Why did I use the tomato clock again?

I recently realized one thing: I am not short of ideas, I am not short of start-up. What I lack is to keep attention on one matter long enough until it produces a full output.

Agent’s in a whirl, not sure when it’s done. When I saw a window running, I felt idle and wasted, so I started another project. The second Agent started running and I went to start the third thing. Soon, I had more than a dozen “ongoing” missions, which seemed to have been busy for days without knowing what had been accomplished.

It reminds me of the days when the tomato method was used completely. High-quality learning during the day and sleeping well at night; every day knowing what they have done, there is no need to remind themselves to move forward with anxiety. The efficiency felt was not delusional, but it wasn’t that famous 25 minutes.

I was learning C++ when I first took tomatoes seriously. In one day, three tomatoes were used to write a chapter code, three to read the next chapter, two to read the tomato work book and one to review the word. Looking back today, output is not necessarily surprising, but there was one thing that was very certain: I knew where time went and how far it had been. That certainty makes me fall asleep at 10 o’clock.

What I am missing is precisely that certainty. Instead of filling up a day, it ends with a few concrete results: I did it.

It’s more simple than being busy today, and it makes me less anxious.

Tomatoes are not 25 minutes of magic, but fixed borders, turning vague, infinite, anxious work into an achievable promise.

Why did the tomato clock work?

“Burn Wiki Today” would be an instinctive delay, because it has no end.” It’s easy to start by simply stating the core of the article.” I am no longer committed to completing the entire project, but to completing the immediate action. When you promise to be small, the brain doesn’t have to calculate the full cost, and it doesn’t have to be perfect.

The fixed border also pulls me back from the results. The anxiety always focuses on when the course is completed, when the project goes online and when the knowledge base is good enough; tomatoes only ask what exactly is done at this point in time, and what is to be successful. The results remain important, but I am no longer concerned.

It also separates implementation, rest and decision-making. It is carried out without re-programming, without pretending to continue working, and at the border to decide what to do next. I used to mix these three things, and the brain was in a semi-executive, semi-judgmental state, and it looked like it didn’t stop, but there was very little real time.

When attention is broken, an hour disappears like five minutes. After slowing down, I rediscovered that 15 minutes had been long: enough to write a judgmental text, to review Agent’s delivery, or to tear a vague task out of the criteria for success. There was no increase in time, except that it was not left in the switch.

So tomato really protects judgment. It lets me finish first and then decide on the next step, instead of doing it at the same time, while discovering B, thinking of C and the last three.

That’s what slow is fast means to me. Rather than deliberately reducing speed, slowness is a willingness to allow an action to take up the time it takes and no longer to pay the switch for every new possibility.

AI, what’s the new delay?

Traditional delays are “no start”. The AI era has seen a more subtle delay:Started all the time, but never stopped.

It is not a low-intensity ejection, but rather a flight with a productive look. I asked questions in one window, spread out in another window, sent a mission to the third Agent, and studied the fourth option. Every conversation has immediate feedback, and every new window gives me the feeling of being moving forward.

The ordinary “resistance distraction” can’t explain it. I didn’t leave work to entertain, and I’ve been doing things that seem serious; but the switch between the different goals has shattered the judgement that really needs me.

The hardest part is that this delay will reward me. A new question was answered immediately, and a new plan came up with a contours, and an Agent started right away. By contrast, reading long scripts, reviewing codes, checking failed paths, changing them to releaseable, is slower, boring, and easier to reveal what I really don’t know. So I’m instinctively going back to the conversation and avoiding the difficult part of the old mission with new possibilities.

Dialogue itself is not an output, nor is it assigned to AI. A real advance must leave something to examine: a document, a clear decision-making, a passing or failed test, a release, a reply, or a justified waiver. If I just talk a lot, get a lot of ideas, get a lot of assignments, I just expand on the product.

The real scarcity of AI is not the generation of capacity, but the ability to stop generating, judging and delivering.

Why is this a capacity expansion illusion and a closed circle debt?

AI makes me look like I can do anything: develop a website, make a product, write a course, study a strange business, start a business. I couldn’t even do the prototype in the past, and now I can get Agent to move on.

This capability is real, but it has expanded the scope for start-up and does not automatically expand my target judgement, field understanding, receiving and inspection capacity, risk taking and ongoing maintenance capacity.AI suddenly amplified my action radius, but it did not simultaneously amplify my ability to bear the results.

To have a real ability, through a ladder:

  1. **AI can generate:**I can get AI to give code, file, or plan.
  2. **I can run:**I know how to run it up.
  3. **I can explain:**I made it clear why it works like this.
  4. **I can accept:**I can judge whether the results have met their goals.
  5. **I can fix it:**After an error, I can locate, debug and recover.
  6. **I can maintain:**I am willing to bear the burden of renewal, risks, costs and consequences over the long term.

The first two levels have only been leveraged to generate and initiate. I began to have it when I entered into interpretation, acceptance, repair and maintenance. If I ignore this ladder, I go into a cycle: the sense of capacity expansion makes me open too many projects at the same time; each project is poorly understood, making it impossible to accept and accept; I go to AI to find new solutions, and then I create more branches, more difficult to close.

That explains why the more prototypes, the more anxious I sometimes get. Each prototype is like a competency voucher, but once I have to go online, face real users, deal with anomalies, and undertake follow-up maintenance, I find it still borrowed. The problem is not by using AI, but by mistaking it as “I can be responsible.”

Every unaccepted task continues to draw attention: I have to remember where it works, worry about its failure and decide when to go back. This is closed-ring debt. My default co-op ceiling is therefore limited to1 Main Delivery + 1 Tasks Pending External ResultsOther new ideas enter the parking lot. The solution is not to make less use of AI, but to make commitment catch up with capacity to generate.

How do you work for a closed-ring tomato?

Ordinary tomatoes are limited in time and are not sufficient to address multiple windows. If five windows, three targets and seven missions are allowed in a tomato, the timer only records how long the confusion lasted.

What I need is a tomato delivery, or a closed cycling tomato:**One target, one main window, one acceptance.**No new targets are opened until the old targets are accepted and accepted. Information and Agent can be multiple, but all actions must serve the same direction of delivery.

It has three phases, and the time can be adjusted to the mandate:

Define closed circle

Write the target, main window and success criteria before starting. “Advance course” is not an objective, but a list of gaps in the Windows 1 screenshot is completed and the next one is marked. If you don’t know the criteria for success, you’ll have to downsize the task and get Agent to work.

Push the ring.

Only for the current receipt and inspection. When Agent is not finished, I can check needs, organize receipt and inspection, read relevant materials, prepare for testing; if there is really nothing to do with the same goal, then rest or end early. Waiting for AI is not working, let alone waiting to start another project.

Receiving and Inspection Closed Rings

At the end of the examination, the receiving and inspection material is located, the standard is met, the next step is what, and only four states are available:

  1. **Delivery:**Acceptances exist and meet the criteria for success.
  2. **To be external:**Registered responsible, current status and conditions for the next inspection.
  3. **Clear blockage:**Clear evidence of obstruction, requirements and the next person in charge have been written.
  4. **Active abandonment of:**Continued input is not worth the judgement and the reasons are documented.

“There’s a lot of talk,” “There’s a lot of new ideas,” “the AI has been assigned” is not a state. The number of tomatoes is also not an achievement; whether a tomato is effective depends only on whether it allows the control ring to move forward.

Which missions are eligible for tomatoes?

Not every thought should immediately become a development task. Before I begin, I would like to ask five questions: what is the goal, what are the criteria for success, what are the main risks, when should we stop, and how will we recover after failure?

It is unclear whether to learn first, to shrink, or to explore borders only once. For example, “to confirm whether this route is worth continuing with a tomato”, rather than letting Agent do it first. Irritated enough to understand the material, continue to walk away, make a clear difference in judgment, and are not entitled to continue to open tomatoes; the right action is rest, not a brush.

Exploration and formal commitment must be separated. Exploration can allow the answer to be “no worth doing” and not to require the output of a full product; formal commitment means that I am willing to be held accountable for acceptance and subsequent results. I used to upgrade my curiosity directly into a project, and now I’m going to make sure that it is entitled to take over the attention of the next few days, weeks or even more.

There is no need to design an ambitious time management system today, but to do three closed loops: to take back a runaway project, leaving behind physical results that can continue to be used; to wait for an AI without opening a second project; to check today ’ s documents, submissions, articles or replies, and to write in three words what has been done, what has not been done, what will be done tomorrow.

The common denominator of these three exercises is not 5/15/5. They all have an end to be seen and accepted.

How do I get my job back on the line?

I have recently taken what I have to do back to the borders that I can touch, understand and bear. Wiki and AI’s little white introductory class is my area of competence, but even these things still need real time.

This is not a retreat, but rather a question of whether “AI can start” and replace it with “can I continue to judge, grind, deliver and maintain”? AI can compress implementation time without eliminating the time required for taste formation, content judgement, user feedback, rectification and long-term maintenance.

Wiki doesn’t produce articles that end. I’m going to decide what is worth making public, where it should be, if the reader can find it, if the old link is bad, if it’s really readable after the release. Nor does it end with scripts, which require real operations, screenshots, trial readings, feedback and repeated revisions. These times cannot be bypassed by a hint.

Slow down and choose only a few of the works that really belong to me. I’m restoring ownership.

Final judgement: less, less, more, more.

I once understood efficiency as moving forward, and the parallel power of AI as I should. As a result, projects are growing, attention is becoming increasingly fragmented and time is becoming scarce.

What I really need now is not to fill every minute, but to leave a period of time for one thing and leave the full result at the end. Tomatoes are the smallest unit to carry the ring; the core is not a timer, but a limit on commitment and distribution.

AI is responsible for expanding implementation and I am responsible for securing the borders. Don’t do it, finish it until you really have it. Slow is fast.