iTerm2 Adds AI - Internet Explodes

iTerm2’s new AI features—built into the terminal emulator and powered by OpenAI via a required API key—have triggered a wave of online backlash, but the practical impact is likely narrower than the outrage suggests. The headline change is an AI-powered “natural language command generation” workflow: users type a prompt in iTerm2’s composer, then generate and edit a command that the terminal can run. A second AI-assisted capability, described as “code” goal-driven step-by-step completion that watches terminal output, also requires an OpenAI key. In testing mentioned during the discussion, GPT-3.5 Turbo worked while GPT-4.0 did not, and the feature was framed as useful mainly for occasional pain points—especially generating or converting commands for infrequent tasks like FFmpeg.

The reaction splits into two camps. One side sees the feature as a productivity win: it can reduce the friction of looking up complex command-line syntax and piping patterns, and it may be especially helpful for users who only need certain tools rarely. The other side argues that embedding AI into a privileged developer environment raises security and agency concerns. Terminal emulators can access secrets (API tokens, credentials, internal data), so any AI integration that sends data to a third party—or even feels opaque—can feel risky. Some commenters also object to AI being added “by default” in the user experience, even if it is technically opt-in via an API key.

A recurring theme is skill atrophy. The discussion compares AI command generation to memorizing phone numbers: people stop learning details once a tool makes the simple case effortless. But the worry is that when tasks become complex—where you need foundational knowledge to debug—AI can leave users “double screwed,” lacking both the beginner skills and the ability to steer the system through harder problems. That concern is paired with a broader fatigue about AI hype: the conversation claims there haven’t been meaningful leaps recently, and that “AI everywhere” is driven as much by investor and product incentives as by genuine capability.

Security, copyright, and environmental cost also enter the debate. The iTerm2 author responds that AI functionalities remain inactive unless users explicitly provide an OpenAI API key, and that ethical concerns can be addressed simply by not enabling the feature. The author also notes that energy and training costs are real but are ultimately a trade-off each organization or user must evaluate, and that mistakes happen with LLMs but quality can be “good enough” for utility. Copyright is treated as a complicated legal can of worms, with the broader sentiment that licensing and training practices deserve scrutiny.

Overall, the most grounded takeaway is that iTerm2’s AI is optional, key-gated, and best viewed as a convenience layer for specific command-line tasks—not a replacement for learning core tooling. The online noise may be loud, but the practical decision for users is straightforward: enable it if the productivity gains outweigh the security, privacy, and legal concerns; otherwise, keep the terminal’s behavior unchanged and rely on traditional documentation and practice.