This piece explains why traditional peak planning no longer works in a world of continuous, agent-driven demand and what replaces it. If your systems are built for seasonal peaks and reactive response, you will experience modern failures as sudden, unexplained events. What follows is a field report detailing those events and how teams survive them.
Every engineering leader will encounter this, your dashboards are green, but your business is bleeding. An upstream recommendation engine re-ranked a category overnight and you find out from finance, not from product. An autonomous pricing agent quietly walks your margins off a cliff, no human approved. An agent-driven procurement chain you didn’t know about is suddenly buying from you twice as much, until it isn’t. You open the war room. You call the team. The team does what twenty years of incident response trained them to do, and the room is full of capable people doing exactly the right things for a kind of incident this is not. Somewhere around minute forty you step out into the hallway for water and feel a specific kind of alone. Not the loneliness of a hard problem. The loneliness of a problem your instruments cannot explain and your peers cannot describe. The loneliness of standing at what you thought was the finish line and discovering that the line is not there. This is a dispatch from that hallway.
The Clearing
24:00 hours. It’s Dark. The rain has been steady for six.
Sarah, an ultra-marathoner, breaks out of the trees into the clearing where the finish line should be. There is nothing there. No banner, no timekeeper, no truck pulled up on the access road, no other runners arriving behind her. Just the trail picking up on the far side of the clearing and continuing into the woods.
Another runner emerges from the trees thirty seconds later. She does not stop. She glances at the clearing once, briefly, the way a hiker checks a known landmark and continues across to the far trail. Her stride does not change. Her pack rides her shoulders the way a pack settles after years of use. There is no hesitation in her crossing. She is not looking for what was supposed to be there. She is heading toward what is.
Sarah watches her cross.
Last November, this community published a series called Peak Season, an ultramarathon with no finish line. At the time, that framing was a metaphor for structural risk, the preparation, response, and learning loops that never close. This dispatch is what the metaphor meant. You will recognize Sarah, our runner, because she is close enough to you that you have probably already been her this morning.
Ninety seconds. That is the length of pause an experienced ultrarunner takes at an aid station that turns out not to be an aid station. You inventory your pack, you hydrate, and you decide what you can no longer afford to carry. Then you cross the clearing.
This piece is about those ninety seconds.
What Sarah does not yet know is that she has a choice about which kind of runner she will be from now on. The runner she just watched cross the clearing, the one who did not stop, whose pack moved with her as though it belonged on her, she is not running a longer version of the race Sarah is running. She is running something else, and she has been for a while.
The Race You Trained For Is Not the Event You Are Inside Of
Before we get to the trail picking up on the far side of the clearing, we owe each other an honest accounting of what just happened.
The community has been through three transitions in the last thirty years that each looked like a category change at the time and, in retrospect, turned out to be additions to the same race.
E-commerce in the 1990s added a new digital channel. It shifted where peaks happened, no longer just physical. The calendar still ran. The drivers were still human. The finish line was still in the same place, now with a new digital channel.
Omnichannel maturity in the 2010s added integration burden across channels. It changed how peaks expressed themselves operationally. The calendar still ran. The drivers were still human. The finish line was still in the same place.
Mobile dominance, which crossed over last year, added a substrate. It changed which devices and contexts drove demand. The calendar still ran. The drivers were still human. The finish line was still in the same place, now with more ways to arrive at the finish.
Three additions to the same race made it harder, longer, and more crowded, yet still recognizable. The runners who took each transition seriously did better than the runners who didn't. The training compounded.
What’s happening now is structurally different, and saying so plainly is the only honest thing to do. Agent-driven demand removes the human driver from peak seasons. The calendar comes apart, and the finish line goes with it. This is not the next chapter of the race the community has been running. It is the early hours of a different kind of event entirely.
The community trained for an ultramarathon. That training is not wasted. It is what earned the right to recognize what comes next, and it’s what makes survival here possible at all. A runner who had not put in the years would not last 24 hours in these conditions. However, the conditions are no longer those of a race we recognize, and the runners who insist on running them as a race are running themselves into the ground.
The race you trained for got you here. What you are inside of now is something else.
Three Peaks Now, Not One
Last year's series taught the community what a human peak looks like at scale. Black Friday, tax season, open enrollment, the school calendar, and holiday cycles are all demand patterns tracking to human capital markers; it’s the paychecks, the holidays, the deadlines you could put on a wall calendar in 1995 and still recognize in 2025. That work still applies. The instruments built for it still work for the peak it described.
What has changed is that two more peaks have moved onto the same trail.
The machine peak has no calendar. It forms from continuous activity, an underlayer beneath your business: millions of agent decisions made in milliseconds, evaluating offers, prices, inventory, and trust signals, converging when their rulesets agree on the same value at the same instant, then taking action. There is no advanced human-readable warning. The peak is a standing wave that emerges from the agents' collective behavior, not a pulse driven by a date. By the time it is measurable in your dashboards, it has happened. The encounter is over before the response. We will spend most of this dispatch on this one because it is the peak your team is most likely to encounter this year, and it is the peak your existing kit was not built to handle.
The hybrid peak occurs when humans rewrite the rules their agents run on. A central bank moves a rate. A regulator changes a disclosure requirement. An influencer reframes a category. A million agent rulesets re-weight overnight. Demand re-pools at a new value point that did not exist twelve hours before. This peak is the most volatile of the three because it lacks both the predictability of the human peak and the substrate-driven physics of the machine peak; it is reflexive, and the rules change faster than the rule-changers anticipate. We do not yet know enough to draw this map; we will return to it in a future dispatch, once there is enough trail behind us to report honestly.
Three peaks. The runner who prepared for one of them and called the work done is the runner now standing in the clearing in the rain.
There is a second shift beneath all three peaks that most teams have not yet instrumented: the ability to prove what the system believed at the moment it made a decision. Not after the fact. At the moment of commitment. This will matter more than any single metric you are currently watching.
The Bear at 9 a.m. Tuesday
Sarah's team had their best peak season ever in 2025. They had instrumented the wallet boundary, drilled the inference fallback, kept their idempotency keys current, and rotated their on-call into the shape the original series recommended. They came back from peak with a clean retrospective and a confident roadmap. They had earned the right to be in this race.
It is now Tuesday, March 17, 2026. Mid-morning. Sunny. A normal day. Sarah is in a one-on-one with her staff engineer at 9:11 a.m. when the first signal arrives, not from a pager, but from a finance channel asking why authorization volume is down 14% in the past hour while traffic is flat. Every operational dashboard on her wall is green.
Here is what happened: the team will not finish the post-incident analysis for another six days, and will reconstruct it from the post-incident analysis.
At 8:47 a.m., a major recommendation provider released a routine model update. The release notes did not flag the change as material. Within seven minutes, comparison agents across three retail intermediaries had started returning re-ranked results for Sarah's product category. By 8:54 a.m., one of Sarah's competitors, previously ranked fourth in her market, was being recommended ahead of Sarah's brand on roughly 38% of agent-mediated comparison queries. Two minutes later, the agents who still recommended Sarah's brand were sending shoppers to a checkout flow that her wallet partner had quietly throttled after detecting anomalous request patterns from the agents themselves. By the time the finance channel pinged at 9:11 a.m., the convergence had been running for 24 minutes, and the day's revenue impact was already in the books.
This is what every runner who encounters a black bear on the trail in fading light knows about wild-animal encounters: the encounter is over before it begins. The animal saw you before you saw it. By the time you register what’s happening, the structure of the moment is already determined. You do not survive these encounters by reacting faster. You survive them because of who you had become as a runner before this morning, your pack, your pacing, your conditioning, the trail you had already walked at every other hour.
The bear in Sarah's clearing is not bad luck. It is her firm's own architectural debt manifesting under conditions it was never designed for. The team that was prepared for human peaks did not know they were responsible for understanding agent peaks until the agents made them responsible. The bear had been waiting on this stretch of trail for as long as the architecture existed; Sarah's team simply had not been here on a Tuesday before.
Sarah runs.
This is what the old training tells her to do. The old training is not wrong. It just does not know what kind of animal she is running from. She runs hard, she calls the team, she opens the war room, she does what twenty years of incident response has rewarded her for doing. And the bear gains ground on her.
The Bank and the Jump
From this point forward, every system failure you encounter will present itself as one of two things: a bear you have created and you try to outrun, or a river with conditions and forces you learn to operate within.
Sarah reaches a riverbank with the bear in close pursuit and faces a choice: behind her is the bear, certainty. In front of her, a river flows wide, fast, continuous, and lethal to a runner who enters it the way old training teaches you—fighting the current and trying to cross at right angles. Expending energy on resistance, uncertainty.
Sarah stops at the edge.
This is when she sees her.
Downstream, on the far bank, the runner from the clearing hours ago is moving along the riverbank. Pack on her shoulders the way a pack rides when it has been there for years. She is not waiting for Sarah, not even looking at Sarah. She is just running, on the far side, in the rain, the way a runner runs who has been on this trail long enough to know it.
Something Sarah cannot name shifts in her chest.
The bear made the river the only direction. The runner on the far bank made the river go in the right direction. The far bank is not a void; Sarah is fleeing toward it. It is inhabited. The runner on the far side has been on this trail longer than Sarah, knows things Sarah does not yet know, and is on the far side because that is where runners who know things end up.
Sarah jumps.
What saves her is not strength. It is yielding. Going where the current wants to take her, conserving energy, reading the water rather than fighting it, trusting that downstream is a direction and not a defeat. The river does not care about her plan. The river has been here for some time. The runners who survive the river are the runners who let it carry them while they read it, and the river does not punish them for it. It punishes the ones who arrived expecting to cross at right angles.
This is what continuous motion teaches a body that has only ever known sprinting. It is not a longer version of the same effort. It is a different physiology. A sprinter running an expedition pace is not slower at sprinting; she is doing something her body has not yet learned to do. The running is closer to swimming. The lungs work differently. The mind has to stop bracing for the next finish, because there is none, and the bracing itself is what costs.
Sarah is not yet a runner who knows this in her body. She is a runner learning it in the river, which is the only way to learn it.
This is the diagnostic frame this dispatch will leave you with: the bear and the river. The same upstream event is a bear for one organization, still in seasonal-sprint mode, and a river for another, in continuous motion. The difference is not what the encounter is. It is what the organization is when the encounter arrives.
The bear is somewhere upstream, still on the bank Sarah came from, no longer the threat it was when she was running. The runner on the far bank is still moving, downstream, in the rain.
Sarah is in the water now.
The Crossing
The current carries her.
She surfaces, breathing, alive, in motion. She is not yet across. She may not reach the far bank for some time. The river is doing what rivers do; it has its own pace, its own pull, its own decisions about where it goes. Her job is to stay in the water, read the current, and not exhaust herself fighting things that cannot be fought from inside a river.
Through the rain, on the far bank, Sarah can see the clearing runner. She’s actively using something Sarah doesn’t recognize. The way the other runner handles it, runs with it, it’s clearly something special.
Sarah will need to learn what it is. Not right now, right now she has a river to cross.
What she can see from the water, even at this distance, is that the runners on the far bank are not running alone. They are not in formation. They are not pacing each other, but they are aware of one another in a continuous and structured way. The runners on Sarah’s side of the river were not. Whatever they are carrying appears to be something they share between them, something that travels between organizations and remains useful, something that lets one runner know what another runner has just encountered without either runner having to stop.
This is the infrastructure for cooperation that Sarah’s old peak-season planning approach doesn’t have. It is what makes the far bank survivable. It is also what the architectural conversation in early June will pick up.
For now, Sarah is in the water. She is breathing. She is alive. The bear is upstream and behind her. The runner on the far bank is downstream and ahead of her, moving at her own pace, running her own race, in the rain. The trail picks up on a bank Sarah has not yet reached. She will reach it.
The current carries her.
We are still here. So is the work.
The tool the clearing runner is holding is not a record of what happened. It is a record of what is happening — what the runner’s organization is committing to, in real time, at the moment of commitment, in a form that an external party could verify. Verification at transaction time, not after. This is the substance of what travels between the runners on the far bank, and what we will pick up architecturally in June.
What Comes Next
A companion checklist will follow this dispatch and the June webinar. It is the artifact runners take home from the far bank, what they carry forward, what they continue to put down, and what they watch for in 2027. The work the checklist asks of you is not work that finishes. That is the point.
The webinar in early June will pick up where this dispatch deliberately stops, on the far bank. What readiness looks like when there is no calendar to read. What does verification at transaction time mean architecturally? What the infrastructure of cooperation requires from the runners building it. Together. The companion checklist will follow as the artifact you take home from that conversation.
Next dispatch in 2027. By then, we will know more about the hybrid peak. We will know which of the protocols on this year's gear list survived their first machine peaks intact. Sarah will be on the far bank by then, with her own perspective on the crossing. So will you.
One last reflection for the leaders reading this. The encounters your teams will face in the next year are already determined by who your organizations are becoming in this quarter. On the morning of Sarah’s bear, the team that lost the encounter was a team still organized for a different kind of trail. There is no in-the-moment decision that can recover an organization from the operational posture this moment demands. The most important business choice you will make this year is not what to do during the next encounter; it is who your team is when the encounter arrives. Continuous motion is not a tactic; it’s a posture. The posture is the work.