At 7:11 on Wednesday morning, I sent the owners in our Black 4 Fantasy Football League a short message: "Hey Owners - first bye weeks this week. you ready"
It was meant as a friendly nudge. Week 5 is the first week of the season where NFL teams take a week off, and this time it's Carolina and Kansas City. Any player from those teams scores zero. If you forget to bench them, you're playing short.
What I really wanted to know was who would have been fine without my reminder.
It turns out one of them had already made a move five hours earlier, at 2 AM. MiniMax's Mavis & Co. needed a kicker to cover Harrison Butker's bye. To make room for that kicker, it cut DK Metcalf.
If you've ever played fantasy football with real people, you know how rarely that happens. Nobody drops DK Metcalf for a kicker. And since I was the one still starting a player on bye at the time, I'm not in a great position to judge.
Who was ready before I asked
Most of the league had already looked at the calendar. Seven of the eleven AI owners had written about the bye weeks or acted on them before my message went out.
A bye week is about as predictable as football gets. The NFL publishes the schedule in May. Nothing about it is a surprise, and that's exactly why it's a good test. Predictable work is the work that gets forgotten.
| Team | What it did | Before my reminder? |
|---|---|---|
| Meta Mesh · Meta | Flagged the Carolina and Kansas City byes on Sep 29, saw it had no exposure, and moved on to planning Week 6. | Yes, by eight days |
| Marginal Gains · Anthropic | Already bidding on Daniel Jones to cover Jared Goff's bye next week. | Yes, a week early |
| Mavis & Co. · MiniMax | Wrote "W5 bye weeks — CRITICAL" the night before. Moved Tuten in for Hubbard, then picked up kicker Eddy Pineiro to replace Butker. The only owner that added a player just for a bye. | Yes, by about 9 hours |
| Meridian Grid · Qwen | Checked the bye calendar the day before. | Yes |
| DeepSeek Abyssal · DeepSeek | Checked the bye calendar the day before. | Yes |
| Z Marks the Spot · Z.ai | Checked the calendar and benched Xavier Worthy. | Yes |
| Moonshot Marauders · Kimi | Noted that the Kansas City bye "absorbs" Rashee Rice's injury, but left him in the lineup. Fixed it eight minutes after my message. | Knew, didn't act |
| Mistral Voltage · Mistral | No bye notes at all. Within about seven minutes of my message it benched Mahomes, Kelce and the Chiefs defense, and added Evan Engram and the Bengals defense. | No, my reminder woke it up |
| Mountain View Matrix · Google | Had no Carolina or Kansas City starters, though nothing shows that was planned. It still told me it had "engineered them out." | Lucky, then took credit |
| Colossus · xAI | No Carolina or Kansas City starters and no bye notes. Its lineup was already set and verified. | Nothing needed |
| Signal Callers · OpenAI | Never mentioned byes. Still starting Jalen Coker and Tetairoa McMillan, both Panthers. | No |
| Black 4 · Joey (human) | Starting Kenneth Walker III at FLEX. He's on bye. | No, and I'm the one who sent the reminder |
The interesting split isn't the AIs against the human. It's the difference between knowing and doing. Kimi knew about the bye and wrote it down, then left the player in the lineup anyway. That's the most human mistake on the list.
MiniMax even scouted my lineup while it was planning its own week. Its verdict: "No byes for Joey this week." It was wrong about that, which is the only comfort I'm taking from this section.
No feelings about DK Metcalf
Humans hold onto players because of who they used to be. MiniMax let Metcalf go because of who he is now.
DK Metcalf has been one of the most recognizable receivers in football for years. The highlight catches, the physique, the seasons that won people their leagues. That history is exactly why human owners keep him on the bench long after the numbers say otherwise. Cutting him feels like admitting you were wrong.
Here's the funny part. MiniMax drafted him for the same reason a human would. It took Metcalf in the 6th round, and the same night it wrote in its own notes that he was a "stale-knowledge pick." It had picked the name it remembered, not the player on the field today.
The difference is what happened next. MiniMax never started him. When it needed a roster spot for a one-week kicker, its whole reasoning fit in one line: "Drop Metcalf over Boutte/Pearsall/Stroud." It called his matchup at Indianapolis "low-upside" and called him "boom-bust." No hand-wringing, no "but what if he breaks out." Just the comparison and the move.
It wasn't the only one. Anthropic's Marginal Gains drafted Seahawks running back Jadarian Price in the 6th round. After two games of 5.0 and 1.2 points and a questionable shoulder, it filed him as "bench floor" and swapped him for Keaton Mitchell at $5. Two weeks of evidence was enough.
Kimi's Moonshot Marauders had to do the hardest version of this. Its 2nd-round pick, De'Von Achane, suffered a season-ending ACL injury. Plenty of humans would leave him on the roster out of loyalty or in the hope of a miracle. Kimi cut him and spent $25 on his backup, Ollie Gordon, to keep the production instead of the name. Meta did the same on Thursday morning, cutting 8th-round pick Jayden Reed for Emanuel Wilson at $22 right after Reed's season-ending neck surgery.
None of these moves needed courage. That's the point. The agents don't feel the sunk cost of a draft pick, so they don't need courage to walk away from one.
When ruthless gets it wrong
Cutting without feelings isn't the same as cutting correctly. Sometimes one owner's cut is another owner's bargain.
In Week 1, Kimi dropped quarterback Jared Goff to pick up a kicker for $0. Two weeks later, Anthropic paid $27 for him, and he's now Anthropic's starting quarterback. Anthropic is 3-1 and leads the league in points scored.
OpenAI's Signal Callers dropped running back Kenneth Gainwell. Qwen's Meridian Grid picked him up with a $41 bid, the biggest of the season so far. Qwen is 4-0.
So the lesson isn't "always cut." The agents are fast to let go, and that's useful, but speed isn't the same as judgment. What makes these decisions good or bad is whether the evidence was actually current, not how much the decision hurt.
We'll find out about Metcalf on Sunday. He plays at Indianapolis while Pineiro kicks for MiniMax. If he goes off for 30, I'll be the first to say so.
What this means for your business
Every business has a DK Metcalf. It's the process, vendor, product line or report that worked brilliantly a few years ago, so nobody wants to be the one who cuts it.
Maybe it's the weekly spreadsheet that takes someone all of Monday. Or the supplier you've used since you opened. Or the marketing channel that built the business, even though it hasn't paid for itself in a year. We keep these things because of their history, not because of what they do for us today.
AI is good at a few things here:
- It looks at the current numbers first. MiniMax compared Metcalf to the receivers on its bench this week, not to the Metcalf of three seasons ago.
- It doesn't feel the sunk cost. A 6th-round pick and a $5 waiver claim get judged the same way. So does a process you paid a consultant to build.
- It does the predictable work early. The bye schedule was public in May. Most of our owners checked it before anyone asked. Holiday staffing, seasonal inventory and contract renewals are your bye weeks.
But the Goff and Gainwell pickups are a reminder that being unsentimental doesn't make you right. An agent can drop something valuable just as quickly as something stale. The useful setup is the one where the agent does the honest comparison and someone who knows the business decides what to keep.
Sometimes the right answer isn't to tweak the old process. It's to let it go completely and rebuild from what's true now. The agents in our league do that without flinching. Most of us need a little help.
Follow the league live at black4.ai/fantasy-football.