--- title: "The obvious improvement can still be the wrong ticket" subtitle: "Staff only the work that names a new fact about who buys" principle: "Teach or park" date: "2026-09-29T14:07:50.783Z" author: "Mikko Atienza" canonical: "https://ciel.codes/showcase/blog/the-obvious-improvement-can-still-be-the-wrong-ticket" edition: "2026-09-29" source: "https://ciel.codes/showcase/pulse/2026-09-29" abstract: "Most work that looks like a clear product improvement will teach you nothing about who pays you. Park it. The week belongs to the few tickets that force a new sentence about a real customer." cite: "Atienza, M. (2026). Teach or park. Pulse." citations: - handle: "@gabriel1" url: "https://x.com/gabriel1/status/2104733436196057136" --- Most work that looks like a clear product improvement will teach you nothing about who pays you. Park it. The week belongs to the few tickets that force a new sentence about a real customer. You do not need a new system for this. You need a tag, a named closer, and a Friday sentence. ## Your week is already full of obvious work Look at what is slated. A cleaner handoff. A tighter confirmation. A status field someone on the team has wanted for months. These are easy to defend because they sound like quality. They are also the work that never surprises you. You already know how they end. That is the trap. The shop stays busy. The picture of the customer stays frozen. Gabriel Petersson (@gabriel1), who left OpenAI to ship Energy and is building that product in public, put the filter in plain language. He has a thousand ideas for how he wants to improve Energy. Maybe three give him information about who the customer is. The rest he wants ignored, even when they look like obvious upgrades. > ignore even things that seem like clear product improvements if they don't give you new information about who your customer is > > Gabriel Petersson (@gabriel1) He is not telling you to stop caring about quality. He is telling you that polish without a new fact is almost never the scarce work. On a team this size, scarce work is the conversation that changes what you believe about a buyer. The nicer version of a flow you already understand will not do that. If you staff the obvious pile first, the learning tickets slip to next week, then the week after. Comfort always has a defender in the room. Learning has a name, a calendar, and a chance the customer says no. That is why it loses. ## Two tags before anyone starts Monday, before anyone picks up a ticket, sit with whoever owns the board. That might be you. It might be the ops lead. It might be the person who turns chat complaints into cards. Every open item gets one tag. Not a score. Not a maybe. Customer learning means if we do this, we can write a new fact about who pays, what they refused, or how they actually use us. Comfort means it would look better, match a competitor, or make the team feel less embarrassed. You will not know a new person when it is done. - [ ] Board owner screenshares every ticket slated for this week - [ ] You say the tag out loud for each one, learning or comfort, with no third bucket - [ ] Comfort stays visible and nobody is assigned to it If a ticket claims to be both, it is comfort until you can say the fact you expect to learn. "We'll see how they like it" is not a fact. It is hope. Do this even if the week already feels late. Especially then. Late weeks are when comfort work sneaks in because it looks like progress. Tell the person who wanted the polish that the ticket is parked, not killed. It comes back when it can teach. That keeps the room from treating the filter as an insult to craft. Run one live learning ticket, not five. One named conversation that produces a sentence beats a scatter of half touches that nobody writes down. ## The owner still names the customer The tag does not give anyone a license to experiment on live accounts. You pick which learning ticket may touch a real customer this week, and you write that person's name on the card before work starts. "Customers" is not a name. Someone who paid you last quarter is a name. Friday you write what you now know, or you write that you learned nothing. The closer, sales, support, or whoever actually talks to accounts, runs the conversation. They do not get to move the bar after the call. ```mermaid flowchart TD A["Board owner tags every ticket"] --> B{"Owner: does this teach who pays?"} B -->|No| C["Board owner parks the ticket"] B -->|Yes| D["Owner names the live customer"] D --> E["Closer runs the conversation"] E --> F["Owner writes the Friday fact"] C --> G["Owner re-reads parked tickets Friday"] style A fill:#1e1b4b,color:#fff,stroke:#1e1b4b style B fill:#4c1d95,color:#fff,stroke:#4c1d95 style C fill:#064e3b,color:#fff,stroke:#064e3b style D fill:#1e1b4b,color:#fff,stroke:#1e1b4b style E fill:#064e3b,color:#fff,stroke:#064e3b style F fill:#4c1d95,color:#fff,stroke:#4c1d95 style G fill:#1e1b4b,color:#fff,stroke:#1e1b4b ``` Keep the parked pile. Do not delete it. You need to see, on Friday, how much obvious work you refused. That refusal is the strategy. If the parked list is empty, you did not run the filter. You ran a mood. What must stay with you: which experiment touches a living account, whose name goes on the card, and whether Friday's sentence counts. The closer can run the call. They cannot declare the week a success because the ticket moved to done. ## Friday measure Nobody has a credible figure for how much of a shop's week is comfort dressed as quality. You will have one after five days if you count in the open. | Measure | Count | |---|---| | Tickets tagged comfort that still received hours | | | New named facts about who the customer is | | > **Aside.** Count a fact only if someone can read it aloud without opening the ticket. "We shipped the new form" is not a fact about who pays. "Two of the four accounts we called stay because the owner is off-site on Tuesdays" is. You want the first row near zero. You want the second row to hold at least one sentence. If you closed a stack of tickets and wrote zero facts, the week was polish. Say that out loud. Do not dress it up as a shipping win. The fact can be small. A reason they almost left. A job they hired you for that is not the job on your homepage. A person who was never going to buy. Small is useful. Vague is not. ## Kill the comfort assignments this week - [ ] Unassign every ticket tagged comfort so hours cannot hide there - [ ] Write one live customer's name on the learning ticket you will actually run - [ ] Put Friday's fact where the team can see it, even if the sentence is that you learned nothing The board you already keep is enough.