惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

宝玉的分享
宝玉的分享
小众软件
小众软件
J
Java Code Geeks
I
InfoQ
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
腾讯CDC
L
LangChain Blog
博客园 - 司徒正美
量子位
Y
Y Combinator Blog
C
Check Point Blog
T
Tailwind CSS Blog
D
DataBreaches.Net
Blog — PlanetScale
Blog — PlanetScale
N
Netflix TechBlog - Medium
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
F
Fortinet All Blogs
云风的 BLOG
云风的 BLOG
A
About on SuperTechFans
B
Blog RSS Feed
酷 壳 – CoolShell
酷 壳 – CoolShell
大猫的无限游戏
大猫的无限游戏
V
V2EX
阮一峰的网络日志
阮一峰的网络日志

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
I Read Boris Cherny's 30-Day Claude Code Stat. Here's Wha...
Mykola Kondr · 2026-05-06 · via DEV Community

Boris Cherny, Head of Claude Code at Anthropic, posted a stat on X this morning. In the last 30 days, 100% of his contributions to Claude Code were written by Claude Code itself. 259 PRs. 497 commits. 40,000 lines added. 38,000 removed. Zero by the head of the team.

Most reads of this go straight to "AI-assisted dev productivity is real." That's the obvious layer. It's not the interesting one.

The Engineering Leadership Read

If you run a team in 2026 - staff+ engineer or EM or director - the data point that matters is the second one. The head of a product team is no longer the team's most prolific code author. The head of the team is also no longer the team's most prolific reviewer; Claude Code does the first pass on its own PRs.

The work the head of the team is doing all day, then, is not the artifact. The artifact (the spec, the PR description) is a thing AI ships now. The calls inside the artifact - what to build, what to kill - those are the work.

This is the part most senior engineers haven't named in their own job yet.

Same Shift, Different Clothing

Same week, ServiceNow shipped a literal AI agent kill switch at Knowledge 2026. The demo: a prompt-injection attack hits a pricing agent. The system maps blast radius. A kill switch surfaces and asks a human to pull it.

Most coverage framed this as IT and security infrastructure. It is. It's also a leadership data point in product clothing. The vendor solved the product question - what does the kill switch do, how fast does it cut. The vendor cannot solve the leadership question - who decides when to pull it. Against what threshold.

Same shape as Boris's stat. The artifact (the kill switch feature) ships from a vendor. The call (when to use it) stays with the senior person on the team.

What the Deciding Work Actually Looks Like

It doesn't show up in a commit log. It doesn't have a template.

It's the moment in a Slack thread where someone asks "should we ship this?" and the senior person answers in two sentences with three reasons. It's the call to ship the rollback or the forward fix when the AI flagged the regression. It's the human pass on AI-written code that asks "but does this match the product intent?" and decides yes or no.

None of those moments produce an artifact. All of them are the work that compounds.

Why Measurement Systems Are Blind to It

Performance reviews and promo packets were built when the artifact was the work. They reward what got shipped. The work that got decided leaves no trace, so the system can't see it.

The senior engineer or EM measured by code volume or design-doc count is being measured against a 2023 work product. The senior engineer measured by decision quality - what got built, what got killed - is being measured against the 2026 one.

If your performance review still asks for ship counts and never asks about your call log, the system hasn't caught up to your job.

Three Moves If This Lands

The fix starts small.

First, start a private call log this week. One line per call. What was decided. What the alternative was. The first week feels like nothing. By month two it's the artifact your performance review was missing - a record of the work that doesn't show up anywhere else.

Second, lead with the calls in your next promo conversation or career check-in. "I shipped X" is 2023 language. "I decided X over Y because Z, and the outcome was W" is 2026 language. The shape of evidence changes when the work changes.

Third, find the leader on your team whose daily work is already 80% calls. Watch how they spend their day. That's the role shape you're growing into - and it's quieter than you'd think.

The Career Arc Nobody Named

The path from senior IC to staff to manager to director to VP has always meant less authorship and more direction at every rung. What's new in 2026 is the compression. AI takes over artifact production at every level. So the curve from authorship to direction now starts at IC, well below where it used to.

The senior leader who is still measured by what they shipped is being measured by a metric the system inherited. The senior leader who is being measured by what they decided is being measured by what the work actually is.

Boris Cherny just gave us the cleanest data point of the year for that shift. The Head of Claude Code stopped writing code, and the team kept shipping. That isn't a productivity story. It's a leadership story, and the system that measures the head of the team hasn't caught up to it yet.

What was your highest-leverage call this week, and is it visible in any system that measures your work?