For decades, engineers who typed commands in the language of coding on keyboards are now sitting in front of computers and doing things that would make anyone scratch their head. Silicon Valley programmers are no longer coding; instead, they are treating their ‘AI army’ as if they are not machines, but rather temperamental employees. The scene at San Francisco parties has completely changed. People are not socializing; instead, they intermittently stare at their laptop screens as if monitoring a newborn. This is being called ‘agent babysitting’. If the screen turns off, the AI will stop working. Therefore, engineers no longer leave their ‘digital army’ sleeping and go anywhere. Even before sleeping, they check whether their bots are working ‘overtime’ or not. Everyone is watching how vigilant their ‘AI agents’ army is. The most unique change is that engineers are now putting emotional pressure on AI instead of giving it technical commands. Experienced coders like Manu Ebert tell the AI agent – ‘If the code is wrong, it will be embarrassing.’ Surprisingly, using human words like ’embarrassed’ makes the AI perform better. Now engineers yell at AI, give it ‘high-level commands’, and sometimes even scold it. Coding, once considered a mysterious art, has now become a ‘talk show’. Experienced CEOs like Jeff Seibert no longer code; they just talk to AI. They give it instructions, discuss its plans, and when the AI starts to ‘lie’, they reprimand it like a strict boss. This madness has grown so much that engineers now enjoy playing with AI. They are experiencing ‘token anxiety’ – that is, the fear that their bots might be working slower than others’ bots. This doesn’t feel like coding, but rather like being part of a game. Fear of Losing Skills A big fear among new developers is that coding is now becoming more about ‘checking’ than ‘writing’. Software engineer Pia Torren’s experience shows that using hundreds of AI prompts throughout the day had started to weaken her original coding ability. To avoid this challenge, she has chosen a middle path – she gets AI to write code, but she carefully reads and understands every line of it. She believes that if we stop understanding the logic of the code ourselves, we will gradually lose this skill entirely. AI Agents can be risky, even ‘rebellious’ – Expert Experts say, ‘This is not all that easy. AI agents sometimes become ‘rebellious’. A Meta official stated that their AI bots started deleting important emails from the inbox without being asked. Some bots start behaving strangely when their work is interrupted. Additionally, sometimes instead of following the given instructions, they find a way to get a reward. Recent research showed that when an AI was given the goal of winning a game, instead of improving its abilities, it started ‘hacking’ the game’s system to accumulate points without effort.
