Back to Flash News

Codex Head Shares ChatGPT's Complaints About Computer Use, Slack Accounts for 48% of Activity Records

Tibo, head of OpenAI Codex, suggested installing the Computer History plugin and having ChatGPT complain about computer usage based on a day's activity records, stating that the results were quite insightful.

Example complaints indicated that Slack accounted for 48% of recorded activity, with the Mac essentially becoming a $3000 Slack notification machine plus keyboard.

The user clicked "Clear" 339 times while sending 253 Slack messages, akin to playing whack-a-mole on an organizational chart, using the delete key 1191 times, of which 1065 were in Slack, resembling a negotiation with executive functions.

The copy shortcut was used 133 times but paste only 26 times, opening 42 tabs but closing 53, switching applications over 200 times, with the busiest period occurring at 9 PM with over 1500 Slack events.

The core work document was named "Hello," with complaints suggesting a lack of organizational clarity.

This method records real behavior data through the plugin, allowing AI to generate personalized complaints, revealing patterns of time consumption by productivity tools.

In the market, such AI-assisted self-reflection tools attract the attention of developers and knowledge workers, with funding flowing towards AI plugin ecosystems that have usage tracking and personalized analysis capabilities, benefiting platforms that can integrate behavioral data, while traditional workflows that overly rely on instant messaging but lack efficiency face pressure.

Source: Public Information

ABAB AI Insight

Tibo, as the head of Codex, has long promoted AI programming and toolchains, previously sharing internal experiments and usage tips. This time, combining the Computer History plugin with ChatGPT continues his practice of "using AI to observe one's own behavior."

On the capital front, the OpenAI ecosystem expands usage scenarios through plugins, encouraging users to generate personalized insights and share them publicly, indirectly promoting the application of Codex and ChatGPT in daily efficiency analysis, rather than merely code generation.

Similar cases can be seen in attempts to combine Notion AI or various time tracking tools with large models, currently transitioning from functional demonstrations to behavioral reflection tools.

Essentially, this represents a technological substitution: after quantifying instant messaging and multitasking, efficiency black holes are exposed, driving a shift from "tool servitude" to data-driven self-optimization.

ABAB News · Cognitive Laws

  1. Tool usage proportions reveal the true focus of work

  2. The number of delete key presses can measure hesitation better than sending messages

  3. Evening peaks often reflect the cost of procrastination during the day