Tech Laid Off 165K in 2026, but ByteDance’s Frontend Hiring Jumped 23%
programs.com updated a dataset last week: In the first seven months of 2026, the global tech industry has laid off over 165,000 people, with more than 91,000 of those positions explicitly attributed to AI replacement. 55 CEOs directly mentioned 'efficiency gains brought by AI' in their layoff announcements.
Oracle cut 21,000 jobs in one go. Microsoft laid off 4,800, stating clearly in its announcement: 'AI enables us to accomplish the same work with fewer people.'
But during the same period, ByteDance's 2026 campus recruitment in China opened over 5,000 positions, with frontend/client-side roles growing 23% year-on-year. Taotian Group expects to issue around 1,000 offers, with technical roles accounting for over 90%.
Layoffs on one side, hiring on the other. Contradictory?
Not at all. The people they are laying off and the people they are hiring are not the same kind.
Disappearing Roles vs. Emerging Roles
I looked at JD.com's frontend job postings on BOSS Zhipin. One obvious phenomenon: a large number of traditional 'Frontend Development Engineer' positions are closed, while the requirements in the job descriptions for the roles still open have already changed.
A frontend role requirement at JD Technology (still open):
'Responsible for the frontend technical construction and implementation of JD Technology's C-end AI products, deeply understand AI business requirements, conduct system analysis and architecture design'
Another frontend role requirement at JD Mall:
'Use frameworks like TensorFlow to reconstruct existing systems or tools'
JD Supply Chain Team:
'Based on natural language processing, knowledge graphs, and deep learning technologies, realize the reverse engineering from consumer-side demand to product design'
These are not backend positions. These are all positions with the title 'Frontend Development Engineer'. But what they require has far exceeded HTML/CSS/JavaScript.
Data: Who Got Laid Off
Breaking down the 165,000 layoff data:
| Company | Layoff Count | Stated Reason |
|---|---|---|
| Oracle | 21,000-30,000 | AI investment + data center transformation |
| Microsoft | 4,800 | 'AI efficiency gains' |
| Cisco | 4,000 | Business restructuring |
| Etsy | 12% of staff | CEO explicitly said 'not AI-driven' but simultaneously increased AI investment |
| Remaining 55+ companies | Total 90,000+ | Most mentioned AI |
These layoffs share a common characteristic: they cut the 'execution layer' and retained or added to the 'judgment layer'.
What does that mean?
- Execution layer: Given a design mockup, write the page, connect the API, go live. Clear inputs, standardized outputs.
- Judgment layer: Make technology choices, define architecture plans, evaluate the quality of AI output, handle edge cases that AI cannot solve.
AI can write code now. It writes it fast and well (most of the time). So 'being able to write code' itself is no longer a scarce skill. What is scarce is the ability to 'judge whether this code should be written this way'.
This explains why ByteDance's frontend roles are growing—they are not hiring 'people who write pages,' but 'people who can command AI to write pages and judge the quality of the result.'
Keywords Disappearing from Job Descriptions
Comparing frontend JDs from 2024-2025 with those from 2026, several obvious changes:
| Disappearing/Weakening | Strengthening/Newly Added |
|---|---|
| 'Proficient in HTML/CSS/JavaScript' | 'AI product frontend technical construction' |
| 'Familiar with Vue/React frameworks' | 'Deep understanding of AI business requirements' |
| 'Slicing images, restoring design mockups' | 'System analysis and architecture design' |
| '1-3 years of experience' | '3-5+ years, architecture capability' |
| 'Complete requirements on time' | 'Technology-driven business development' |
| 'Cross-browser compatibility' | 'TensorFlow/NLP/Knowledge Graphs' |
| 'Responsive layout' | 'AI output verification and quality control' |
It's not that React/Vue are no longer important—they are still foundational. But 'only knowing React/Vue' is no longer enough. The wording in JDs has shifted from 'proficient use' to 'deep understanding of its design principles,' from 'completing development' to 'architecture design.'
The bar hasn't been lowered; it has shifted to a different dimension.
40% of Jobs Will Make AI Skills 'Mandatory' by Year-End
A recent report from a Big 4 consulting firm contains one line: 'By December 2026, 40% of current jobs will require AI skills as mandatory, not preferred.'
Not a 'plus,' but a 'must.'
Translated to frontend hiring, this means:
- In 2025, JDs wrote 'experience with AI tools preferred'
- In the second half of 2026, JDs write 'must have AI-assisted development experience'
ByteDance's campus recruitment data also verifies this: among the 5,000+ positions, AI-related roles account for nearly 50% (Taotian Group data). The number of frontend positions is growing, but the growth is entirely in 'AI + Frontend' hybrid roles.
What Kind of Frontend Engineers Are Still Being Hired
Based on currently open positions (BOSS Zhipin, Lagou data for August 2026), five frontend sub-fields are still growing:
1. AI Application Frontend
The core keyword in new positions at JD, ByteDance, and Alibaba. It's not about training models, but:
- Integrating large model capabilities into frontend products
- Building AI conversation interfaces, real-time streaming rendering
- Handling the uncertainty of AI output (loading states, error fallbacks, retry mechanisms)
// This kind of code was rarely written before, now it's daily routine
const ChatStream: React.FC = () => {
const [chunks, setChunks] = useState<string[]>([]);
useEffect(() => {
const source = new EventSource('/api/chat/stream');
source.onmessage = (e) => {
setChunks(prev => [...prev, e.data]);
};
source.onerror = () => {
// Fallback for AI output interruption — one of the core skills of the new role
source.close();
showFallbackUI();
};
return () => source.close();
}, []);
return <MarkdownRenderer content={chunks.join('')} />;
};
2. Data Visualization / Large Screen Direction
JD Supply Chain and Ant Financial are both hiring. Requirements:
- Three.js / WebGL / D3.js
- Real-time data stream processing
- Complex chart interaction
AI finds this direction hard to replace—because it requires understanding the meaning of data and aesthetic judgment, not just 'writing a page from a design mockup.'
3. Cross-Platform Architect
Not 'writing mini-programs,' but designing cross-platform solutions:
- Architectural decisions for React Native / Flutter
- Trade-off judgments for rendering one codebase on multiple ends
- Locating and optimizing performance bottlenecks (AI currently does this poorly)
4. Frontend Engineering / DevOps
Build tools, CI/CD, monitoring systems from the frontend perspective. Requirements:
- Source-code-level understanding of Webpack/Vite/Turbopack
- Ability to design team-level developer experience
- Performance monitoring and alerting systems
5. Full-Stack AI Engineer (Frontend Background)
A newly emerged title. Not traditional full-stack (Frontend + Node CRUD), but:
- Frontend + AI API integration
- Frontend + Prompt Engineering
- Frontend + Post-processing and display of AI output
'Pure Frontend' Isn't Dead, It's Been Redefined
People who say 'frontend is dead' and people who say 'frontend hasn't changed' are both only half right.
A more accurate description is: the title 'Frontend Development Engineer' is being split into two categories—
| 'Execution-type' Frontend | 'Judgment-type' Frontend | |
|---|---|---|
| Core Work | Write pages per mockups, connect APIs | Architecture design, AI integration, quality control |
| AI Impact | 80%+ replaced | Demand actually increasing (verifying AI output) |
| JD Status | Mass closures/downsizing | Growing (ByteDance +23%) |
| Salary Trend | Stagnant/Declining | BOSS Zhipin: 3+ years architecture type avg salary 32.5K |
| Experience Required | 1-3 years | 3-5 years minimum |
The vast majority of the 160,000 laid off were 'execution-type.' The newly added positions are almost entirely 'judgment-type.'
What to Do Now
It's not about learning TensorFlow tomorrow. But if you are a frontend engineer with less than 3 years of experience, a few actions are worth starting now:
1. Upgrade from 'writing code' to 'reviewing code'
AI can generate code now. Your value is no longer 'being able to write it,' but 'being able to spot the problems.' Deliberately practice Code Review skills—not looking at formatting and naming, but at architectural rationality, edge cases, and performance risks.
2. Do at least one AI-integrated frontend project
Nothing complex. A chat interface, a streaming render, an error fallback for AI output. Being able to talk about practical 'AI frontend' experience in an interview is far more valuable than 'I've used React for three years.'
3. Understand the principles of the AI tools you use
Not about reading papers. But at least know: why a Prompt works better written this way, what a Token is, the limits of the Context Window, why AI hallucinates. This knowledge determines whether you can 'command' AI rather than just 'use' AI.
4. Increase your 'judgment' density
Every day when writing code, ask yourself: What is the trade-off of this solution? If I let AI write it, how do I verify it didn't plant a landmine? Can I spot the core issue in a colleague's PR within 30 seconds?
Judgment ability isn't learned from books; it's built through daily deliberate practice.
Frontend Skill Transition Cheat Sheet
| Current Skill | Transition Direction | Learning Priority | Corresponding Role |
|---|---|---|---|
| React/Vue Development | + AI API Integration + Streaming Rendering | 🔴 Immediate | AI Application Frontend |
| CSS/Responsive Layout | + Three.js/WebGL + Data-Driven | 🟡 Within 6 months | Data Visualization |
| Mini-Program Development | + Cross-Platform Architecture Design + Performance Optimization | 🟡 Within 6 months | Cross-Platform Architect |
| Webpack Configuration | + CI/CD Design + Monitoring System | 🟢 Understand | Frontend Engineering |
| Node.js APIs | + Prompt Engineering + AI Post-Processing | 🔴 Immediate | Full-Stack AI Engineer |
| 'Proficient in using frameworks' | → 'Understanding framework design principles' | 🔴 Immediate | All Senior Roles |
It's Not 'Frontend is Dead,' It's 'The Page-Only Frontend is Dead'
The 165,000 layoffs are not the end—programs.com's tracking is still updating weekly. But ByteDance's +23% is also real.
Two data points being true simultaneously doesn't mean 'frontend is finished'; it means the frontend role is undergoing a filtration. The only criterion for the filter: Among the things you can do, how much is AI unable to do?
If the answer is 'almost nothing'—then your position is indeed in danger. If the answer is 'architectural judgment, quality control, trade-off decisions for complex interactions'—then you are not only safe, but will become more valuable.
The difference isn't whether you use AI, but whether you are AI's operator or AI's referee.