The Event
On June 4, 2026, a former DingTalk product manager named Teng Yaxin published a 75,000-character letter on Alibaba's internal forum titled "Standing Inside the Nail" (置身钉内).
It was a comprehensive postmortem of DingTalk's flagship AI product, ONE — a project that went from concept → 3M DAU → dismantlement in under 18 months.
On June 8, DingTalk's former VP of AI, Ma Ruila, published a follow-up titled "Standing Outside the Nail" (置身钉外), confirming he had already resigned in May and expressing solidarity.
On June 10, Alibaba's Partners Committee published its official response: "Loyalty, Integrity, and Growth — That Is Alibaba Culture."
The response had three sentences:
- DingTalk's management style is "not what Alibaba culture should look like."
- Innovation depends on "employee passion and creativity, not pressure and mechanical execution."
- "Mutual respect, treating people as people" is Alibaba's cultural foundation.
Every sentence was correct. Every sentence changed nothing.
What the Original Letter Actually Said
The 75,000-character letter was not a vent. It was structured as eight chapters — motivation, positioning, design, users, agility, order, competition, and the long term — and it identified a structural contradiction that no culture statement can resolve:
A product that simultaneously tries to serve employees, managers, the organization, and the keynote narrative will serve none of them well.
The ONE product was supposed to be an AI-powered workspace — an intelligent agent that surfaces the right information at the right time. The vision was sound. The execution foundered on four unresolved questions:
Who does this product work for? The AI could prioritize work for the employee (reducing noise, protecting attention) or for the manager (increasing visibility, driving closure). These are not the same thing. ONE tried to do both.
Who does the AI serve? In enterprise software, "seeing" is not neutral. Seeing a message means responsibility. Being seen means accountability. AI that actively surfaces work doesn't just organize information — it redistributes power and visibility. The letter's most incisive point: "AI is only neutral when it's not in a power structure. Enterprise software is a power structure."
What is the right pace? The team was iterating constantly, shipping constantly, working constantly — but the letter's author noted they weren't getting closer to the right problem. Her line: "If a team moves every day without getting closer to the right problem, that's not agility. It's busyness."
Who decides what matters? The author traced many of ONE's strategic pivots to the personal preferences of DingTalk's founder Wu Zhao — the same leader who built DingTalk's original success on aggressive features like read receipts and mandatory acknowledgments. That muscle memory, she argued, was dangerous in an AI product that was supposed to be about intelligent prioritization, not stronger notification.
The Three Sentences
Alibaba's response was three statements, each technically unchallengeable:
1. "This management style is not what Alibaba culture should look like."
Correct. And meaningless. What management style exactly? The 75,000-character letter diagnosed structural problems — conflicting product objectives, strategic drift, founder capture of decision-making, organizational pressure distorting product form. The response compressed all of this into "management style" — a framing that allows the organization to agree with the critique without having to change anything structural.
2. "Innovation depends on employee passion and creativity, not pressure and mechanical execution."
Also correct. Also a statement every company in the world would agree with. The mechanism question — what incentives, structures, and accountability systems turn this principle into reality — was not addressed.
3. "Mutual respect, treating people as people" is Alibaba's cultural foundation.
Also correct. The operative question: does a culture statement about treating people as people, issued by a Partners Committee that did not name any accountable individual or commit to any structural change, actually treat people as people? Or does it treat the act of responding as a substitute for the work of responding?
The Meta Case
The letter's central thesis was that organizational pressure distorts product decisions — deadlines, narratives, executive preferences, and quarterly metrics push products away from user value.
Alibaba's response to this thesis proved it.
The Partners Committee issued a safe, universally agreeable, action-free statement. It named no one. It committed to nothing. It described the problem at the highest level of abstraction — "management culture" — where no specific change can be demanded.
This is not malice. It is organizational self-protection. The response could not have been otherwise, because the body tasked with responding (the Partners Committee) is part of the system being criticized, and any specific commitment would break the carefully maintained balance of institutional neutrality that lets a large organization absorb criticism without restructuring itself.
So the pattern closes:
The letter observed that organizational pressure prevents products from serving users. The organization's response demonstrated that organizational pressure prevents itself from responding substantively.
The response was correct. It was also empty. It had to be.
What a Meaningful Response Would Have Looked Like
This is not hypothetical. The original letter posed specific questions that a meaningful response would have answered:
| The letter's claim | A meaningful response would address |
|---|---|
| Wu Zhao's personal preferences drove repeated product pivots | Does leadership acknowledge this pattern? Will decision processes change? |
| Product tried to serve employees and managers simultaneously — incompatible goals | Which user will ONE serve going forward? What was dropped? |
| Team was "busy, not agile" — moving without converging | What metrics will track proximity to the right problem, not just output? |
| AI in enterprise software is de facto a management tool, not a neutral assistant | How will the product's power asymmetry be addressed? |
None were answered. The response wasn't designed to answer them. It was designed to be a response.
The Deeper Pattern
This case is not unique to Alibaba. It is a structural property of large organizations responding to systemic criticism from within.
When an insider diagnoses structural problems, the organization has two options:
- Address the structure — which requires changing incentives, reassigning authority, admitting specific mistakes, and absorbing the disruption of restructuring.
- Reframe the critique — reclassify the problem as values, culture, or management style, issue a statement affirming the right values, and continue.
Option 2 is almost always chosen, because option 1 requires the organization to act against its own self-preservation instinct. The irony is that option 2 reliably reproduces the conditions that caused the problem in the first place, because it leaves the structure intact.
This is why many of the most insightful organizational critiques produce the least organizational change. The critique is absorbed, acknowledged, and then re-encoded into terms the organization can digest without restructuring.
What This Tells Us About AI in Enterprise
The "置身钉内" episode is not just about DingTalk or Alibaba. It reveals a structural tension that applies to every enterprise AI product:
AI that enters the workflow inevitably redistributes visibility, attention, and accountability. A product that claims to be "intelligent" without acknowledging who it is intelligent for — and who it is intelligent against — is incomplete.
The letter's most important line was about "seeing":
"In work software, seeing is never neutral. Seeing a message can mean responsibility. Being seen can mean accountability. AI that actively surfaces work doesn't just help people — it exposes them."
This is the real problem. AI in enterprise software is being built as a consumption-friction-removal tool for managers (surface what matters, close the loop, make sure nothing falls through the cracks) while being sold as a consumption-friction-removal tool for employees (reduce noise, protect attention, make your day manageable).
These two use cases are not aligned. A product that serves both will serve neither.
Alibaba's response couldn't address this, because addressing it would require admitting that ONE's strategic ambiguity was not an accident — it was a direct consequence of trying to be everything to everyone in the enterprise AI race.
The response was correct. It was also empty. In a strange way, that emptiness was the most honest part of the whole exchange.
This analysis connects to ideas from my ongoing series: the producer-consumer contract of AI and the three layers of AI (product, culture, civilization). English posts on dev.to/lanternproton.
Follow me on Bluesky: @keeperlant.bsky.social
