Tech postings up 20%, tech employment flat

Two US data releases about tech work came out last week. One is Dice's October tech jobs report: tech job postings up 20% from a year earlier, and AI and machine learning postings up 91%. The other is the Bureau of Labor Statistics' October Employment Situation, released November 6: employment in "computer systems design and related services" was down 1.1% from a year earlier.

One says openings are rising, the other says headcount is not. Both can be true, but only if you know what each one measures. This post puts them side by side, then looks at the part of the Dice report I find most interesting: employers are starting to name specific AI coding tools in job postings, and right next to them, technical debt, maintainability and design reviews. A caveat up front: both are US data, and Taiwan's market will not necessarily move the same way.

1. What happened

Dice: October tech jobs report (released in November)

• The data comes from Lightcast's job postings database. Dice pulled it on November 3 and analyzed more than 7 million US tech postings.
• Tech postings rose 4% month over month (after 9% in September) and 20% year over year. AI/ML postings grew 91% year over year, more than four times the overall rate.
• By industry, tech postings in insurance and software rose more than 30% month over month; year over year, software, aerospace/defense and manufacturing were all at or above 50%.
• By region, the New York metro area rose 48% month over month and Chicago 46%, while San Jose fell 12% for the month and 4% for the year.
• Job titles with month-over-month growth above 150% cluster in three groups: training and running AI models (AI Trainer, ML Data Engineer, MLOps Engineer), senior technical roles (Lead, Staff, Distinguished Engineer), and core enterprise systems (mainframe, SQL DBA, Workday, Salesforce).
• The list of skills up 200% or more year over year includes Claude AI and GitHub Copilot, alongside Technical Debt, Maintainability, Design Reviews, Software Technical Review and Defect Management. Dice reads this as employers wanting engineers who can review and maintain AI-assisted code.
• The same list includes Agentic AI, AI Agents, RAG and Responsible AI, plus production skills such as Telemetry, Traceability and Exception Handling.

BLS: October Employment Situation (released Nov 6)

• Nonfarm payrolls rose by 165,000 and the unemployment rate was 4.1%. August and September were revised up by a combined 84,000. Professional and business services added 70,000 jobs in October.
• Seasonally adjusted figures from the industry employment table (Table B-1), compared with October 2025 (September and October are preliminary), in thousands:

Industry	Oct 2025	Oct 2026	Year over year	October change
Information	2,853	2,734	about −4.2%	−1
Computer systems design and related services	2,379.7	2,354.1	about −1.1%	+3.2
Computing infrastructure, data processing, web hosting	480.7	454.8	about −5.4%	0.0
Publishing industries (includes software publishers)	903.5	880.0	about −2.6%	−4.1
Management, scientific and technical consulting	1,860.0	1,905.7	about +2.5%	+17.3
Semiconductor and other electronic components	369.5	376.8	about +2.0%	+3.3

The year-over-year changes are my own calculations from the table.

2. What was not said, or is still open

• A posting is not a hire. Dice states that its report is based entirely on real-time posting data, is independent of BLS employment statistics, and that prior months may be revised. A posting can stay up for months, be reposted, or never be filled, and Dice does not report fill rates.
• The categories differ. Dice's "software industry" is the industry of the employer posting the job. The BLS rows are headcounts by NAICS industry. You cannot subtract one from the other.
• BLS figures for September and October are preliminary. The next release is December 4 and will revise them.
• Dice does not say what share of postings mention Claude AI or GitHub Copilot, only that they grew more than 200%. For a skill starting from a small base, 200% growth can mean going from very few to few.
• Comparable Taiwan data for the same period, at the level of engineering roles, is not something I found this time, so I am not drawing a parallel.

3. My take

"Postings are recovering" and "employment isn't growing" can both be true. If postings rise while headcount in tech-related industries stays flat or falls, the more likely explanation is that companies are swapping people and changing the mix of roles rather than simply expanding. That matches the Indeed data I covered in engineering jobs in the AI era: the total has not come back, the composition is shifting, entry-level share is shrinking and senior share is growing. Among Dice's titles growing more than 150% this month, Lead, Staff and Distinguished titles stand out, which points the same way. This is my inference; neither dataset proves it directly.

Employers are naming tools, and naming the finishing work too. Postings used to say "familiar with AI tools." Now specific tools appear. More telling is what sits next to them: technical debt, maintainability, design reviews, defect management. That echoes the verification gap in 81% of teams have merged AI code they didn't fully understand: once producing code gets cheap, checking it matters more. Employers are not just buying "can write with AI." They are buying "can judge whether what AI wrote is fit to ship, and keep it running afterward."

The steadiest demand is not always where the hype is. Mainframe, SQL DBA, Workday and Salesforce titles are on the list too; Dice's explanation is that AI and data projects eventually have to connect back to core business systems. Consulting employment is up 2.5% year over year and semiconductor manufacturing about 2%, and neither is what people usually mean by "tech companies." For job seekers, the map of tech jobs is more spread out than it used to be.

4. Practical advice

Engineers job hunting or changing careers

• Show evidence of review and maintenance on your resume. How you review AI-generated code, what technical debt you have paid down, what tests or monitoring you have set up. That says more than "proficient with Copilot."
• Use the tools, but lead with judgment. When postings name Claude and Copilot, using them is the baseline. Being able to explain when you rejected an AI suggestion, and why, is the differentiator.
• Learn what happens after launch. Telemetry, Traceability and Exception Handling are on the same list. Beyond building features, knowing how to monitor, trace and debug is the part junior engineers prepare for least.
• Treat domain knowledge as leverage. Tech postings in insurance, finance and manufacturing are growing. Understanding the business process plus engineering skills is easier to place than coding alone.

Engineering managers and hiring leads

• If a job description names AI tools, put "how do you review AI output" into the interview too, or you will only filter for people who generate a lot of code quickly.
• Add a hands-on technical debt or code review exercise. It is closer to what the team does every day than another algorithm question.
• If you are cutting entry-level roles, think about where your senior engineers will come from in three years.

Recent graduates without work experience

• When contributing to open source, pick maintenance work such as fixing bugs, adding tests or cleaning up docs. It maps directly onto the skills employers are naming.
• A portfolio project where you took over someone else's code and improved it shows these skills better than another new to-do app.

5. What it means for job seekers

Put together, what I read in these two reports is this: the market has not closed, but what it pays for is changing. More and more people can produce code quickly. There are still not enough people who can check it, maintain it, and connect AI to real systems.

For engineers in Taiwan, this is US data and does not translate directly into local salaries or posting counts. But if you are applying to foreign companies or remote roles, or your company is rolling out AI coding tools, this skills list is worth holding up against your resume. The finding in the Stack Overflow 2026 Developer Survey that 48% of developers trust AI only when the answer is easy to verify makes the same point: you have to understand the work to know whether the AI got it right.

---

Posting numbers will keep bouncing around, but the ability to judge and maintain code is unlikely to go out of style. If you are working out which skills to build next, or want your team's hiring criteria to keep up with AI coding tools, book a 30-minute intro call.