On 23 September, hours before Xi Jinping landed in Washington for a three-day state visit, US Treasury Secretary Scott Bessent went on television and announced that the trade truce between the two countries would be extended. Not by six months, which is what most observers expected. Not by three, which the US Trade Representative had hinted at. By two: from 10 November to 10 January 2027.
The number is the story. A short extension in a negotiation this size is not a technicality. It is a message, and it tells you more about how the two sides see each other than any joint statement could.
What actually happened
The truce dates from October 2025, when Trump and Xi met in Busan after a year in which tariffs on both sides had climbed past 100 per cent at the peak. Under that arrangement China suspended its retaliatory tariffs on many US farm products, committed to buying US soybeans, paused new export controls on rare earths and critical minerals, and acted on fentanyl precursors. The US suspended or reduced a set of its own tariffs and export restrictions. Both sides kept the underlying disputes open.
By September, the scorecard was mixed. Bessent said China was meeting its commitment to buy 25 million tons of soybeans, but was behind on a pledge to buy 17 billion dollars of other farm goods. US officials said rare-earth deliveries were "not up to par". One analysis citing unnamed officials put them around a quarter below the averages seen before China introduced its controls. Talk of a Chinese order for US aircraft had shrunk from 500 planes to 200, with no types or delivery schedule attached.
Bessent's own summary, after meeting Vice Premier He Lifeng: "There are some deliverables that have not been perfect on the Chinese side."
What came out of the visit itself was thin. The truce was extended. The two sides agreed to set up a channel for serious AI safety incidents. Trade Representative Jamieson Greer described a "Board of Trade" process that might identify goods for reciprocal tariff cuts. Tariff levels after January, farm purchases beyond soybeans, rare-earth volumes and technology restrictions were all left for later rounds, with two multilateral summits in Shenzhen and Miami pencilled in before the year ends.
Four things in that sequence are worth taking apart, because each one shows up in negotiations a thousand times smaller.
1. When nobody can enforce the deal, the deadline is the enforcement
There is no court above the United States and China. No arbitrator who can order a shipment of rare earths. Every commitment in the Busan arrangement rests on the other side's willingness to keep it, and on what it costs to stop.
In that setting the expiry date is the only real lever either side holds. A six-month extension would have said: we trust the trajectory, go and plan. A two-month extension says: perform, then we talk about longer. Washington kept the deal alive and kept the pressure on, in the same move.
The cost is paid by everyone else. Every company that imports from China, sells into it, or sits in a supply chain that passes through it now has a new cliff on 10 January. Leverage for the negotiator is uncertainty for the people downstream, and a short deadline buys the first by spending the second.
A deadline you control is the cheapest form of enforcement you will ever get. Use it, but know who pays for it.
At your own table this is the review date. Not "let's see how it goes", but "we review on the first of March, and the renewal depends on what we see". The renewal date is yours to set. Set it close enough to matter.
2. A commitment you can't measure isn't one
Look at the three Chinese commitments that were in the news and notice which ones caused arguments.
- 25 million tons of soybeans. A quantity, a product, a clear way to count. Bessent said it was being met, and nobody disputed it.
- 17 billion dollars of "other farm goods". A number attached to a category. What counts, at what prices, over what period? Behind schedule, said Washington. Still in progress, says Beijing.
- Suspending rare-earth export controls. A promise about a process, not an outcome. The controls are suspended; the deliveries are short. Both sides can claim to be right, because the commitment never said how many tons would leave the port.
The pattern is not about trade. It is about how terms are written. A term that a stranger could check with a spreadsheet holds. A term that needs interpretation turns into the next negotiation.
When you close your next deal, run every commitment through four questions. How much? By when? Who measures it, with what? And what happens if it is missed? If any answer is "we'll figure it out", you have not agreed on that point yet. You have agreed to argue about it later.
3. Linkage cuts both ways
The trade talks did not stay trade talks. Xi, according to Chinese state media, asked Washington to handle Taiwan "with prudence" as a basis for cooperation. Earlier in the year Trump had called a pending arms package for Taiwan a negotiating chip. Security moved onto the trade table, and trade onto the security table.
Linking issues is one of the oldest moves in negotiation, and for good reason: it creates trades that would not exist inside a single topic. I give on something cheap to me and dear to you, you give on the reverse. The pie gets bigger.
It also blurs accountability. Once Taiwan and soybeans sit on the same table, a miss on soybeans can be answered with a complaint about Taiwan, and the question "did you deliver what you promised?" has no clean answer. The two sides end up negotiating about what the negotiation is about.
So link, but say what you are linking. "If the volume commitment is met, we extend the exclusivity" is a linkage. "We'll look at the whole relationship" is a fog. Name the trade out loud and write it down in the same sentence.
4. "We'll keep talking" is itself a deal
The easiest thing to miss in Washington is that the most concrete outcome was procedural. A Board of Trade process. An AI incident channel. Two summits with dates. None of that moves a ton of soybeans, and all of it matters.
When substance stalls, the professional move is to negotiate the process: who meets, when, with what agenda, reporting to whom. A process agreement keeps the counterpart engaged, creates checkpoints you can hold them to, and gives both sides something to announce that is true. It is not a consolation prize. It is the scaffolding the next substantive agreement gets built on.
If you leave a meeting with no deal, leave with a date. "We reconvene on the 14th, you bring the revised volumes, I bring the pricing" is an outcome. "Let's stay in touch" is how negotiations die quietly.
At your own table
You will probably never negotiate rare earths. You will negotiate a vendor renewal, a salary review, a partnership with a company that keeps saying "in principle, yes". The moves are the same, and the Washington case gives you a checklist.
- Set a review date instead of accepting open-ended goodwill. Keep it short enough that the other side has to perform before it.
- Write every deliverable so that a stranger could check it: quantity, date, who measures, consequence.
- If you link issues, name the link. If they link issues, make them name it.
- Never leave the room without the next conversation booked, with an agenda.
None of this is hard to understand. All of it is hard to do under pressure, with someone across the table who would rather you accepted the fog. That gap, between knowing the move and making it when it counts, is the one worth practising, and the only safe place to practise it is against a counterpart that is not real. On Pacct that counterpart is an AI, and the debrief afterwards tells you where you drifted. Try the checklist there before you need it.
