AI coders half-open laptops through airports, offices, and ice rinks are now a recognizable sight in 2026 — developers walking around with their screens cracked just enough to keep a background AI agent alive. It looks absurd. It is also a completely rational response to a real technical constraint that the AI tooling industry has not yet solved. Here is the full story of why this behavior is happening, what it reveals about the current limits of AI agent infrastructure, and what the industry needs to build before developers can finally close their lids.
The Behavior: What People Are Actually Doing
The Business Insider report that set off this week’s viral conversation documented something that any developer in a major tech hub has probably already noticed: colleagues walking through public spaces with their laptops deliberately left ajar, screen glowing just enough to confirm the machine has not gone to sleep.
Business Insider reports that some developers are walking through airports, offices, and ice rinks with their laptops left slightly open so background AI agents can keep running. The article profiles users including Geoff Chan, identified as a 39-year-old head of product at Raven.AI, who says he leaves his laptop ajar during his daughters’ skating practices to avoid interrupting OpenAI Codex-assisted work, telling Business Insider, “I have to put it up on a shelf.” Engadget
The ice rink detail is what made this story travel. There is something about the image of a parent at their kid’s skating practice, laptop propped on a rink-side shelf at a specific angle to prevent sleep mode, that crystallizes exactly how far AI agent workflows have burrowed into daily life in 2026. This is not an edge case of a developer working from a coffee shop. This is a parent at an ice rink doing his best to keep an AI coding session alive because closing the lid would cost him hours of work.
The Technical Root Cause: Why Closing the Lid Kills the Agent
To understand why developers are doing this, you need to understand what happens when a laptop goes to sleep while a long-running AI agent is active.
Business Insider reports that the practical reason for this behavior is that many agent workflows run locally or rely on unstable WiFi, so closing or sleeping a laptop interrupts long-running processes and developer progress. The article references OpenAI Codex as an example of the kind of tooling integrated into these sessions. Engadget
When a laptop sleeps, its operating system suspends active processes to conserve power. For most applications, this is seamless — the app resumes where it left off when the lid opens. But AI agents are different. A long-running agent session may be in the middle of executing a multi-step task: analyzing a codebase, writing functions, running tests, iterating on failures. When the process is suspended mid-execution, the agent loses its working state. Depending on the tool and the task, resuming may mean starting over from a checkpoint or losing the entire context of what the agent was doing.
For agents that depend on a live internet connection — including cloud-based tools like OpenAI Codex that require continuous API calls — the problem is compounded. Sleep mode drops the network connection. When the laptop wakes, the session may have timed out, the API context may be lost, and the agent’s progress may be unrecoverable.
OpenAI Codex and the Tools at the Center of This Behavior
OpenAI Codex is the AI coding tool most directly referenced in the Business Insider reporting, but it is not the only product driving this behavior. The broader category of agentic AI coding tools — which includes tools like Claude Code, Cursor, Devin, and various custom agent frameworks — all share the same fundamental architectural limitation.
Developers who rely on long-running local agents create operational pressure on device uptime, battery management, and network stability. In comparable developer workflows, practitioners often adopt containerized runtimes, remote session keep-alive utilities, or dedicated edge machines to preserve state without carrying a laptop in public. Engadget
The containerized runtime approach — running agents inside Docker containers on a remote server rather than locally — is the most robust technical workaround currently available. But it requires infrastructure setup, cost management for cloud compute, and a level of DevOps knowledge that not every developer using AI coding tools possesses. The appeal of local agent tools is exactly that they do not require that overhead. When the workaround to a local tool’s limitations is “set up a cloud server,” many users will simply carry the laptop instead.
The Workplace and Airport Dimension
The social dimension of the AI coders half-open laptops behavior is as interesting as the technical one. These developers are not just carrying open laptops at home. They are doing it in workplaces, airports, school hallways, and public venues — spaces where an open laptop screen attracts attention and creates both practical and security concerns.
Business Insider also cites other users who describe walking through airports, offices, and school hallways while keeping machines active to avoid losing agent state. Engadget
An open laptop in an airport with an active AI coding session is a visual security consideration. The screen may display proprietary code, API keys, internal tool configurations, or sensitive project context that is visible to anyone who glances in the right direction. Workplace information security policies at many companies explicitly address unattended or publicly visible screens. The fact that developers are weighing those risks against the cost of losing an agent session — and deciding to keep the lid open — tells you something about how much value they are placing on uninterrupted AI agent time.
What Geoff Chan’s Ice Rink Moment Actually Represents
Geoff Chan shelving his laptop at an ice rink while his daughters skate is a microcosm of a much larger shift in how AI tools are integrating into the rhythms of work and life in 2026.
A year ago, an AI coding tool was something you opened at your desk, used for a specific task, and then closed. The session was short enough that laptop sleep was not a meaningful concern. Today, AI agents are running multi-hour tasks autonomously — refactoring large codebases, writing and testing entire feature sets, iterating through failure modes without human intervention. The session length has extended dramatically. And the infrastructure supporting those sessions has not kept pace.
The anecdotal trend is notable because it illustrates how user behavior adapts around the limits of current toolchains rather than the other way around. For tool vendors and platform teams, recurring reports of preserved foreground sessions point to demand for durable local state, checkpointing, and more robust background execution models. Engadget
That last point is the most important one for the AI development industry. When users adapt their physical behavior to work around software limitations — when they prop laptops on ice rink shelves rather than use a feature that does not exist yet — that is a product signal. It is not subtle. It is a visible, public, viral demonstration of a gap in the tooling that the market is waiting for someone to fill.
What the Industry Needs to Build
The AI coders half-open laptops behavior will go away when the underlying tooling problem is solved. The solution set is already understood, even if the products do not yet exist at scale.
Observers should watch for product responses such as improved checkpoint APIs, background agent services that survive sleep cycles, and smarter mobile-to-cloud handoffs. Also watch whether these user stories motivate security or workplace-policy discussions about unattended, visible screens in public settings. Engadget
Checkpoint APIs that save granular agent state — not just the output but the agent’s internal working context — would allow a developer to close their laptop at the ice rink and reopen it at home with the session continuing exactly where it stopped. Background agent services that survive sleep cycles would eliminate the need to keep the screen active at all. Cloud handoff capabilities that migrate a local agent session to a cloud runtime when the laptop is about to sleep would preserve both portability and session continuity.
All of these are engineering problems, not research problems. The technology to build them exists. What has not yet happened is the product prioritization to build them into mainstream AI coding tools. The Business Insider viral moment may be the catalyst that changes that.
Broader Implications: The Gap Between AI Ambition and Infrastructure Reality
The AI coders half-open laptops story is ultimately a story about the gap between what AI tools promise and what the underlying infrastructure currently delivers. Companies are releasing agentic AI tools that run long, complex, autonomous tasks. They are marketing those tools around the vision of a developer who can delegate entire workstreams to an AI and walk away. But the walk-away part only works if the AI does not stop when you close your laptop.
The user behavior documented in Business Insider’s reporting is the market’s response to that gap. Developers want the vision. They are compensating for the infrastructure with physical workarounds. The question for every AI coding tool company in 2026 is how long they are willing to let ice rink shelf sessions be the state of the art in agent persistence. For more on the biggest stories in AI and developer tooling, visit The Tech Marketer.
Latest Updates
The AI coders half-open laptops story is spreading fast across the developer community. Here is where to follow the full conversation:
- Business Insider has the original reporting on developers carrying half-open laptops through airports, offices, and ice rinks to keep AI agents running, including first-person accounts from developers using tools like OpenAI Codex. Read more at Business Insider
- AsatuNews has coverage of how AI developers are keeping their laptops open in public spaces to sustain AI agent sessions and what the behavior signals about the current limits of AI developer tooling. Read more at AsatuNews
- Let’s Data Science has the full editorial breakdown of the Business Insider report including technical analysis of why sleep mode interrupts agent sessions, what workarounds currently exist, and what the industry needs to build next. Read more at Let’s Data Science
FAQ: AI Coders Half-Open Laptops Agents
1. Why are AI coders keeping their laptops half-open in public? Developers using AI agent tools like OpenAI Codex are keeping their laptops ajar in airports, offices, and other public spaces to prevent the machine from sleeping. When a laptop sleeps, it suspends active processes and drops network connections, which interrupts long-running AI agent sessions that can take hours to complete.
2. What is OpenAI Codex and why does it require an open laptop? OpenAI Codex is an AI coding tool that can autonomously execute complex programming tasks over extended periods. Because it requires an active internet connection and an uninterrupted local process, closing or sleeping the laptop mid-session can cause the agent to lose its working state and context, forcing developers to restart.
3. Who is Geoff Chan and what did he do at the ice rink? Geoff Chan is a 39-year-old head of product at Raven.AI who was profiled by Business Insider. He described leaving his laptop propped open on a shelf at his daughters’ skating practices to avoid interrupting an ongoing OpenAI Codex session, illustrating how AI agent workflows are now extending into everyday life.
4. What technical solutions exist for the AI agent laptop sleep problem? Current workarounds include running agents in containerized cloud environments that do not depend on a local machine, using remote session keep-alive utilities, or dedicating a separate edge machine to agent tasks. The products that would fully solve the problem — including checkpoint APIs and background services that survive sleep cycles — do not yet exist at mainstream scale.
5. What should AI tool companies build to fix the half-open laptop problem? Experts point to three key product needs: checkpoint APIs that save the full agent state so sessions can resume after sleep, background execution services that keep agents running without requiring an active screen, and mobile-to-cloud handoff capabilities that migrate local sessions to cloud runtimes automatically.
