children should place in AI-generated media. AI in business and infrastructure (Priority: 4/5): The hosts discuss the gap between flashy corporate AI initiatives and basic IT failures, plus the idea of space data centers as a response to energy, cooling, and regulatory constraints. AI models, hallucinations, and competition (Priority: 4/5): Questions about Copilot, DeepSeek, Grok, and possible model poisoning lead to a broader discussion of how the hosts choose which models matter and why hallucinations are usually simpler than conspiracy theories. AI milestones and games (Priority: 3/5): The episode closes with reflections on the Turing test’s cultural obsolescence and a playful exchange about Balatro strategy and exponential scoring.

Key Arguments: Casey’s meditation resolution failed not because meditation was useless, but because it never became a habit; burnout improved more through rest and renewed purpose than through AI-assisted reflection. Kevin believes short-form video is becoming central to media consumption, and journalists should learn the medium rather than dismiss it as inherently unserious. The hosts argue that online posting feels safer when one accepts that niche audiences matter and that it is impossible to please everyone. Chatbots can create real liability: if a company’s bot makes a promise, the company may still be responsible, as shown by the Air Canada case. Space data centers are not completely lawless; international treaties and national responsibility still impose constraints, even if regulation avoidance is part of the appeal. Corporate AI adoption is often disconnected from basic operational problems, and AI does not solve mundane infrastructure issues like broken Wi-Fi or legacy software. The Turing test has effectively been passed in its classical form, but its cultural significance faded because people moved the goalposts and because conversational fluency no longer equals intelligence. Robots can be useful for chores and sleep-related baby care, but intimate bonding moments with children should not be fully outsourced. The hosts prioritize discussing frontier models and major capability shifts over covering every chatbot, though they acknowledge the importance of widely used enterprise tools like Copilot.

Data Points: Casey’s meditation resolution: major flop - His 2025 goal of using AI to get better at meditation did not become a lasting practice. Duration of flight with child: 5 hours - Kevin describes a cross-country flight where expired downloads nearly left him with no entertainment for his three-year-old. Casey’s AI productivity system testing period: about 4 months - He has been using his Capacities-based note and journal system long enough to trust it as stable. Random blips shown daily: 5 - Casey’s live query in Capacities randomly selects five “blips” each morning for spaced repetition. LM Arena rank for Copilot’s MAI1: 62nd - Kevin cites the ranking to illustrate that Copilot is not a frontier-leading model. GPT-4 judged human: 54% - Kevin says a May 2024 study found GPT-4 was judged human slightly more than half the time. Newer models judged human: 73% - Kevin cites a 2025 study showing better-than-chance human-likeness in Turing test-style evaluations. Age of children in Santa question: 3 and 6 - A listener asks about using AI-generated Santa footage for kids of these ages. Age of baby in robot-care question: 7 months - A new mother asks whether a humanoid robot might affect attachment development. Years since Outer Space Treaty: 1967 - Used to explain that space activities are still subject to international legal responsibility. Anti level reached in Balatro: 13 - Casey says he has reached Anti 13 but has not yet won beyond that point. Score threshold mentioned for Balatro: 300 billion - He says passing the first ante at that stage requires roughly this amount. Turing-test-human rate: better than chance - The hosts interpret 54% and 73% human judgments as evidence the classical Turing test has been surpassed.

Pivotal Quotes: "I think I finally cracked it, Kev." — Casey Noon: Casey introduces his resolution to stop changing his productivity system for a year. "I want to get good at short-form video." — Kevin Russ: Kevin states his main professional resolution for 2026. "The way it was when I was covering the aftermath of 9-11, and this is the way it is now as I cover the United States of today." — Dan Barry: A New York Times subscription promo emphasizing factual reporting and journalistic mission.

Implications: The episode suggests AI is becoming unavoidable in media, work, parenting, and consumer tech—but success will depend on using it carefully, legally, and with clear boundaries around trust, quality, and human judgment.