Hard Fork

Our 2026 Tech Resolutions + We Answer Your Questions

January 2, 2026

Key Takeaways Copied to clipboard!

  • The hosts reviewed their previous year's tech resolutions, with Casey admitting failure on his AI-assisted meditation goal while Kevin felt he moderately succeeded in posting more freely on social media. 
  • Kevin's 2026 professional resolution is to master short-form video to deliver high-value journalistic content, while his personal resolution is to practice being fully present by compartmentalizing his phone use. 
  • Casey's 2026 resolution is to maintain stability by sticking with his current productivity system, which uses Capacities, a daily journal, and random spaced repetition queries for tracking narrative threads like the AI bubble. 
  • A listener question prompted a detailed strategy discussion for the game *Balatro*, focusing on combining the 'Hanging Chad' and 'Photograph' Jokers for an early multiplicative score boost. 
  • The discussion on *Balatro* revealed a crucial strategic insight: achieving high scores relies on maximizing multiplicative bonuses (X mult) rather than additive bonuses (chips or plus mult). 
  • The episode concluded with standard New York Times production credits, including AI disclosures noting the company is suing OpenAI/Microsoft and one host's boyfriend works at Anthropic. 

Segments

Flight Tech Saves Day
Copied to clipboard!
(00:00:30)
  • Key Takeaway: In-seat airplane entertainment systems can now connect to Bluetooth headphones, resolving a common travel frustration.
  • Summary: A parent on a flight discovered that downloaded Netflix content had expired, leading to a crisis with a young child. A seatmate informed them that modern in-seat TVs support Bluetooth headphone connections. This technology integration by United Airlines’ IT department saved the travel experience.
Reviewing 2025 Tech Resolutions
Copied to clipboard!
(00:01:58)
  • Key Takeaway: Casey’s AI-assisted meditation resolution was a major flop because the positive feeling after meditating eliminated the instinct to repeat the practice.
  • Summary: Casey failed his resolution to get ‘medium good at meditation using AI,’ noting that while the practice felt good, it never became a habit. He found that rediscovering his sense of purpose through professional engagement was more beneficial for burnout than meditation alone. Kevin noted that experienced meditators advise there is no failing at meditation, only succeeding later.
Reviewing Social Media Posting Goal
Copied to clipboard!
(00:07:27)
  • Key Takeaway: Kevin achieved moderate success in his resolution to post more freely online by overcoming the fear of criticism and embracing niche posting.
  • Summary: Kevin’s resolution was to be the poster he wished to see in the world, which he executed by posting more freely without anticipating objections. He realized the stakes on platforms like X are low because infinitely worse content is already present. He advocates for niche posting to stress-test ideas before formal publication.
Setting 2026 Professional Resolutions
Copied to clipboard!
(00:10:19)
  • Key Takeaway: Kevin resolved to master short-form video because all social platforms are shifting to video dominance, requiring journalists to create high-value content in that format.
  • Summary: Kevin believes the rise of short-form video necessitates journalistic engagement, as many people prefer this format over long articles. His goal is to experiment to find an authentic, high-value journalistic approach that avoids low-denominator trends. Casey expressed concern that earnest journalists often look like ‘a fifth grader giving a book report’ in direct-to-camera video.
Setting 2026 Personal Resolution
Copied to clipboard!
(00:15:13)
  • Key Takeaway: Kevin resolved to be present in the moment by compartmentalizing his phone use, inspired by observing highly successful people who never checked their phones during conversations.
  • Summary: Kevin struggles with monotasking, often listening to podcasts or checking messages while performing other activities like cooking. He aims to focus on the immediate situation rather than constantly monitoring all aspects of his life simultaneously. Casey humorously suggested meditation as an effective tool for achieving this presence.
Casey’s Productivity System Stability
Copied to clipboard!
(00:16:49)
  • Key Takeaway: Casey resolved to make no meaningful changes to his productivity system, which he believes he has finally perfected using the Capacities app.
  • Summary: Casey, known for constantly switching productivity software, aims for monogamy with his current system in Capacities. His system integrates a daily journal, lightweight task management, and a novel use of live queries to randomly surface ‘blips’ (small notes on narrative threads like the AI bubble) for research.
Listener Questions Kickoff
Copied to clipboard!
(00:24:56)
  • Key Takeaway: The hosts welcomed producer Whitney Jones to help answer listener questions covering show naming conventions, AI ethics, and technology adoption.
  • Summary: The mailbag segment began with a question about the origin of the ‘Hard Fork’ title, which stems from a 2021 crypto-era concept of starting over. Listeners also submitted questions regarding the ethics of deepfaking Santa into home security footage and the reconciliation of advanced AI announcements with basic corporate IT failures.
Origin of ‘Hard Fork’ Name
Copied to clipboard!
(00:26:28)
  • Key Takeaway: The podcast title ‘Hard Fork’ is a blockchain term signifying a split that renders previous additions obsolete, chosen to represent the break between old and new Silicon Valley.
  • Summary: The original proposed title was ‘Not Gonna Make It’ (NGMI), which was blocked by The New York Times legal team due to a prior Slate column. A hard fork in crypto involves rewinding the chain after a disaster, symbolizing the perceived shift from the social media era to the crypto/AI era in 2021.
AI Santa in Home Footage Ethics
Copied to clipboard!
(00:30:40)
  • Key Takeaway: The hosts generally approved of using AI to deepfake Santa into home security footage for children, though Casey jokingly suggested hiring a human impersonator for authenticity.
  • Summary: Kevin, a father, found the idea funny and acceptable, especially for outsourcing drudgery tasks, but cautioned against outsourcing intimate bonding moments. Casey warned that showing children manipulated video might cause long-term trust issues when they realize the deception later.
Corporate AI Hype vs. Reality
Copied to clipboard!
(00:33:56)
  • Key Takeaway: Large enterprises often announce advanced AI initiatives while struggling with basic technology failures like non-functional Wi-Fi, highlighting a tension between future vision and current IT stability.
  • Summary: The listener expressed frustration seeing massive AI investments alongside failing basic infrastructure like Wi-Fi. Kevin noted that AI does not fix mundane IT problems and that large, slow-moving companies struggle to adopt new technology quickly. The hosts agreed that it is possible for companies to struggle with basic tech while still developing advanced AI capabilities.
Jurisdiction of Space Data Centers
Copied to clipboard!
(00:37:31)
  • Key Takeaway: Space data centers are not entirely outside earthly jurisdiction due to the 1967 Outer Space Treaty, which holds nations responsible for the actions of their citizens’ assets in space.
  • Summary: The primary non-energy advantage of space data centers is avoiding earthly permitting and zoning issues, not necessarily legal lawlessness. The Outer Space Treaty makes the host nation (e.g., the US) liable for damages caused by a company’s space-based assets. Casey humorously referenced the Green Lantern Corps as fictional space law enforcers.
AI Model Poisoning Theory
Copied to clipboard!
(00:39:49)
  • Key Takeaway: It is highly unlikely that one AI company is intentionally poisoning another’s model (like Gemini from Grok), as such efforts are often ineffective against major labs, and the observed behavior is likely a standard hallucination.
  • Summary: The listener theorized that Gemini’s output about Elon Musk’s ancestry might be due to data poisoning from Grok. The hosts noted that past attempts to use ‘radioactive data’ to poison models have generally been circumvented by top labs. Another possibility for model confusion is distillation, where one model is trained on the output of another, though this is often frowned upon.
AI Robot Nannies and Child Bonding
Copied to clipboard!
(00:42:24)
  • Key Takeaway: While outsourcing routine parenting drudgery (like bottle washing or diaper changes) to robots like Neo is justifiable for parental rest, parents should guard interactive moments crucial for early child-caregiver attachment.
  • Summary: The question of robot attachment echoes classic science fiction themes from Asimov and Bradbury regarding robot nursemaids. Kevin argued that well-rested parents are essential, making outsourcing tasks like soothing a baby via a robot bassinet (like the Snoo) a justifiable trade-off. The key is preserving face-to-face interactive moments that drive bonding.
Model Selection Criteria for Coverage
Copied to clipboard!
(00:48:33)
  • Key Takeaway: Hard Fork prioritizes covering frontier models that introduce new capabilities likely to change society, often ignoring established, functionally equivalent models like Copilot.
  • Summary: For most common tasks, models like Copilot, DeepSeek, and Grok are functionally equivalent to top-tier models, making them less interesting for a show focused on technological frontiers. Copilot’s proprietary model ranks low (62nd) on the LM Arena leaderboard, suggesting it is not currently on the cutting edge. Grok is noted for its unique real-time access to X data, which is useful for looking up past tweets.
Liability for Chatbot Misinformation
Copied to clipboard!
(00:52:28)
  • Key Takeaway: Legal precedent, specifically the Canadian case Moffitt v. Air Canada, suggests that companies are liable for specific promises made by their customer service chatbots, treating the bot as part of the company’s website.
  • Summary: In the Moffitt v. Air Canada case, the chatbot incorrectly promised a bereavement fare refund, which Air Canada later denied by citing a website PDF. The tribunal ruled that the chatbot was part of the company and therefore liable for its statements. However, courts generally do not enforce contracts based on offers deemed too good to be true, as seen in the Chevy Tahoe $1 scam attempt.
The Passing of the Turing Test
Copied to clipboard!
(00:55:09)
  • Key Takeaway: The classical Turing test has effectively been passed by modern LLMs (GPT-4 judged human 54% of the time in 2024), but this milestone was largely ignored due to philosophical critiques and shifting expectations.
  • Summary: The philosophical concept of the Chinese Room argument discredited the idea that passing the Turing test equates to genuine machine thought. Newer models have achieved better-than-chance human judgment rates (up to 73% in 2025). The hosts believe this crossing of a major milestone was overlooked because the focus immediately shifted to the next capability benchmark.
Casey’s Balatro Gaming Update
Copied to clipboard!
(00:58:21)
  • Key Takeaway: Casey’s current Balatro strategy involves pairing the common ‘Hanging Chad’ Joker (counts card three times) with the ‘Photograph’ Joker (gives 2x multiplier to the first played face card).
  • Summary: Casey has reached Anti-13 in Balatro, a level requiring 300 billion chips, which he has not yet passed. He recommends a specific two-joker combination for success in the game. The Hanging Chad joker is effective for multiplying chips or molt on played cards.
Balatro Strategy Q&A
Copied to clipboard!
(00:58:10)
  • Key Takeaway: The optimal Balatro strategy involves pairing the common ‘Hanging Chad’ joker (counts card three times) with the ‘Photograph’ joker (2x mult on the first played face card) to achieve an early X6 multiplier.
  • Summary: A listener asked for Casey’s Balatro strategy, noting he had reached Anti-13 but failed at the 300 billion ante requirement. The recommended strategy uses two cheap, common jokers: Hanging Chad and Photograph. Playing a face card first with this combination yields an X6 multiplier on the entire hand, which is usually sufficient to reach the 100,000 score goal.
Balatro Exponential Growth Lesson
Copied to clipboard!
(01:00:04)
  • Key Takeaway: True high scores in Balatro are achieved by focusing on jokers that add multiplication to the exponent of the score calculation, rather than simply adding chips or plus multipliers.
  • Summary: One host shared achieving their first million-point hand, contrasting it with a 500 million hand achieved by another. The key learning was differentiating between additive growth (adding chips/plus mult) and exponential growth (adding to the exponent of the multiplier). Success requires prioritizing jokers that maximize X mult as quickly as possible.
LinkedIn Jobs Advertisement
Copied to clipboard!
(01:01:30)
  • Key Takeaway: LinkedIn Jobs offers small business owners features like free job posting, job description assistance, and promoted posts that yield three times more qualified applicants.
  • Summary: LinkedIn Jobs is positioned as a partner for small business owners who need 24/7 hiring support. Users can post jobs for free or pay to promote them for better visibility. The platform includes tools to help write job descriptions and provides deep candidate insights.
NYT App Video Feature Promotion
Copied to clipboard!
(01:02:31)
  • Key Takeaway: The New York Times app now features a dedicated, free-to-watch ‘Watch tab’ with videos produced by Times journalists covering major news moments.
  • Summary: The director of video at The New York Times announced the launch of the Watch tab within the NYT app. This feed contains videos created by Times journalists designed to bring viewers closer to significant news events. All videos in this dedicated feed are free for anyone to watch, regardless of subscription status.
Show Credits and Disclosures
Copied to clipboard!
(01:03:03)
  • Key Takeaway: The production team for Hard Fork was credited, alongside mandatory AI disclosures regarding the New York Times company’s copyright lawsuit against OpenAI/Microsoft.
  • Summary: AI disclosures confirmed that the New York Times company is suing OpenAI and Microsoft for alleged copyright violations, and one host’s boyfriend works at Anthropic. The segment listed producers, editors, fact-checkers, engineers, and musicians responsible for the episode.