The feed shows another next move: one lab opens weights, another speeds releases, another escalates the safety story. You read it like intelligence. Your hand already wants to copy it into the organization. Variation in Tactics does not ask who copied better. It asks 途有所不由 — which public road you should refuse today. You write the refusal before you follow the news.
Public moves are not your menu
You did not invent this impulse. Pages already ask whether you should open-weight like a lab, chase release cadence, or turn a safety story into a slogan. The more the title looks like a playbook, the more you are replaying someone else’s fight.
So you can refuse “copy their next move” before you follow the headline. Public moves are visible structure: open-weight posture, rapid release cadence, safety-narrative escalation. Both labs publish model releases and version notes in public; OpenAI has at times published open-weight efforts; Anthropic keeps publishing Constitutional AI and responsible-scaling safety narratives. You read those layers to see the board — not to treat their news as your order of battle.
The Art of War is not a license for you to reproduce a headline. It is a constraint you can still use while the feed argues about who moved first.
You can also refuse “they can walk it, so can we” as a reason. OpenAI’s and Claude’s public actions are their resources, partners, and narratives rewriting a choice. Your organization has another object, goal, and resistance. So you can leave the playbook to pages that sell a follow-the-lab war. You keep 不由 — the road not taken — from 九变.
途有所不由: write which road you refuse
The live judgment you use on this site is short. When conditions change, tactics change. Some roads and some fights should be refused. The aim is preservation, not pride.
You read that judgment on Refuse. The original lines stay there as evidence, not as decoration. There are roads that should not be taken, armies that should not be attacked, cities that should not be besieged, ground that should not be contested — and orders that need not be obeyed when they violate that variation.
Three publicly verifiable moves help you change the question. Open-weight posture is one road: opening weights changes distribution, governance, and where responsibility sits. Rapid release cadence is another: the calendar gets walked by versions, and verification and stop-loss get squeezed. Safety-narrative escalation is a third: the room starts mobilizing on slogans while the plan may still be blank.
All three are public chess moves. They are loud, screenshot-friendly, easy to narrate as “must follow.” Completeness is not you retelling whose next step in the group chat. Completeness is whether your object, goal, and resistance can still be preserved before you copy.
So you can write those three moves on the news in front of you. Open weights are posture. Releases are rhythm. Safety narrative is the story layer. If one step is fog or a trap for you, you do not have a must-reproduce problem. You have a road to refuse. Stopping is not falling behind. Stopping is 途有所不由.
The 用间 essay covers missing fact before you move. This piece is different. The facts may already be enough: both labs publish releases; posture and narrative sit in public. There is still a road you refuse. Enough intelligence does not decide whether you should copy.
Write object, goal, resistance — then refuse one road
You name the object: the person, team, or system that must actually move in this “should we copy” choice — not an abstract “AI giant.” You name the goal: a result that makes the next cycle better, not “looked like we kept up.” You name the resistance: the real block, not the headline you dislike.
Then you ask one question. Will this public move burn your calendar, your responsibility boundary, or an irreversible state with it? If yes, you write: this road is one you refuse. If no, you still write a trigger: what fact must appear before you move; what condition means you stop.
If the refusal is clear, you do not need the news’s next move as a command. You write one refusal. You verify only that condition. Preservation outranks looking alike.
Then you stop treating laboratory news as a menu of operations. Your review is three sentences: the judgment, what you verified, what would make the next cycle better. If you skipped copycat follow-through, the review is shorter. That is the point.
People will tell you that refusing a copy means you are behind. The chapter does not say you never move. It says some roads should not be taken, some forces should not be struck. If you skip 不由 because you want to look like you read the board, you chose the expensive road on purpose.
So you can keep the constraint without announcing it. Ask which road you refuse first; then change the method. You do not owe the room a lecture that retells two laboratories. You owe the next cycle one fewer wasteful fight.
Waging War covers consumption and the calendar; Attack by Fire covers anger and what cannot be restored. This piece covers organizational “don’t copy”: facts may be enough, and there is still a road you refuse. Do not make stop-loss dates or venting the main frame here. What you stop today is copying a public move.
知彼知己×AI covers your tool boundaries. This piece covers whether the organization should reproduce someone else’s news rhythm. Different doors. If you arrived looking for commercial warfare or “how to beat a model,” this is the wrong door.
This site is not an AI shopping guide. The case exists only to serve fewer unnecessary fights. You are here to open one fewer wasteful fight — not to collect a follow-the-lab menu.
Questions people actually ask when “next move” tempts a copy
If both labs are moving and I refuse one step, will I look behind?
You might. The chapter does not protect your image in the next hour. It protects the whole. Some roads should not be taken. If you hold the room by retelling the news and skip object, goal, and resistance, the round was a small follow-the-lab fight. So you can let “must look like them” go. You write which road you refuse. Preservation outranks looking current. Keeping up is not copying a playbook.
Open-weight posture is loud — must I open something too to stay “industry”?
No. A public open-weight move is their resource and governance choice, not your default menu. Variation in Tactics asks 途有所不由 first. So you can write whether opening weights would rewrite your responsibility boundary, distribution, and verification cost. If it would, and the next cycle cannot stay whole, you write the road you refuse. Leave “must open-source to look advanced” to pages that sell a posture war. Your job is one refusal, not reproducing the news.
Releases move fast and the team says lagging equals losing — how do I pull talk back to refuse?
You do not need to win a speed debate first. You put three lines on the table: who the object is, what the goal is, where the resistance sits. Then you ask what calendar a forced follow would burn, and whether you can still stop. So you verify a trigger and a refusal, not who ships more often. A release is visible rhythm, not your order of battle. You keep the road you refuse before you debate the follow.
Does copying a safety or alignment story into slogans mean we “did it right”?
Not automatically. Public safety narratives are one layer of the move: they change how others read risk and roadmap. You do not have to swallow every line, and you should not ignore the whole layer. You check whether the narrative rewrote the plan and boundary on your table. If you only escalate slogans while the plan stays blank, you already chose an expensive road. You verify the plan, not how good the slogan sounds.
If the facts are already clear, why refuse a road — how is this different from 用间?
用间 covers missing fact before you move. This piece admits the public move is often already on the table; the facts may be enough. Enough still leaves 途有所不由. So you write which road you still should not take, strike, or contest. Full intelligence does not decide whether you should copy. You stop the copy, not the knowing.
Leadership ordered “do their next move” — must every order be taken as-is?
九变 says some ruler’s orders need not be obeyed when they violate the advantage of variation. You are not staging defiance theater. You make the refusal concrete: object, goal, resistance, and what a copy would burn. So you write a verifiable condition, not an emotional fight. An order can be discussed; fog is not strategy. You keep the judgment before the execution speech.
How is this different from the “who wins” essay or 知彼知己×AI — which should I read?
The “who wins” essay refuses siege scorekeeping and reads 伐谋 and 伐交. 知彼知己×AI covers tool boundaries. This piece covers whether the organization should copy a public move, and lands on Refuse from Variation in Tactics. So if “next move” is walking you into reproducing the news, use this one. You need one refusal, not another encyclopedia. You pick the page that stops an unnecessary fight.
The next move is not reproducing laboratory news. You write one public move you will refuse this week, and the aim that stays. If it fits, Contact Us — teaching, training, and content, not crushing a rival, and not winning a follow-the-lab war. We reply within two business days. You stop copying public moves. Let the next cycle open with one fewer wasteful fight.