Farewell Stack Overflow

You know what they say, "a time-series chart is worth a thousand words":

The above chart tells a clear story about Stack Overflow's popularity. The site-wide monthly Q+A volume peaked in March 2014 at 664,154 posts (questions + answers, including deleted). The latest complete month in this dataset (February 2026) is 11,280 posts.

That is a 98.3% drop from peak monthly activity.

Stack Overflow's decline appears to have started around the COVID-19 period and accelerated further with the arrival of LLMs.

The chart above is based on this Data StackExchange query and CSV export:

• Query: Stack Overflow monthly post counts: Questions vs Answers (including deleted)
• CSV: QueryResults.csv (our copy)

For completeness, this is the exact query used:

SELECT
    DATEFROMPARTS(YEAR(CreationDate), MONTH(CreationDate), 1) AS Month,
    SUM(CASE WHEN PostTypeId = 1 THEN 1 ELSE 0 END) AS Questions,
    SUM(CASE WHEN PostTypeId = 2 THEN 1 ELSE 0 END) AS Answers
FROM PostsWithDeleted
WHERE PostTypeId IN (1,2)
GROUP BY
    YEAR(CreationDate),
    MONTH(CreationDate)
ORDER BY
    Month

My personal Stack Overflow journey

Click here to view my profile on Stack Overflow

For years, Stack Overflow was part of my daily engineering loop. I asked questions, provided answers, and got a lot back from the community.

For a while, I was in the top 1% contributors by reputation. I proudly put the 18k+ reputation milestone on LinkedIn and in my own bio here on devillers.nl. I still consider that an achievement worth mentioning.

Over the years, engineers in my network would message me that they had stumbled upon one of my answers and it fixed their issue.

I even stumbled upon my own answer once or twice.

Some of the more popular answers became blog posts on this site, for example:

• Improving DbEntityValidationException
• Mockito, @BeforeEach and UnnecessaryStubbingException
• Nodejs, TypeScript, ts-node-dev & top-level await

Let's crunch some more data.

Personal stats

Snapshot as of March 9, 2026:

Focusing on my tenure and daily activity:

• Joined period analyzed: 2011-01-09 to 2026-03-09 (15 years, 2 months)
• Total calendar days in that window: 5539
• Business days (Mon-Fri) in that window: 3956
• Days visited: 3008

That means:

• I visited Stack Overflow on 54.3% of all days in that span.
• I visited Stack Overflow on 76.0% of business days in that span.

In other words, over the last 15 years of my life I visited Stack Overflow every other day.

I don't what that means, but it must mean something, right? 😬

Let's go a little deeper

Stack Overflow has many hidden/niche features. One of them is the "visit history", which is this little jQuery DatePicker that pops up when you click the "Visited x days" on your personal profile. Embedded within this widget is a full history of every single day you visited Stack Overflow. To grab this data, you take the following URL https://stackoverflow.com/users/daily-site-access/546967, substitute the user ID for your own and fetch the raw HTML. On line 3 you will see a variable visited with your visit history in the following format:

var visited = {2010:{12:{18:1,19:1,24:1}},2011:{1:{17:1,20:1},2:{12:1},3:{13:1,15:1,16:1,17:1,28:

Leave it to the SO devs to use a sparse three-dimensional map keyed by (year, month, day) as a data structure for space-efficient storage and fast lookups.

This data structure encodes dates in a Year → Month → Day hierarchy.

visited
 └── Year (2010, 2011)
      └── Month (1–12)
           └── Day (1–31)
                └── Value (1)

Example lookup:

visited[2010][12][18] === 1

Which means that on December 18, 2010 I visited Stack Overflow.

Armed with this full history, I can visualize my visits using a neat per-month heatmap:

End of an era

Like many engineers, I used Stack Overflow less and less.

Part of that is positive: tooling and online resources improved, open-source ecosystems matured, and increasingly I could just read the source code directly.

Then LLMs arrived. For day-to-day implementation work, they became my go-to source for typical engineering Q&A and much more. That further reduced how often I used Stack Overflow.

I have immense respect for what Stack Overflow gave the industry, and I feel privileged to have been part of it. A lot of that knowledge, including some of my own answers, now also shows up through LLMs.

If you open your favorite AI tool and ask, "How can I see the validation errors from DbEntityValidationException?", it will usually return my approach, sometimes even including my sassy remarks that its default behavior is frustrating.

Stack Overflow gave LLMs a massive high-quality training set with 18 years of collective engineering knowledge. The open question now is how we keep building shared, durable technical knowledge as usage shifts to LLMs.

Farewell, Stack Overflow.

Thanks for the years.