diff --git a/src/blog/posts/2024/5/learning-go-day-12.md b/src/blog/posts/2024/5/learning-go-day-12.md
new file mode 100644
index 0000000..1b38261
--- /dev/null
+++ b/src/blog/posts/2024/5/learning-go-day-12.md
@@ -0,0 +1,99 @@
+---
+title: "Learning Go: Day Twelve"
+date: 2024-05-14T08:00:00.0Z
+tags:
+    - learning
+    - go
+excerpt: "A quick one today on improving the data that's output"
+---
+
+Yesterday I [actually started storing and printing responses](/post/learning-go-day-eleven) from the database. Today as a short exercice, I'm going to improve that data and group it by site. 
+
+## The data model
+
+Now I've introduced a new data model, `ping.SiteResponse`. Originally this was going to live inside the `sites` package, but this introduced a circular dependency. I'll need to resolve that soon regardless, because this data model sucks, but for now this code will live in the `ping` package.
+
+```go
+// ping/ping.go
+type SiteResponse struct {
+	created_at uint64
+	Name       string
+	Url        string
+	Pings      []Ping
+}
+```
+
+## Retrieving the data
+
+My first instinct was to use an aggregate function to retrieve a site along with all of it's pings, however SQLite only has [a small subset of aggregate functions](https://www.sqlite.org/lang_aggfunc.html). There are some workarounds, but for this part I'm just going to do it as two separate queries, even if it is a bit inefficient[^1].
+
+So, I add a `List` function to the `sites` package, which gets all of the site records:
+
+```go
+// sites/sites.go
+
+func List(db *sql.DB) []Site {
+	rows, err := db.Query(`SELECT url, name, created_at FROM sites ORDER BY created_at DESC`)
+
+	if err != nil {
+		panic(err)
+	}
+	defer rows.Close()
+
+	sites := make([]Site, 0)
+
+	for rows.Next() {
+		s := Site{}
+		rows.Scan(&s.Url, &s.Name, &s.created_at)
+		sites = append(sites, s)
+	}
+
+	return sites
+}
+```
+
+And then in `ping.ListGroupedBySite`, I iterate over this list and retrieve all of the pings for each site:
+
+```go
+// ping/ping.go
+func ListGroupedBySite(db *sql.DB) []SiteResponse {
+
+	responses := make([]SiteResponse, 0)
+
+	for _, site := range sites.List(db) {
+		s := SiteResponse{Url: site.Url, Name: site.Name, created_at: site.Created_at}
+		s.Pings = listForSite(db, site)
+		responses = append(responses, s)
+	}
+
+	return responses
+}
+
+func listForSite(db *sql.DB, site sites.Site) []Ping {
+	rows, err := db.Query(`SELECT ping.timestamp, ping.status FROM ping WHERE site = ? ORDER BY timestamp DESC`, site.Url)
+
+	if err != nil {
+		panic(err)
+	}
+	defer rows.Close()
+
+	pings := make([]Ping, 0)
+
+	for rows.Next() {
+		p := Ping{}
+		rows.Scan(&p.Timestamp, &p.Status)
+		p.Site = site
+		pings = append(pings, p)
+	}
+
+	return pings
+}
+```
+
+I realised I had to make a couple of changes to the `Site` struct to make this work. Namely `created_at` needed to be a string, not an int, as SQLite stores the timestamps as strings[^2]. I also had to make it public so I could copy the value to the `SiteResponse` struct[^3].
+
+But that works, now when accessing https://oopsie.lewisdale.dev, there's an array of sites, with a timestamped list of `pings` attached.
+
+[^1]: I've got like 12 database records. I'd spend more time writing a query and code than it would save me right now.
+[^2]: I could also have used [datetime functions](https://www.sqlite.org/lang_datefunc.html) to convert it to a unix timestamp, but what's the point?
+[^3]: Really this should be the other way round and `Site` has a mapping function, as `SiteResponse` is the publicly-accessible type
\ No newline at end of file