What even is an OS now?
A prominent voice in tech declares AI is irrevocably redefining the very nature of computing, questioning what an operating system even is when users can 'conjure' bespoke applications. This provocative stance, coupled with a veiled startup announcement for a new AI-first phone, ignites fervent debate among HN users about the future of software, security, and the role of the 'app' itself. It challenges fundamental assumptions about how we interact with technology and who creates it, resonating deeply with a technically sophisticated audience grappling with AI's implications.
The Lowdown
The author, transitioning from Fly.io to a new venture, asserts that the profound impact of AI on computing remains largely unappreciated, even as it obliterates traditional boundaries in software development and the very definition of an operating system.
- AI is dissolving the lines between backend/frontend, web/native, systems/applications, and most importantly, between programmers and users.
- The author envisions a future where individuals use English to "conjure" their own highly specific applications, addressing problems too niche for commercial software.
- This shift means software will increasingly be for audiences of 1-2 people, distributed not as prefabricated apps, but as building blocks.
- The traditional role of a modern OS—partitioning and controlling applications from "strangers"—becomes less relevant when users are the primary creators of their own malleable software.
- His new project is to build a phone designed for this world, one where the device itself enables app creation on demand, rather than merely running fixed-function applications.
- He acknowledges the common "startup announcement" skepticism but emphasizes that the industry is experiencing a destabilization akin to the advent of the personal computer, with AI rapidly breaking every perceived limit.
The piece concludes by urging a return to a childlike imagination about computers, unlearning the limitations imposed by decades of traditional software development, as AI empowers computing to finally match our earliest, wildest dreams.
The Gossip
Operating Systems, Old Definitions, New Realities
The title of the article sparks a debate about the fundamental definition of an operating system in the age of AI. Some commenters argue that the core purpose of an OS remains unchanged, while others, including the author, contend that its role must evolve significantly, particularly concerning permissions, sandboxing, and resource management for highly malleable, user-generated software. The discussion touches on whether the OS itself becomes invisible or needs re-architecting for 'conjurable' experiences.
Agent or Application: The AI Dichotomy
A central discussion revolves around whether the future of interaction with AI will primarily involve a single, all-encompassing AI assistant, or if users will still want and create bespoke applications generated by AI. Proponents of the 'AI assistant' model suggest that asking the AI to perform tasks directly is more efficient than having it generate an app. However, others argue for the necessity of deterministic, repeatable results that applications offer, contrasting with the potentially non-deterministic nature of direct LLM interactions. A hybrid approach, where AI agents build and manage applications, is also proposed.
Security Skepticism and Software Provenance
The concept of AI-generated, user-conjured software immediately raises concerns about security and trust. Commenters question how an OS should handle software whose 'provenance' is an LLM, even if prompted by the user. The consensus points to a pressing need for new OS design innovations, particularly in permissions, access control, and sandboxing, to ensure safety and accountability in a world of highly fluid and personalized code, rather than simply relaxing existing security models.
Startup Stigma and the Slop Squabble
The author's self-aware framing of the post as a 'startup announcement' elicits comments referencing his prior online persona and the common trope. A term like 'slop' is introduced (and later clarified by the author as an external comment), sparking a minor debate on the perceived quality and value of small, user-generated applications. The author pushes back on the notion that solving niche, everyday problems with bespoke software is somehow less 'serious' or merely a 'toy,' drawing parallels to the widespread and often underestimated productivity of tools like Excel.