OpenAI CEO Sam Altman deleted TikTok from his devices after becoming addicted to the platform, he revealed during a recent podcast appearance reported May 27 local time by Business Insider. Altman began using TikTok to better understand its recommendation algorithms while OpenAI developed Sora, an AI-based video social media app similar to TikTok. The CEO's usage escalated from an intended 10 minutes before bed to sessions lasting up to three hours, prompting him to remove the app entirely despite initially believing he could maintain control over his consumption habits.
Altman Details TikTok Usage Escalation From Research to Addiction
Altman told the podcast he "made myself addicted to TikTok to learn" and quickly understood why the platform's recommendation algorithm is considered industry-leading. He initially thought he maintained control, using the app for approximately 10 minutes before sleep as a relaxation tool. However, usage time gradually increased—one night extending to one hour, and on a Saturday afternoon, Altman spent approximately three hours watching TikTok videos while sitting on his couch. He described the experience as "very enjoyable in the moment, like a drug," despite recognizing the habit was unhealthy. After a period of regaining control and limiting usage to 5-10 minutes nightly before bed, Altman ultimately decided to stop entirely. OpenAI discontinued its Sora app approximately six months after launch in March, announcing the decision to focus on core business operations.
Hedge Funds Allocate Capital to Citadel Alumni-Managed Firms
Major US hedge funds are expanding their search for top traders and investment ideas beyond internal resources to external managers, particularly those with Citadel backgrounds, Business Insider reported May 27 local time. According to data from institutional investor data provider Allwellabs, regulatory filings, and industry sources, at least 12 funds managed by former Citadel traders are receiving capital from asset managers. Multi-strategy firms are investing in external talent pools, with most Citadel alumni-run firms known as equity long-short investors trading in market-neutral strategies. Millennium, Schonfeld, Balyasny, and Qube have deployed capital into funds managed by former Citadel personnel, according to the report.
PwC CEO Griggs Credits Curiosity and Extra Effort for Career Advancement
PwC US CEO Paul Griggs told Business Insider in a May 27 local time interview that career success principles remain unchanged despite AI transforming work environments, emphasizing "you get out what you put in." Griggs, who started as a PwC intern approximately 30 years ago, attributed his advancement to CEO to proactive engagement when opportunities arose and self-directed learning during off-hours. He stressed the importance of investing time when others are not watching, stating that people who maintain curiosity and seek additional learning ultimately advance further. Griggs cited a former partner who drew a line down the center of notes—recording meeting content on one side and questions requiring further study or research on the other—demonstrating how post-meeting self-directed answer-seeking built competitive advantage. He told new employees that differences in weekend time usage create significant gaps over time: one person rests after Friday departure while another reviews work-related content, resulting in the latter understanding tasks faster on Monday and naturally receiving more opportunities. Griggs emphasized these small repeated differences lead to increased responsibility, new roles, and eventual promotion. Visa Group President Oliver Jenkins similarly invests weekly time learning new skills and prioritizes curiosity in hiring evaluations.
Anthropic Developer Recommends Goal-Based AI Instructions Over Step-by-Step Commands
Boris Chernyi, developer of Anthropic's AI coding agent Claude Code, advised users to provide goals and criteria rather than detailed step-by-step instructions when using advanced AI models, Business Insider reported May 27 local time. Chernyi stated many users underestimate AI model capabilities and make the mistake of issuing overly specific sequential instructions like "do this, then do that, and definitely follow steps 1, 2, 3, 4 in order." He explained this approach is inappropriate for the latest AI models. Instead, clearly presenting task objectives, defining constraints (guardrails) and success criteria, then allowing the AI to determine work methods independently produces better results. Chernyi stated: "You should describe the task, set guardrails, define success criteria, and then let the AI work on its own. Come back after a while and you'll get results better than expected." He assessed that past prompting methods may become increasingly outdated as AI model performance rapidly improves. Chernyi noted that dozens or hundreds of additional useful capabilities likely remain undiscovered, as new use cases continue emerging during internal Anthropic model experimentation. He emphasized that providing appropriate goals and criteria while granting autonomy, rather than excessively controlling AI, represents the effective method for utilizing the latest AI systems.
US States Lower Income Tax Rates While Raising Sales Tax Burdens
Many US state governments are lowering taxes on high earners while raising taxes on everyday goods, moving in the opposite direction of voter demands for increased taxation on ultra-high-income individuals and corporations. According to an Institute on Taxation and Economic Policy (ITEP) analysis cited by Business Insider on May 26 local time, top income tax rates have generally declined since 1990 while sales and excise taxes increased, shifting tax burdens from high earners and corporations to consumer goods. The US Census Bureau noted some states are moving toward flat income tax rates or eliminating income taxes entirely, while sales taxes have become the largest component of state revenue. Tax Foundation senior policy analyst Avir Mandal explained this trend fueled competition to attract workers from states with high income tax rates like California and New York, with states like North Carolina seizing opportunities to become attractive to both businesses and individuals. Mandal added that Mississippi and Oklahoma wanted to join this trend after North Carolina and South Carolina succeeded in attracting workers and businesses. He noted arguments exist that raising sales taxes over income taxes makes more sense, as income tax increases can induce income reductions and income volatility creates revenue uncertainty. However, this strategy carries risks. In 2025, Louisiana reversed previous tax cuts, raising sales tax from 4.45% to 5% while lowering income tax to a 3% flat rate, resulting in budget deficits due to reduced income and corporate tax collections.
FAQ
Why did Sam Altman delete TikTok from his devices?
Altman deleted TikTok after losing control over his usage time. He initially intended to use the app for approximately 10 minutes before bed but found his consumption escalating to one hour some nights and approximately three hours during weekend afternoons. Despite recognizing the habit was unhealthy and attempting to regain control, he ultimately decided to remove the app entirely.
What investment trend are major hedge funds following regarding Citadel alumni?
Major US hedge funds including Millennium, Schonfeld, Balyasny, and Qube are allocating capital to funds managed by former Citadel traders. According to data reported May 27 local time, at least 12 funds run by Citadel alumni are receiving investments from asset managers, with most operating as equity long-short investors using market-neutral trading strategies.
How does Anthropic's developer recommend using advanced AI models effectively?
Boris Chernyi, Claude Code developer, advises users to provide task objectives, define constraints (guardrails), and establish success criteria, then allow the AI to determine work methods independently rather than issuing detailed step-by-step instructions. He stated this approach produces better results with the latest AI models compared to overly specific sequential commands.
