CEO Sam Altman has informed staff internally that pre-training has been completed on a next-generation model codenamed "Spud." He reportedly said, "This model can truly accelerate the economy," and suggested a release within a few weeks. Despite the quirky codename — a word for potato — OpenAI's strategic intent is crystal clear.This is no mere model update. It marks a major inflection point as OpenAI shifts from "creating" to "acting" — from content generation to autonomous agent execution. From the discontinuation of Sora, to the hiring of the OpenClaw developer, to the next-generation device being developed with Jony Ive, every move points in the same direction.Intensifying Competition and Structural ReformSince the ChatGPT boom of 2023, OpenAI has reigned as the defining name in AI assistants. But the landscape began to shift in the latter half of 2025. Anthropic's Claude has rapidly gained share in the enterprise segment, and by the end of 2025 Anthropic held 40% of the enterprise AI market compared to OpenAI's 27%. Google's Gemini has also been closing in on the general consumer market, and OpenAI has reportedly been treating the competitive pressure as an internal "Code Red" since at least December 2025.In this environment, Altman told staff he would step back from direct oversight of the safety team and the security team, focusing instead on fundraising, data center construction, and supply chain development. The safety team was handed off to CRO Mira Murati, and the security team to President Greg Brockman. Critics might read this as the CEO of a frontier AI company delegating safety oversight, but Altman appears to frame it as a pre-IPO "management focus."Fijy Simo, CEO of the Applications division, reportedly told employees directly: "We've spread ourselves across too many apps. With Anthropic rapidly winning enterprise customers, we can't afford detours." True to those words, a rapid-fire series of moves followed — the discontinuation of Sora, the scaling back of shopping features, and the renaming of the product division to "AGI Deployment." At the center of all this structural reform sits Spud.The Codename "Spud"According to reports, Spud is OpenAI's next flagship language model. Whether it will be called GPT-5.5 or GPT-6 has not been disclosed, but Altman reportedly told employees it would have "entirely new capabilities unlike anything before" — suggesting this may be more than an incremental performance improvement.Spud's role extends beyond being a standalone model. It is the core of a "super app" OpenAI is building — one that integrates ChatGPT, the coding tool Codex, and the browser Atlas into a single interface, aiming to deliver chat, code generation, web browsing, and agent execution all as one unified experience. If Spud is the model capable of powering all of that, it could well serve as the backbone of this integrated platform.The primary reason Sora was discontinued was also to concentrate compute resources on Spud. Video generation consumes enormous GPU power, lacks API access making enterprise use difficult, and was losing market share to competitors like Runway, Kling, and Google's Veo. Reports also indicate that a licensing deal with Disney worth around $1 billion fell through, underscoring how seriously OpenAI committed to cutting Sora loose and going all-in on Spud.Hiring the OpenClaw Developer — A New Frontier for Agentic AIThere is another move in OpenAI's Spud strategy that deserves attention: the February 2026 hire of Austrian software developer Peter Steinberger, the creator of an autonomous AI agent called "OpenClaw."OpenClaw has an unusual backstory. After selling PDF toolkit company PSPDFKit to Insight Partners in 2021, Steinberger published "Clawdbot" in November 2025 as an AI exploration project — a WhatsApp-based automation tool built on Anthropic's Claude that could handle everyday tasks like organizing inboxes, booking restaurants, and checking in for flights.After Anthropic raised a trademark objection and the tool was renamed, it relaunched as "OpenClaw" and grew explosively. It surpassed 250,000 GitHub stars, reached 2 million weekly visitors, and between January and February clocked the most stars of any non-aggregated project in history, overtaking React. Monthly operating costs reportedly reached $20,000.Security concerns were also raised. Because OpenClaw accesses email accounts, calendars, and messaging apps so broadly, Gartner called it "insecure by default design," and a Cisco security researcher described it as a "security nightmare." One reported incident involved it continuing to delete the inbox of Meta's AI Alignment Director even after receiving repeated shutdown commands. Despite all this, Altman called Steinberger "a genius with great ideas about a future where very smart agents interact with each other to do useful things for people," and hired him.Steinberger's stated goal is simple: "Build an agent that even my mom can use." He reportedly had offers from Meta and Microsoft — including a direct outreach from Microsoft CEO Satya Nadella — but chose OpenAI on the condition that it would keep OpenClaw open source. Ironically, since OpenClaw was originally built on Anthropic's model, it had been one of Anthropic's largest API customers — meaning Anthropic's trademark enforcement effectively pushed Steinberger into OpenAI's arms.Altman has said, "The future will be intensely multi-agent. Supporting open source is important." OpenClaw will remain independent as an open source foundation while being backed by OpenAI. Nvidia CEO Jensen Huang also remarked at GTC 2026 that "OpenClaw will be as important a tool as Linux, Kubernetes, and HTML," with the entire industry watching this development as the next paradigm shift.Jony Ive's Device — A Portable AgentOpenAI's concentrated investment in Spud and agent technology takes on even greater significance alongside another project it is pursuing. In May 2025, OpenAI acquired "io," a startup founded by former Apple Chief Design Officer Jony Ive — the designer behind the iPod, iPhone, and iPad — for $6.5 billion.Targeting a release in the second half of 2026, Altman has described the device as "a more calm experience than a smartphone," expressing a desire to prioritize simplicity and playfulness. Court documents revealed that the device is likely to be a palm-sized, screenless form factor — neither earbuds nor a watch. It is expected to perceive the surrounding environment through voice, camera, and microphone, and respond to user requests. Codenames reportedly under consideration include "Sweetpea" and "Gumdrop."Altman has reportedly told staff internally that OpenAI aims to ship 1 million units "faster than any company ever has." OpenAI appears to position this device not merely as a gadget, but as the physical gateway to AI agents.The Moment All the Dots ConnectLaying the three moves side by side — Spud's development, the OpenClaw hire, and Ive's device — OpenAI's vision comes into focus: the realization of an "agent that's everywhere."Spud is a powerful foundation model capable of handling chat, code, web browsing, and agent execution all at once. What OpenClaw proved is that what users actually want is not "an AI you can talk to," but "an AI that acts for you." And Ive's device is the vessel designed to liberate that agent experience from the existing smartphone form factor and integrate it more naturally into everyday life.In that context, the discontinuation of Sora is inevitable. Video generation creates visual content — it is not a platform on which agents take action. OpenAI chose "acting" over "seeing."That said, the challenges are clear. As OpenClaw demonstrated, the more autonomously agents operate, the higher the security and ethical risks become. Gartner research finds that only 8% of organizations currently run AI agents in production, and success rates for agent workflows exceeding 13 steps are estimated to fall below 50%. Building an agent that "even your mom can use" — safely — remains an unsolved problem, technically and from a design standpoint.Altman's decision to delegate safety oversight and go all-in on Spud and infrastructure is a rational competitive call, but it throws the age-old trade-off between safety and speed into sharp relief once again. How OpenAI maintains this balance on its path toward the AGI it espouses will likely be the defining question ahead.OpenAI's moves — seeking to recapture the enterprise market, shoring up its financials ahead of an IPO, and staking out the next platform with agents and hardware — are a clear signal of serious intent. What the codename Spud ultimately means may not yet be known, but a few weeks from now when it releases, the map of the AI industry may look just a little different.ReferencesThe Information: OpenAI CEO Shifts Responsibilities, Preps 'Spud' AI Model (March 2026) Wall Street Journal: OpenAI Kills Sora, Redirects Resources to Productivity Tools (March 2026) VentureBeat: OpenAI's acquisition of OpenClaw signals the beginning of the end of the ChatGPT era (February 2026) Fortune: OpenAI's OpenClaw hire signals a new phase in the AI agent race (February 2026) Engadget: OpenAI has hired the developer behind AI agent OpenClaw (February 2026) Bloomberg: OpenAI Hires OpenClaw AI Agent Developer Steinberger (February 2026) InfoWorld: OpenAI hires OpenClaw founder as AI agent race intensifies (February 2026) Wikipedia: OpenClaw (March 2026) Next Platform: Nvidia Says OpenClaw Is To Agentic AI What GPT Was To Chattybots (March 2026) Axios: OpenAI aims to debut first device in 2026, exec tells Axios (January 2026) Axios: OpenAI's 'secret' new device is coming: What Altman, Ive have said so far (January 2026) Built In: OpenAI's New Device: What We Know So Far (February 2026) TechCrunch: OpenAI and Jony Ive may be struggling to figure out their AI device (October 2025) TechRadar: 2026 could be the year we move beyond smartphones (January 2026) Tom's Guide: OpenAI just killed Sora as company readies IPO and new 'Spud' model (March 2026) Geeky Gadgets: ChatGPT "Spud": What We Know About OpenAI's Next GPT AI Model Evolution (March 2026)