{
  "source": {
    "revision_source_version": "1.0.0",
    "canonical_url": "https://theproductiveprompter.com/docs/todoist-nirvana-is-not-the-outcome.html",
    "slug": "todoist-nirvana-is-not-the-outcome",
    "title": "What Todoist Nirvana Taught Me About AI",
    "original_published_timestamp": "2026-11-11T12:00:00-05:00",
    "source_html_sha256": "19e9d5dce688ed28d02e18bf8017730ebe847aaae53fff4d0407aad0916260b4",
    "source_markdown_sha256": "280702897f6e42a62bcfdb7272ad2bc57581441009c8dc17417a10bdd7945f06",
    "source_commit": null,
    "source_observed_at": "2026-11-13T15:43:20Z"
  },
  "article_markdown": "My Todoist Karma screen says Enlightened. That is the milestone behind “Todoist nirvana”: a level [reached at 50,000 Karma points](https://www.todoist.com/help/todoist/features/introduction-to-karma-OgWkWy). The screenshot below shows 51,651 points and 25,236 completed tasks since September 19, 2020.\n\nI got there while completing tasks, building dashboards, and working out how to keep several projects moving at once. I learned useful habits along the way. I also spent time getting better at maintaining systems that AI now helps me maintain.\n\nLooking back, the lesson I want to carry forward is to define the outcome first, then build a system around achieving it. Giving every new idea its own process can turn organizing into another project. I want to understand what deserves that investment before I start building.\n\n![Todoist Karma showing Enlightened, 51,651 points, and 25,236 completed tasks since September 19, 2020.](/assets/articles/todoist-nirvana-is-not-the-outcome/todoist-nirvana-20261111.png)\n\nMy Todoist Karma screen: Enlightened, with 25,236 completed tasks. Reaching that milestone gave me a reason to reflect on the habits I had built and the work they helped me do.\n\n## What I learned while organizing everything\n\nBuilding those systems taught me to capture commitments and keep several aims, projects, and goals in view. My capacity to focus on multiple outcomes feels greater than it was before. I can describe the habits I gained, and Todoist records the tasks I completed. Evaluating my productivity means looking at what that work accomplished.\n\nI valued those habits enough to feel frustrated when other people seemed less productive. That was a personal reaction, shaped by my own experience. Assessing their work would require looking at their goals and results, with room for a different way of working.\n\nI also have to look at how I spent my own time. I could move tasks, revise labels, and redesign a dashboard while leaving a difficult decision untouched. Preparing to do something had become an activity I could keep improving.\n\nThere are good reasons to build a system. A schedule can protect time, a dashboard can expose a neglected commitment, and a metric can reveal whether a process is working. Each also gives me something else to maintain. A cleaner view can make a project feel more manageable while the uncertainty inside it remains.\n\nThe productive people I notice seem to spend more of their attention on actions that advance their aims. That observation gives me a useful way to look at my own habits: name the outcome, identify the next action, and work out what support that action needs.\n\nResearch on intention and behavior helps explain why organizing can feel productive while the work remains unfinished. Strengthening an intention does not produce an equal change in what people actually do ([Psychological Bulletin](https://pubmed.ncbi.nlm.nih.gov/16536643/)).\n\nPlanning can help when it connects a real situation to a specific response. Implementation intentions are often expressed as “if this happens, then I will do that,” and a recent meta-analysis found an average benefit across many tests. The benefit varied with plan design, and the researchers detected publication bias, leaving the success of an individual detailed plan uncertain ([European Review of Social Psychology](https://doi.org/10.1080/10463283.2024.2334563)).\n\nFor me, the useful question is which action a proposed improvement will help me take, and when. If I keep struggling to name that action, I probably need to stop designing and act.\n\n## What AI has changed in my day\n\nAI has changed the work of maintaining these systems dramatically. I now manage my entire workflow through centralized processes. Much of the organization that once involved building my own dashboards, tools, and mechanisms happens through those processes.\n\nThat includes the recurring work of reorganizing information, restoring its context, and archiving it, along with the weekly routine of reviewing emails, dashboards, and metrics. For the most part, AI has helped me take that system maintenance out of my day-to-day tasks. The processes support more complex workflows and bring forward the information that needs my attention.\n\nThis is what continuous action means to me in practice. The work can keep moving, with its information organized and its context carried along. I can attend to what needs a decision or a check as it comes up. Much of the effort I once put into arranging the work now supports the work as it proceeds.\n\nThe processes themselves can still become a distraction. There is plenty of opportunity to keep refining them, just as there was with a dashboard. I want an improvement to begin with an outcome that needs it. That gives me a reason to change the system and a result against which to judge the change.\n\nMuch of the organizing I learned to do is now work that AI helps handle. I still use the understanding behind it: which information matters, what context a task needs, and where a workflow needs attention. I have far less reason to practice every part of that skill manually.\n\n## Where I still need to pay attention\n\nAs more work passes through these processes, I still need ways to tell whether it is going well. AI can help carry out parts of a computer workflow, with its capabilities depending on the task and environment. A clear, worthwhile goal and a suitable setup can improve the conditions for execution. An agent can still lack a permission, misunderstand a constraint, miss a change, or produce something polished that is wrong.\n\nThe improvement in my own workflow gives me a reason to keep adapting. The effect on productivity still depends on the work involved. In one randomized study of experienced open-source developers using early-2025 tools, participants took longer with AI. That finding belongs to the people, tools, tasks, and environment studied. Other workflows need to be evaluated in their own settings ([METR](https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study-paper.pdf)).\n\nIn the 2026 OSWorld 2.0 benchmark, computer-use agents attempted 108 long-horizon workflows. The best evaluated configuration completed only a minority under the primary all-or-nothing measure. Failures included losing track of constraints, missing information introduced during a task, guessing instead of asking for clarification, skipping verification, and failing to recover hidden state ([OSWorld 2.0](https://arxiv.org/abs/2606.29537)).\n\nA general estimate of how much computer work AI can perform would need evidence beyond those tasks and configurations. A model’s task-completion “time horizon” also describes performance on selected, unusually well-specified tasks. The task, model, context, and agent setup all affect the measurement. Reliability during autonomous work elsewhere needs to be assessed in that environment ([METR](https://metr.org/time-horizons/)).\n\nWhen I hand work over, I still need to choose an outcome worth producing and supply the context, constraints, access, and definition of success. Checkpoints should reveal drift before correction becomes expensive. At the end, I need to compare the result with evidence and judge its quality, consequences, and tradeoffs. If the work fails, stalls, or changes the environment unexpectedly, I need a way to recover.\n\nThe evidence might be a test, a source record, a reconciled total, an expected state, or a qualified person’s review. This is what ground truth means in practice. The task and its risk determine the appropriate check. NIST’s guidance similarly recommends evaluating accuracy, quality, reliability, and authenticity against known ground truth using suitable human and automated methods ([NIST](https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf)).\n\nMonitoring helps me catch trouble during execution. Verification tests the properties that matter in the result, including when it looks plausible. Judgment determines whether that result is useful and acceptable, and recovery gives me a way forward when something fails.\n\n## A candle maker learning about electricity\n\nI think about this as a candle maker encountering electricity. The candle maker understands light, where people need it, and where a candle will be useful. Producing light with a bulb calls for a different set of skills. Some of the old understanding carries over, and some of the effort spent perfecting the old method could have gone toward learning the new one.\n\nThat is how I feel about some of my system-building. I learned a lot, and some of that time was wasted. I can value what I learned while reconsidering how much attention the old routines deserve.\n\nFor the next idea, I want to choose the outcome first. The habits I learned can help me shape the work, with AI handling much of the routine organizing. I can improve the system when the outcome calls for it.\n"
}