NOTES · 2026-04-28

When you spend 4 hours in ChatGPT and have one paragraph to show for it

I opened ChatGPT to draft a client proposal last Tuesday. Four hours later, I had one polished paragraph about sustainable agriculture and seventeen open tabs on regenerative farming practices. The proposal remained unwritten. We know this moment—when the tool that was supposed to save time becomes the time sink itself. I built Deskrune's systems after too many afternoons lost to AI rabbit holes, chasing the perfect prompt instead of the completed task. This is AFM-1 Hyperfocus Tunnel territory, where curiosity overrides intention.

The AI tool that was supposed to save time became the time sink itself. AFM-1 Hyperfocus Tunnel protocol for when you come back.

The AFM-1 Hyperfocus Tunnel in AI work

AFM-1 Hyperfocus Tunnel isn't just about deep work—it's about work that deepens in the wrong direction. You start with a clear objective: 'Write project summary.' ChatGPT suggests an outline. You ask for examples. Then you're comparing Claude's narrative style against Gemini's bullet-point approach. The original task fades as you optimize an irrelevant detail.

We call this the specificity trap. The more precise your follow-up questions become, the further you drift from completion. I've watched fellow ADHD adults spend hours refining AI-generated content that never needed refinement, caught in the gap between 'good enough' and 'perfect.' The tunnel narrows until all you see is the texture of the sentence, not the shape of the document.

Why ChatGPT amplifies our time-blindness

ChatGPT's conversational interface mimics the way our brains jump between connections. Each response suggests new avenues, and we follow them because the dopamine hit of discovery outweighs the satisfaction of finishing. I've tracked my own sessions: what begins as 'research market trends' becomes 'analyze 19th-century economic patterns' within six exchanges.

The absence of natural stopping points compounds the problem. Unlike a book chapter or meeting agenda, AI conversations have no built-in conclusion. We continue prompting because we can, not because we should. This is particularly dangerous for those of us with AFM-7 Time Agnosia, where four hours feels like twenty minutes in the flow state.

The 12-minute timer protocol

I developed the 12-minute timer after realizing that hyperfocus needs boundaries, not elimination. The protocol is simple: set a visible countdown before any AI session. When the timer expires, you must stop interacting and assess progress against your original goal. Not a suggested pause—a hard stop.

Twelve minutes is the sweet spot between deep engagement and time loss. It's long enough to accomplish meaningful work but short enough to prevent tunnel vision. We built a free web-based 12-Minute Timer that doesn't require login or track data—just a straightforward tool for the community.

How to structure your AI sessions

Before opening any AI tool, write your objective on paper. Not in a digital note—physical ink creates commitment. 'Complete budget section' is valid; 'explore financial concepts' is not. This pre-commitment forces specificity before the conversation begins.

During your 12-minute session, track prompts versus output. If you've asked more than three follow-up questions without adding to your document, you're in the tunnel. The solution isn't better prompts—it's closing the tab. I keep a sticky note with 'IS THIS MOVING THE NEEDLE?' above my monitor as a visual checkpoint.

When to switch AI tools (and when not to)

Tool-hopping is another form of avoidance. I've watched people restart the same request in Claude, ChatGPT, and Gemini seeking some mythical perfect response. The truth is that most modern AI tools produce comparable quality for straightforward tasks.

The exception is specialized work. Notion AI excels at reorganizing existing content. Custom GPTs trained on your writing style can maintain voice consistency. But for drafting and brainstorming, pick one tool and stick with it for the entire project. We maintain a living comparison of real AI tools updated monthly, focusing on practical ADHD workflows rather than feature lists.

The paragraph completion metric

I now measure AI sessions by paragraphs completed, not time spent. If I haven't added at least one substantive paragraph to my document within 12 minutes, the session has failed. This output-focused metric prevents me from mistaking activity for achievement.

The metric works because it's binary. Either the paragraph exists or it doesn't. There's no gray area for 'almost finished' or 'needs refinement.' We apply this same principle to our Deskrune AI OS—every module has a clear completion state so you always know where you stand.

What to do with accumulated research

Those four hours of agricultural research weren't wasted—they were misallocated. I now have a 'Curiosity Bank' where I paste interesting tangents for future exploration. The key rule: banking happens after the primary task is complete.

This system acknowledges our brain's desire for exploration without derailing deadlines. I've linked unexpected research to later projects, like connecting medieval crop rotation patterns to modern productivity cycles in our 92-Day Bank App Mistake analysis. The ideas matter—their timing matters more.

Key takeaways

Set a 12-minute hard stop for all AI interactions using our free timer tool
Define your output metric before starting—paragraphs completed beats time spent
Physical pre-commitment to objectives reduces speculative prompting
Tool-hopping rarely improves quality enough to justify time cost
Bank interesting tangents for later instead of following them immediately