Stack Overflow developer survey 2026 highlights
Stack Overflow developer survey 2026 highlights
The 2025 Stack Overflow Developer Survey, released December 29, 2025 and now driving 2026 hiring conversations, polled 49,000+ developers across 177 countries. The headline numbers: 84% of devs now use AI tools and 51% of pros use them daily, but trust hit an all-time low (46% distrust output, only 3% "highly trust" it). PostgreSQL crossed 55.6%, Cursor debuted at 18%, and Python jumped 7 points year over year.
That is the 60-second version. The longer story matters more, because the survey is the closest thing the dev labor market has to a census, and 2026 budgets, hiring rubrics, and tooling decisions are all being rewritten around it.
What the Stack Overflow developer survey 2026 actually measures
Stack Overflow's annual survey is a self-selected sample, but it is the largest one in the industry. The 2025 edition (the one shaping the 2026 cycle) ran from May to August 2025, asked 62 questions across 314 technologies, and collected responses from 49,000+ developers in 177 countries.
It is imperfect. Self-selection skews toward Stack Overflow users, which still over-indexes on web and back-end devs. But for tracking year-over-year shifts, no other public dataset matches it. Founders pricing comp, agency owners benchmarking rates, and CTOs writing interview rubrics all read this thing the week it drops.
This year's release came late (December instead of the usual mid-year), but that timing makes it the de-facto 2026 reference, which is why everyone is calling it the "2026 survey" in conversation even though the data was collected in 2025.
AI tool adoption: 84% use it, 51% use it daily
The biggest year-over-year story is AI usage solidifying as a baseline, not a novelty.
- 84% of respondents are using or plan to use AI tools, up from 76% in 2024.
- 51% of professional developers use AI tools daily.
- 47.1% of all respondents (pro + learner) use AI tools daily.
- Early-career developers actually lead daily adoption at 55.5%.
The IDE leaderboard
The IDE rankings are where 2025 broke from prior years. VS Code still rules, but Cursor and Claude Code crashed the top tier on their first survey appearance.
| Tool | % using | Notes |
|---|---|---|
| Visual Studio Code | 75.9% | Still the default. Many run Cursor on top. |
| Visual Studio | 29% | .NET shop default, mostly enterprise. |
| Notepad++ | 27.4% | Long tail of Windows ops engineers. |
| IntelliJ IDEA | 27.1% | JVM and Kotlin. |
| Cursor | 18% | First appearance. Top AI-native IDE. |
| Claude Code | 10% | First appearance. Anthropic's CLI agent. |
| Windsurf | 5% | Codeium's agent IDE, first appearance. |
The interesting read is that Cursor and Claude Code are both growing without cannibalizing VS Code; many devs run both. That matches what we see at Cadence, where every engineer is vetted on Cursor / Claude / Copilot fluency before they unlock bookings, and most of them have all three installed.
The LLM leaderboard
Model usage rankings are equally informative for anyone evaluating an engineer's stack:
- OpenAI GPT models: 81.4% usage
- Claude Sonnet: 42.8% usage (45% among pros, 30% among learners)
- Gemini Flash: 35.3% usage
- Most admired: Claude Sonnet at 67.5%, edged only by Gemini Reasoning at 65.2%
Translation: GPT is still the default chat model, but when developers actually like the model they are working with, it skews Anthropic. That maps to Anthropic's pricing power and to why Claude Code grew fast enough to make the IDE list in its first year.
The trust gap: AI usage is up, AI confidence is down
Here is the most counterintuitive finding in the whole survey, and it has direct implications for how you screen engineers in 2026.
| Metric | 2023-24 | 2025 |
|---|---|---|
| Positive sentiment toward AI | 70%+ | 60% |
| Distrust accuracy | ~40% | 46% |
| "Highly trust" output | ~10% | 3% |
Adoption keeps climbing, but enthusiasm is collapsing. The top frustration: 66% of devs cite "the solution is almost right, but not quite" as their main AI gripe. 45% say debugging AI-generated code eats serious time. Combined with the wave of engineering layoffs through 2026, the message to the average developer is unambiguous: get good with AI, or get cut.
Two ceilings are showing up clearly:
- 76% of developers refuse to use AI for deployment or production monitoring.
- 69% refuse to use it for project planning.
The bifurcation matters when you hire. The survey effectively splits developers into two camps: AI-fluent pragmatists (who use it daily, distrust it appropriately, and ship around its weaknesses) and AI-skeptics (who refuse it and miss the productivity floor it now sets). The 2026 hiring playbook is to hire the former and screen out the latter, because every credible study, including the data behind AI changing developer salaries, shows AI-fluent devs commanding a 12% to 56% pay premium for a reason.
Languages and frameworks: Python's 7-point jump and TypeScript's quiet domination
The language ranking moved more in 2025 than it has in five years.
Most-used languages (all respondents):
- JavaScript: 66%
- HTML / CSS: 61.9%
- SQL: 58.6%
- Python: 57.9% (+7pp YoY, the largest single-year jump)
- TypeScript: still rising fast, 43.6% admired
- Rust and Go: +2pp each
Python's surge is the AI story showing up in the data. Every LLM tooling stack, every notebook, every agent framework is Python-first, so devs who used to live in JavaScript or Go are picking up Python as a second language to ship AI features. That single fact reshapes hiring: a "TypeScript engineer" without Python now reads as narrower than the same role did 18 months ago.
Web frameworks (all respondents):
- Node.js: 48.7%
- React: 44.7%
- Next.js: 20.8% (still the dominant React meta-framework)
- Svelte: 7.2%
- Astro: 4.5% (small base, but growing fastest in admired)
For founders building net-new in 2026, the safe-default stack the survey points to is: TypeScript on Next.js with a Postgres back end, deployed on Vercel or Cloudflare, with Python for any AI-heavy service. That maps to roughly 80% of new bookings on Cadence and is the stack we recommend most often when founders ask "what should I build on?"
Databases and cloud: PostgreSQL ate the database market
The biggest infrastructure story in the survey is that PostgreSQL is no longer "one of the popular databases." It is the database.
| Database | 2024 usage | 2025 usage | Shift |
|---|---|---|---|
| PostgreSQL | 48.7% | 55.6% | +7pp |
| MySQL | ~41% | 40.5% | flat |
| SQLite | ~36% | 37.5% | +1.5pp |
| Redis | ~20% | 28% | +8pp |
| MongoDB | ~28% | 24% | -4pp |
Postgres opened a 15-point gap over MySQL (and an 18.6-point gap among pros), and it tops both the most-admired (65%) and most-desired (46%) lists. This is the largest annual expansion in PostgreSQL's history.
MongoDB sliding while Redis surges tells you what shape the new stack is: relational primary store, key-value cache for the hot path, and AI features bolted on through pgvector or Pinecone. The "default to MongoDB" era is over.
Cloud platforms (all respondents):
- AWS: 43.3%
- Azure: 26.3%
- Google Cloud: 24.6%
- Cloudflare: 20.1%
- Vercel: 10.6%
AWS still dominates, but Cloudflare cracking 20% (and Vercel hitting double digits) is the genuinely new pattern. Both are eating workloads at the edge that used to belong to AWS Lambda and CloudFront. If you are evaluating a back-end engineer in 2026, "comfortable on Cloudflare Workers" should be on the rubric, not just AWS.
Work modality and salary: remote held the line
Despite a year of return-to-office press releases, the survey shows remote work did not collapse. It did the opposite: it stabilized.
- 32.4% of developers work fully remote
- 17.9% are fully in-person
- The remaining ~50% are split across hybrid configurations
Remote pay has also closed in on parity. Salary medians keep climbing for AI-skilled roles, and AI engineer global median pay reached roughly $185k. For deeper salary segmentation, our breakdown of the engineering hiring market in 2026 shows where the bands actually sit by region and seniority right now.
The takeaway for founders: remote-first is the working modality of the global dev labor market. Designing your hiring process around an in-person assumption is, statistically, narrowing the pool by roughly 80%.
Reading the survey as a 2026 hiring spec
The most useful way to use this survey is not "interesting numbers" but as a hiring rubric. Every headline stat translates into a screening question.
| Survey finding | Hiring implication |
|---|---|
| 51% pros use AI daily | If a candidate is not AI-fluent, you are below the median. |
| Cursor 18%, Claude Code 10% | Ask which tools they use day-to-day, not "do you use AI". |
| 46% distrust AI accuracy | Hire pragmatists who verify, not zealots who copy-paste. |
| Postgres 55.6% | Default to Postgres skill, not "any SQL". |
| Cloudflare 20.1% | Edge compute is now mainstream, screen for it. |
| Python +7pp | Even non-AI roles benefit from Python literacy. |
| 32.4% remote | Don't filter for in-person without a reason. |
This is the rubric every Cadence engineer is screened against before they unlock bookings. Every engineer on Cadence is AI-native by default, vetted on Cursor, Claude, and Copilot fluency in a live voice interview, with a real bias toward pragmatists who treat AI output as a draft, not a deliverable. Our pool is roughly 12,800 engineers worldwide, and the trial-to-active conversion rate sits around 67%, which is a useful sanity check on how often a vetted engineer actually fits the spec.
If you want a faster benchmark of what a 2026 engineering team should cost against these survey medians, run the numbers on Cadence's ROI calculator. It uses the same survey data plus our own pool to translate "I need 2 senior engineers for 12 weeks" into a real number.
What to do with this survey
If you are hiring or budgeting in 2026, four concrete next steps:
- Update your interview rubric to include AI-tool fluency as a baseline check, not a bonus. Ask which IDE they use, which model, and how they verify output.
- Standardize your team's tooling. Cursor or Claude Code, plus Copilot, plus a defined model preference. Mixed loadouts cost shipping speed.
- Rebuild your salary band against 2026 medians. The numbers above plus regional data (Latin America or Eastern Europe) will keep your offers competitive without overpaying.
- Decide whether you build a team or book on demand. For projects under 12 months, weekly booking on Cadence at $500 (junior), $1,000 (mid), $1,500 (senior), or $2,000 (lead) usually beats hiring full-time. For 5-year strategic capability, headcount still wins. The survey just lowered the bar for what a competent week of work should look like.
Want to skip the rubric work? Every engineer on Cadence is already vetted against this survey: AI-fluent on Cursor, Claude, and Copilot, Postgres-first by default, and matched in 2 minutes. Run the numbers on your next role or start a 48-hour trial at no cost.
FAQ
When was the 2025 Stack Overflow Developer Survey released?
Stack Overflow published the 2025 survey on December 29, 2025, with results from data collected May to August 2025. It captures 49,000+ developers across 177 countries and is now the de-facto reference for 2026 hiring decisions.
What is the most-used AI coding tool in the 2025 Stack Overflow survey?
ChatGPT leads at 82% adoption among developers using AI tools, followed by GitHub Copilot at 68%. Cursor (18%) and Claude Code (10%) made their first survey appearance and rank among the most-used AI-enabled IDEs.
What percentage of developers use AI tools daily?
51% of professional developers use AI tools daily, with 84% using or planning to use them overall. Trust, however, declined: 46% distrust output accuracy vs 33% who trust it, and only 3% report "highly trusting" the output.
What is the most popular database in 2026?
PostgreSQL is the most-used database at 55.6%, up nearly 7 percentage points from 2024. It opened a 15-point gap over MySQL (40.5%) and is also the most admired (65%) and most desired (46%) database.
Did remote work decline in the 2025 Stack Overflow survey?
No. 32.4% of developers work fully remote and only 17.9% are fully in-person; the rest are hybrid. The remote pay penalty also continued to narrow toward parity with onsite roles.