The headline from Boris Cherny’s recent post on X might sound like another AI productivity boost story: 100% of his code contributions over 30 days were written by AI. The data is staggering—259 pull requests, 497 commits, 40,000 lines added, and 38,000 removed—all generated without a single line from him. But the real insight isn’t about AI taking over coding tasks. It’s about what this means for engineering leadership in 2026 and beyond.
The invisible shift in leadership workloads
If you manage a team today, the most critical takeaway from Cherny’s stat isn’t that AI can write code—it’s that the head of a product team no longer needs to be the most prolific contributor. More surprisingly, they may no longer be the most active reviewer either. AI systems now handle the initial pass on pull requests, leaving human leaders to focus on higher-level decisions.
This transition fundamentally changes what leadership looks like. The artifacts—specs, PR descriptions, and even some code—are increasingly generated by AI. What remains is the decision-making process: determining what to build, what to deprioritize, and when to intervene. These calls are the new core of leadership work, yet they leave little trace in traditional performance metrics.
Leadership decisions in an AI-driven workflow
Consider another recent example: ServiceNow’s AI agent kill switch demo at Knowledge 2026. The system demonstrated how an AI agent could detect a prompt-injection attack, assess the blast radius, and surface a kill switch for human approval. Most coverage focused on the security infrastructure, but the deeper implication is about leadership responsibility.
The artifact—the kill switch feature—was designed and shipped by a vendor. The critical decision—when to activate it—rests with human leaders. This mirrors Cherny’s stat: AI handles the production, but the judgment calls remain with people. The challenge for leaders isn’t building the system; it’s defining the thresholds and criteria for intervention.
The work that doesn’t leave a trace
Traditional performance reviews reward visible outputs: lines of code, merged pull requests, or completed design documents. But the most valuable work in 2026 often doesn’t produce artifacts at all. It happens in Slack threads, quick calls, or spontaneous decisions that shape a product’s direction.
For example:
- A senior engineer responds to a question about shipping a feature in two sentences with three clear reasons. No commit log captures this.
- A manager decides to roll back a deployment after AI flags a regression. The decision is documented in a post-mortem, but the real work happened in the moment.
- A tech lead approves AI-generated code but asks, Does this align with the product’s intent? The answer isn’t a line of code—it’s a judgment call.
These moments compound over time, yet they’re invisible to systems that measure productivity by output. The senior engineer measured by code volume or design-doc count is being evaluated against a 2023 standard. The leader measured by decision quality is operating in 2026.
Why measurement systems are failing leaders
Performance reviews and promotion processes were built for an era when the artifact was the work. They reward what got shipped, not what got decided. When a senior engineer’s contributions are evaluated by lines of code, they’re being measured against yesterday’s definition of leadership.
The problem isn’t the engineers—it’s the systems. If your performance review asks for ship counts but never probes your call log, the framework hasn’t evolved with the work. The most impactful decisions often leave no digital footprint, making them invisible to traditional metrics.
Three actions to align with 2026 leadership
Adapting to this shift starts with small, deliberate changes:
- Start a private call log this week. For each significant decision, write a single line: What was decided, what was the alternative, and why? In the first week, it may feel trivial. By month two, it becomes the artifact your performance review lacks—a record of work that otherwise vanishes.
- Lead with your decisions in promotions and career conversations. Instead of saying, I shipped X, frame it as, I decided X over Y because Z, and the outcome was W. The evidence you present changes when the work changes.
- Observe the leaders already operating this way. Identify the team member whose daily work is 80% judgment calls rather than artifact creation. Their routine is the role you’re evolving toward—and it’s often quieter than expected.
The career arc no one predicted
The path from senior individual contributor to staff engineer to manager has always meant trading authorship for direction. What’s new in 2026 is the speed of this transition. AI now handles artifact production at every level, compressing the curve from coding to leadership decisions. The senior leader still measured by lines of code is being evaluated against an outdated standard.
Cherny’s stat is the clearest data point yet: The head of a team stopped writing code, and the team kept shipping. This isn’t a productivity story—it’s a leadership story. The systems measuring these leaders haven’t caught up to the work they’re actually doing. The question isn’t whether AI will transform engineering leadership. It’s whether we’ll adapt our measurements before it’s too late.
AI summary
Claude Code’nin başkanı 30 günde 40 bin satır kod yazmadı. Peki, liderliğin geleceği koddan değil, karar almaktan geçiyor olabilir mi?
