diff --git a/calibrate.go b/calibrate.go
index f858bcd..5e94713 100644
--- a/calibrate.go
+++ b/calibrate.go
@@ -2,9 +2,9 @@ package i18n
 
 import (
 	"context"
-	"fmt"
 	"time"
 
+	"dappco.re/go/core"
 	log "dappco.re/go/core/log"
 	"forge.lthn.ai/core/go-inference"
 )
@@ -66,7 +66,7 @@ func CalibrateDomains(ctx context.Context, modelA, modelB inference.TextModel,
 	// Build classification prompts from sample texts.
 	prompts := make([]string, len(samples))
 	for i, s := range samples {
-		prompts[i] = fmt.Sprintf(cfg.promptTemplate, s.Text)
+		prompts[i] = core.Sprintf(cfg.promptTemplate, s.Text)
 	}
 
 	// Classify with model A.
@@ -93,7 +93,7 @@ func CalibrateDomains(ctx context.Context, modelA, modelB inference.TextModel,
 		if agree {
 			stats.Agreed++
 		} else {
-			key := fmt.Sprintf("%s->%s", a, b)
+			key := core.Sprintf("%s->%s", a, b)
 			stats.ConfusionPairs[key]++
 		}
 		stats.ByDomainA[a]++
@@ -140,7 +140,7 @@ func classifyAll(ctx context.Context, model inference.TextModel, prompts []strin
 
 		results, err := model.Classify(ctx, batch, inference.WithMaxTokens(1))
 		if err != nil {
-			return nil, 0, log.E("classifyAll", fmt.Sprintf("classify batch [%d:%d]", i, end), err)
+			return nil, 0, log.E("classifyAll", core.Sprintf("classify batch [%d:%d]", i, end), err)
 		}
 
 		for j, r := range results {
diff --git a/classify.go b/classify.go
index b6395fa..4dc8dbe 100644
--- a/classify.go
+++ b/classify.go
@@ -3,12 +3,10 @@ package i18n
 import (
 	"bufio"
 	"context"
-	"encoding/json"
-	"fmt"
 	"io"
-	"strings"
 	"time"
 
+	"dappco.re/go/core"
 	log "dappco.re/go/core/log"
 	"forge.lthn.ai/core/go-inference"
 )
@@ -63,7 +61,7 @@ func mapTokenToDomain(token string) string {
 	if len(token) == 0 {
 		return "unknown"
 	}
-	lower := strings.ToLower(token)
+	lower := core.Lower(token)
 	switch {
 	case lower == "technical" || lower == "tech":
 		return "technical"
@@ -107,7 +105,7 @@ func ClassifyCorpus(ctx context.Context, model inference.TextModel,
 		}
 		prompts := make([]string, len(batch))
 		for i, p := range batch {
-			prompts[i] = fmt.Sprintf(cfg.promptTemplate, p.prompt)
+			prompts[i] = core.Sprintf(cfg.promptTemplate, p.prompt)
 		}
 		results, err := model.Classify(ctx, prompts, inference.WithMaxTokens(1))
 		if err != nil {
@@ -119,13 +117,12 @@ func ClassifyCorpus(ctx context.Context, model inference.TextModel,
 			stats.ByDomain[domain]++
 			stats.Total++
 
-			line, err := json.Marshal(batch[i].record)
-			if err != nil {
-				return log.E("ClassifyCorpus", "marshal output", err)
-			}
-			if _, err := fmt.Fprintf(output, "%s\n", line); err != nil {
-				return log.E("ClassifyCorpus", "write output", err)
+			mr := core.JSONMarshal(batch[i].record)
+			if !mr.OK {
+				return log.E("ClassifyCorpus", "marshal output", mr.Value.(error))
 			}
+			line := mr.Value.([]byte)
+			core.Print(output, "%s", line)
 		}
 		batch = batch[:0]
 		return nil
@@ -133,7 +130,7 @@ func ClassifyCorpus(ctx context.Context, model inference.TextModel,
 
 	for scanner.Scan() {
 		var record map[string]any
-		if err := json.Unmarshal(scanner.Bytes(), &record); err != nil {
+		if r := core.JSONUnmarshal(scanner.Bytes(), &record); !r.OK {
 			stats.Skipped++
 			continue
 		}
diff --git a/classify_test.go b/classify_test.go
index 08e91c7..5858760 100644
--- a/classify_test.go
+++ b/classify_test.go
@@ -3,11 +3,10 @@ package i18n
 import (
 	"bytes"
 	"context"
-	"encoding/json"
 	"iter"
-	"strings"
 	"testing"
 
+	"dappco.re/go/core"
 	"forge.lthn.ai/core/go-inference"
 )
 
@@ -85,7 +84,7 @@ func TestClassifyCorpus_Basic(t *testing.T) {
 		},
 	}
 
-	input := strings.NewReader(
+	input := core.NewReader(
 		`{"seed_id":"1","domain":"general","prompt":"Delete the file"}` + "\n" +
 			`{"seed_id":"2","domain":"science","prompt":"Explain gravity"}` + "\n",
 	)
@@ -102,15 +101,15 @@ func TestClassifyCorpus_Basic(t *testing.T) {
 		t.Errorf("Skipped = %d, want 0", stats.Skipped)
 	}
 
-	lines := strings.Split(strings.TrimSpace(output.String()), "\n")
+	lines := core.Split(core.Trim(output.String()), "\n")
 	if len(lines) != 2 {
 		t.Fatalf("output lines = %d, want 2", len(lines))
 	}
 
 	for i, line := range lines {
 		var record map[string]any
-		if err := json.Unmarshal([]byte(line), &record); err != nil {
-			t.Fatalf("line %d: unmarshal: %v", i, err)
+		if r := core.JSONUnmarshal([]byte(line), &record); !r.OK {
+			t.Fatalf("line %d: unmarshal: %v", i, r.Value)
 		}
 		if record["domain_1b"] != "technical" {
 			t.Errorf("line %d: domain_1b = %v, want %q", i, record["domain_1b"], "technical")
@@ -133,7 +132,7 @@ func TestClassifyCorpus_SkipsMalformed(t *testing.T) {
 		},
 	}
 
-	input := strings.NewReader(
+	input := core.NewReader(
 		"not valid json\n" +
 			`{"seed_id":"1","domain":"general","prompt":"Hello world"}` + "\n" +
 			`{"seed_id":"2","domain":"general"}` + "\n",
@@ -157,7 +156,7 @@ func TestClassifyCorpus_DomainMapping(t *testing.T) {
 		classifyFunc: func(_ context.Context, prompts []string, _ ...inference.GenerateOption) ([]inference.ClassifyResult, error) {
 			results := make([]inference.ClassifyResult, len(prompts))
 			for i, p := range prompts {
-				if strings.Contains(p, "Delete") {
+				if core.Contains(p, "Delete") {
 					results[i] = inference.ClassifyResult{Token: inference.Token{Text: "technical"}}
 				} else {
 					results[i] = inference.ClassifyResult{Token: inference.Token{Text: "ethical"}}
@@ -167,7 +166,7 @@ func TestClassifyCorpus_DomainMapping(t *testing.T) {
 		},
 	}
 
-	input := strings.NewReader(
+	input := core.NewReader(
 		`{"prompt":"Delete the file now"}` + "\n" +
 			`{"prompt":"Is it right to lie?"}` + "\n",
 	)
diff --git a/compose_test.go b/compose_test.go
index 275fef9..e4cd0e7 100644
--- a/compose_test.go
+++ b/compose_test.go
@@ -1,9 +1,9 @@
 package i18n
 
 import (
-	"fmt"
 	"testing"
 
+	"dappco.re/go/core"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -123,7 +123,7 @@ func TestSubject_String_Good(t *testing.T) {
 
 func TestSubject_String_Good_Stringer(t *testing.T) {
 	// Use a type that implements fmt.Stringer
-	subj := S("error", fmt.Errorf("something broke"))
+	subj := S("error", core.NewError("something broke"))
 	assert.Equal(t, "something broke", subj.String())
 }
 
diff --git a/core_service.go b/core_service.go
index 30ae2f9..e0aa993 100644
--- a/core_service.go
+++ b/core_service.go
@@ -4,7 +4,6 @@ package i18n
 
 import (
 	"context"
-	"fmt"
 	"io/fs"
 	"sync"
 
@@ -59,7 +58,7 @@ func NewCoreService(opts ServiceOptions) func(*core.Core) (any, error) {
 
 		if opts.Language != "" {
 			if langErr := svc.SetLanguage(opts.Language); langErr != nil {
-				return nil, fmt.Errorf("i18n: invalid language %q: %w", opts.Language, langErr)
+				return nil, core.Wrap(langErr, "NewCoreService", core.Sprintf("i18n: invalid language %q", opts.Language))
 			}
 		}
 
diff --git a/go.mod b/go.mod
index 66c487d..da264cf 100644
--- a/go.mod
+++ b/go.mod
@@ -5,15 +5,12 @@ go 1.26.0
 require golang.org/x/text v0.35.0
 
 require (
-	dappco.re/go/core v0.5.0
+	dappco.re/go/core v0.8.0-alpha.1
 	dappco.re/go/core/log v0.1.0
 	forge.lthn.ai/core/go-inference v0.1.4
 )
 
-require (
-	forge.lthn.ai/core/go-log v0.0.4 // indirect
-	github.com/kr/text v0.2.0 // indirect
-)
+require github.com/kr/text v0.2.0 // indirect
 
 require (
 	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
diff --git a/go.sum b/go.sum
index 4cf7d1f..bc16ba9 100644
--- a/go.sum
+++ b/go.sum
@@ -1,5 +1,5 @@
-dappco.re/go/core v0.5.0 h1:P5DJoaCiK5Q+af5UiTdWqUIW4W4qYKzpgGK50thm21U=
-dappco.re/go/core v0.5.0/go.mod h1:f2/tBZ3+3IqDrg2F5F598llv0nmb/4gJVCFzM5geE4A=
+dappco.re/go/core v0.8.0-alpha.1 h1:gj7+Scv+L63Z7wMxbJYHhaRFkHJo2u4MMPuUSv/Dhtk=
+dappco.re/go/core v0.8.0-alpha.1/go.mod h1:f2/tBZ3+3IqDrg2F5F598llv0nmb/4gJVCFzM5geE4A=
 dappco.re/go/core/log v0.1.0 h1:pa71Vq2TD2aoEUQWFKwNcaJ3GBY8HbaNGqtE688Unyc=
 dappco.re/go/core/log v0.1.0/go.mod h1:Nkqb8gsXhZAO8VLpx7B8i1iAmohhzqA20b9Zr8VUcJs=
 forge.lthn.ai/core/go-inference v0.1.4 h1:fuAgWbqsEDajHniqAKyvHYbRcBrkGEiGSqR2pfTMRY0=
diff --git a/grammar.go b/grammar.go
index 3a97674..3465375 100644
--- a/grammar.go
+++ b/grammar.go
@@ -2,9 +2,10 @@ package i18n
 
 import (
 	"maps"
-	"strings"
 	"text/template"
 	"unicode"
+
+	"dappco.re/go/core"
 )
 
 // GetGrammarData returns the grammar data for the specified language.
@@ -55,7 +56,7 @@ func getVerbForm(lang, verb, form string) string {
 	if data == nil || data.Verbs == nil {
 		return ""
 	}
-	verb = strings.ToLower(verb)
+	verb = core.Lower(verb)
 	if forms, ok := data.Verbs[verb]; ok {
 		switch form {
 		case "past":
@@ -72,7 +73,7 @@ func getWord(lang, word string) string {
 	if data == nil || data.Words == nil {
 		return ""
 	}
-	return data.Words[strings.ToLower(word)]
+	return data.Words[core.Lower(word)]
 }
 
 func getPunct(lang, rule, defaultVal string) string {
@@ -98,7 +99,7 @@ func getNounForm(lang, noun, form string) string {
 	if data == nil || data.Nouns == nil {
 		return ""
 	}
-	noun = strings.ToLower(noun)
+	noun = core.Lower(noun)
 	if forms, ok := data.Nouns[noun]; ok {
 		switch form {
 		case "one":
@@ -126,7 +127,7 @@ func currentLangForGrammar() string {
 //	PastTense("run")    // "ran"
 //	PastTense("copy")   // "copied"
 func PastTense(verb string) string {
-	verb = strings.ToLower(strings.TrimSpace(verb))
+	verb = core.Lower(core.Trim(verb))
 	if verb == "" {
 		return ""
 	}
@@ -140,16 +141,16 @@ func PastTense(verb string) string {
 }
 
 func applyRegularPastTense(verb string) string {
-	if strings.HasSuffix(verb, "ed") && len(verb) > 2 {
+	if core.HasSuffix(verb, "ed") && len(verb) > 2 {
 		thirdFromEnd := verb[len(verb)-3]
 		if !isVowel(rune(thirdFromEnd)) && thirdFromEnd != 'e' {
 			return verb
 		}
 	}
-	if strings.HasSuffix(verb, "e") {
+	if core.HasSuffix(verb, "e") {
 		return verb + "d"
 	}
-	if strings.HasSuffix(verb, "y") && len(verb) > 1 {
+	if core.HasSuffix(verb, "y") && len(verb) > 1 {
 		prev := rune(verb[len(verb)-2])
 		if !isVowel(prev) {
 			return verb[:len(verb)-1] + "ied"
@@ -189,7 +190,7 @@ func shouldDoubleConsonant(verb string) bool {
 //	Gerund("run")     // "running"
 //	Gerund("die")     // "dying"
 func Gerund(verb string) string {
-	verb = strings.ToLower(strings.TrimSpace(verb))
+	verb = core.Lower(core.Trim(verb))
 	if verb == "" {
 		return ""
 	}
@@ -203,10 +204,10 @@ func Gerund(verb string) string {
 }
 
 func applyRegularGerund(verb string) string {
-	if strings.HasSuffix(verb, "ie") {
+	if core.HasSuffix(verb, "ie") {
 		return verb[:len(verb)-2] + "ying"
 	}
-	if strings.HasSuffix(verb, "e") && len(verb) > 1 {
+	if core.HasSuffix(verb, "e") && len(verb) > 1 {
 		secondLast := rune(verb[len(verb)-2])
 		if secondLast != 'e' && secondLast != 'y' && secondLast != 'o' {
 			return verb[:len(verb)-1] + "ing"
@@ -232,20 +233,20 @@ func Pluralize(noun string, count int) string {
 
 // PluralForm returns the plural form of a noun.
 func PluralForm(noun string) string {
-	noun = strings.TrimSpace(noun)
+	noun = core.Trim(noun)
 	if noun == "" {
 		return ""
 	}
-	lower := strings.ToLower(noun)
+	lower := core.Lower(noun)
 	if form := getNounForm(currentLangForGrammar(), lower, "other"); form != "" {
 		if unicode.IsUpper(rune(noun[0])) && len(form) > 0 {
-			return strings.ToUpper(string(form[0])) + form[1:]
+			return core.Upper(string(form[0])) + form[1:]
 		}
 		return form
 	}
 	if plural, ok := irregularNouns[lower]; ok {
 		if unicode.IsUpper(rune(noun[0])) {
-			return strings.ToUpper(string(plural[0])) + plural[1:]
+			return core.Upper(string(plural[0])) + plural[1:]
 		}
 		return plural
 	}
@@ -253,28 +254,28 @@ func PluralForm(noun string) string {
 }
 
 func applyRegularPlural(noun string) string {
-	lower := strings.ToLower(noun)
-	if strings.HasSuffix(lower, "s") ||
-		strings.HasSuffix(lower, "ss") ||
-		strings.HasSuffix(lower, "sh") ||
-		strings.HasSuffix(lower, "ch") ||
-		strings.HasSuffix(lower, "x") ||
-		strings.HasSuffix(lower, "z") {
+	lower := core.Lower(noun)
+	if core.HasSuffix(lower, "s") ||
+		core.HasSuffix(lower, "ss") ||
+		core.HasSuffix(lower, "sh") ||
+		core.HasSuffix(lower, "ch") ||
+		core.HasSuffix(lower, "x") ||
+		core.HasSuffix(lower, "z") {
 		return noun + "es"
 	}
-	if strings.HasSuffix(lower, "y") && len(noun) > 1 {
+	if core.HasSuffix(lower, "y") && len(noun) > 1 {
 		prev := rune(lower[len(lower)-2])
 		if !isVowel(prev) {
 			return noun[:len(noun)-1] + "ies"
 		}
 	}
-	if strings.HasSuffix(lower, "f") {
+	if core.HasSuffix(lower, "f") {
 		return noun[:len(noun)-1] + "ves"
 	}
-	if strings.HasSuffix(lower, "fe") {
+	if core.HasSuffix(lower, "fe") {
 		return noun[:len(noun)-2] + "ves"
 	}
-	if strings.HasSuffix(lower, "o") && len(noun) > 1 {
+	if core.HasSuffix(lower, "o") && len(noun) > 1 {
 		prev := rune(lower[len(lower)-2])
 		if !isVowel(prev) {
 			if lower == "hero" || lower == "potato" || lower == "tomato" || lower == "echo" || lower == "veto" {
@@ -295,14 +296,14 @@ func Article(word string) string {
 	if word == "" {
 		return ""
 	}
-	lower := strings.ToLower(strings.TrimSpace(word))
+	lower := core.Lower(core.Trim(word))
 	for key := range consonantSounds {
-		if strings.HasPrefix(lower, key) {
+		if core.HasPrefix(lower, key) {
 			return "a"
 		}
 	}
 	for key := range vowelSounds {
-		if strings.HasPrefix(lower, key) {
+		if core.HasPrefix(lower, key) {
 			return "an"
 		}
 	}
@@ -322,7 +323,7 @@ func isVowel(r rune) bool {
 
 // Title capitalises the first letter of each word.
 func Title(s string) string {
-	var b strings.Builder
+	b := core.NewBuilder()
 	b.Grow(len(s))
 	prev := ' '
 	for _, r := range s {
@@ -345,8 +346,8 @@ func Quote(s string) string {
 func TemplateFuncs() template.FuncMap {
 	return template.FuncMap{
 		"title":      Title,
-		"lower":      strings.ToLower,
-		"upper":      strings.ToUpper,
+		"lower":      core.Lower,
+		"upper":      core.Upper,
 		"past":       PastTense,
 		"gerund":     Gerund,
 		"plural":     Pluralize,
diff --git a/handler.go b/handler.go
index 962d3f2..236a289 100644
--- a/handler.go
+++ b/handler.go
@@ -1,19 +1,18 @@
 package i18n
 
 import (
-	"fmt"
-	"strings"
+	"dappco.re/go/core"
 )
 
 // LabelHandler handles i18n.label.{word} -> "Status:" patterns.
 type LabelHandler struct{}
 
 func (h LabelHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.label.")
+	return core.HasPrefix(key, "i18n.label.")
 }
 
 func (h LabelHandler) Handle(key string, args []any, next func() string) string {
-	word := strings.TrimPrefix(key, "i18n.label.")
+	word := core.TrimPrefix(key, "i18n.label.")
 	return Label(word)
 }
 
@@ -21,11 +20,11 @@ func (h LabelHandler) Handle(key string, args []any, next func() string) string
 type ProgressHandler struct{}
 
 func (h ProgressHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.progress.")
+	return core.HasPrefix(key, "i18n.progress.")
 }
 
 func (h ProgressHandler) Handle(key string, args []any, next func() string) string {
-	verb := strings.TrimPrefix(key, "i18n.progress.")
+	verb := core.TrimPrefix(key, "i18n.progress.")
 	if len(args) > 0 {
 		if subj, ok := args[0].(string); ok {
 			return ProgressSubject(verb, subj)
@@ -38,14 +37,14 @@ func (h ProgressHandler) Handle(key string, args []any, next func() string) stri
 type CountHandler struct{}
 
 func (h CountHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.count.")
+	return core.HasPrefix(key, "i18n.count.")
 }
 
 func (h CountHandler) Handle(key string, args []any, next func() string) string {
-	noun := strings.TrimPrefix(key, "i18n.count.")
+	noun := core.TrimPrefix(key, "i18n.count.")
 	if len(args) > 0 {
 		count := toInt(args[0])
-		return fmt.Sprintf("%d %s", count, Pluralize(noun, count))
+		return core.Sprintf("%d %s", count, Pluralize(noun, count))
 	}
 	return noun
 }
@@ -54,11 +53,11 @@ func (h CountHandler) Handle(key string, args []any, next func() string) string
 type DoneHandler struct{}
 
 func (h DoneHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.done.")
+	return core.HasPrefix(key, "i18n.done.")
 }
 
 func (h DoneHandler) Handle(key string, args []any, next func() string) string {
-	verb := strings.TrimPrefix(key, "i18n.done.")
+	verb := core.TrimPrefix(key, "i18n.done.")
 	if len(args) > 0 {
 		if subj, ok := args[0].(string); ok {
 			return ActionResult(verb, subj)
@@ -71,11 +70,11 @@ func (h DoneHandler) Handle(key string, args []any, next func() string) string {
 type FailHandler struct{}
 
 func (h FailHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.fail.")
+	return core.HasPrefix(key, "i18n.fail.")
 }
 
 func (h FailHandler) Handle(key string, args []any, next func() string) string {
-	verb := strings.TrimPrefix(key, "i18n.fail.")
+	verb := core.TrimPrefix(key, "i18n.fail.")
 	if len(args) > 0 {
 		if subj, ok := args[0].(string); ok {
 			return ActionFailed(verb, subj)
@@ -88,14 +87,14 @@ func (h FailHandler) Handle(key string, args []any, next func() string) string {
 type NumericHandler struct{}
 
 func (h NumericHandler) Match(key string) bool {
-	return strings.HasPrefix(key, "i18n.numeric.")
+	return core.HasPrefix(key, "i18n.numeric.")
 }
 
 func (h NumericHandler) Handle(key string, args []any, next func() string) string {
 	if len(args) == 0 {
 		return next()
 	}
-	format := strings.TrimPrefix(key, "i18n.numeric.")
+	format := core.TrimPrefix(key, "i18n.numeric.")
 	switch format {
 	case "number", "int":
 		return FormatNumber(toInt64(args[0]))
diff --git a/i18n.go b/i18n.go
index e0d3188..d444636 100644
--- a/i18n.go
+++ b/i18n.go
@@ -2,9 +2,9 @@ package i18n
 
 import (
 	"bytes"
-	"errors"
-	"strings"
 	"text/template"
+
+	"dappco.re/go/core"
 )
 
 // T translates a message using the default service.
@@ -24,7 +24,7 @@ func Raw(messageID string, args ...any) string {
 }
 
 // ErrServiceNotInitialised is returned when the service is not initialised.
-var ErrServiceNotInitialised = errors.New("i18n: service not initialised")
+var ErrServiceNotInitialised = core.NewError("i18n: service not initialised")
 
 // ErrServiceNotInitialized is deprecated: use ErrServiceNotInitialised.
 var ErrServiceNotInitialized = ErrServiceNotInitialised
@@ -109,7 +109,7 @@ func executeIntentTemplate(tmplStr string, data templateData) string {
 }
 
 func applyTemplate(text string, data any) string {
-	if !strings.Contains(text, "{{") {
+	if !core.Contains(text, "{{") {
 		return text
 	}
 	if cached, ok := templateCache.Load(text); ok {
diff --git a/integration/calibrate_test.go b/integration/calibrate_test.go
index f63b71a..161d6f4 100644
--- a/integration/calibrate_test.go
+++ b/integration/calibrate_test.go
@@ -2,10 +2,10 @@ package integration
 
 import (
 	"context"
-	"fmt"
 	"sort"
 	"testing"
 
+	"dappco.re/go/core"
 	i18n "dappco.re/go/core/i18n"
 	"forge.lthn.ai/core/go-inference"
 	_ "forge.lthn.ai/core/go-mlx" // registers Metal backend
@@ -549,7 +549,7 @@ func TestCalibrateDomains_1Bvs27B(t *testing.T) {
 			disagreements++
 			truth := ""
 			if r.TrueDomain != "" {
-				truth = fmt.Sprintf(" [truth=%s]", r.TrueDomain)
+				truth = core.Sprintf(" [truth=%s]", r.TrueDomain)
 			}
 			t.Logf("  DISAGREE: 1B=%s 27B=%s%s | %.60s", r.DomainA, r.DomainB, truth, r.Text)
 			if disagreements >= 50 {
diff --git a/integration/classify_test.go b/integration/classify_test.go
index 839f2e2..37e73bd 100644
--- a/integration/classify_test.go
+++ b/integration/classify_test.go
@@ -3,12 +3,10 @@ package integration
 import (
 	"bytes"
 	"context"
-	"encoding/json"
-	"fmt"
-	"strings"
 	"testing"
 	"time"
 
+	"dappco.re/go/core"
 	i18n "dappco.re/go/core/i18n"
 	"forge.lthn.ai/core/go-inference"
 	_ "forge.lthn.ai/core/go-mlx" // registers Metal backend
@@ -24,9 +22,9 @@ func TestClassifyCorpus_Integration(t *testing.T) {
 	// Build 50 technical prompts for throughput measurement
 	var lines []string
 	for i := 0; i < 50; i++ {
-		lines = append(lines, fmt.Sprintf(`{"id":%d,"prompt":"Delete the configuration file and rebuild the project"}`, i))
+		lines = append(lines, core.Sprintf(`{"id":%d,"prompt":"Delete the configuration file and rebuild the project"}`, i))
 	}
-	input := strings.NewReader(strings.Join(lines, "\n") + "\n")
+	input := core.NewReader(core.Join("\n", lines...) + "\n")
 
 	var output bytes.Buffer
 	start := time.Now()
@@ -58,10 +56,10 @@ func TestClassifyCorpus_Integration(t *testing.T) {
 		}
 
 		// Also inspect the output JSONL for misclassified entries
-		outLines := strings.Split(strings.TrimSpace(output.String()), "\n")
+		outLines := core.Split(core.Trim(output.String()), "\n")
 		for _, line := range outLines {
 			var record map[string]any
-			if err := json.Unmarshal([]byte(line), &record); err == nil {
+			if r := core.JSONUnmarshal([]byte(line), &record); r.OK {
 				if record["domain_1b"] != "technical" {
 					t.Logf("  misclassified: id=%v domain_1b=%v", record["id"], record["domain_1b"])
 				}
diff --git a/loader.go b/loader.go
index 3ca917c..41c02df 100644
--- a/loader.go
+++ b/loader.go
@@ -1,12 +1,12 @@
 package i18n
 
 import (
-	"encoding/json"
 	"io/fs"
 	"path"
 	"strings"
 	"sync"
 
+	"dappco.re/go/core"
 	log "dappco.re/go/core/log"
 )
 
@@ -29,8 +29,8 @@ func NewFSLoader(fsys fs.FS, dir string) *FSLoader {
 func (l *FSLoader) Load(lang string) (map[string]Message, *GrammarData, error) {
 	variants := []string{
 		lang + ".json",
-		strings.ReplaceAll(lang, "-", "_") + ".json",
-		strings.ReplaceAll(lang, "_", "-") + ".json",
+		core.Replace(lang, "-", "_") + ".json",
+		core.Replace(lang, "_", "-") + ".json",
 	}
 
 	var data []byte
@@ -47,8 +47,8 @@ func (l *FSLoader) Load(lang string) (map[string]Message, *GrammarData, error) {
 	}
 
 	var raw map[string]any
-	if err := json.Unmarshal(data, &raw); err != nil {
-		return nil, nil, log.E("FSLoader.Load", "invalid JSON in locale: "+lang, err)
+	if r := core.JSONUnmarshal(data, &raw); !r.OK {
+		return nil, nil, log.E("FSLoader.Load", "invalid JSON in locale: "+lang, r.Value.(error))
 	}
 
 	messages := make(map[string]Message)
@@ -72,11 +72,11 @@ func (l *FSLoader) Languages() []string {
 			return
 		}
 		for _, entry := range entries {
-			if entry.IsDir() || !strings.HasSuffix(entry.Name(), ".json") {
+			if entry.IsDir() || !core.HasSuffix(entry.Name(), ".json") {
 				continue
 			}
-			lang := strings.TrimSuffix(entry.Name(), ".json")
-			lang = strings.ReplaceAll(lang, "_", "-")
+			lang := core.TrimSuffix(entry.Name(), ".json")
+			lang = core.Replace(lang, "_", "-")
 			l.languages = append(l.languages, lang)
 		}
 	})
@@ -106,9 +106,9 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 
 		switch v := value.(type) {
 		case string:
-			if grammar != nil && strings.HasPrefix(fullKey, "gram.word.") {
-				wordKey := strings.TrimPrefix(fullKey, "gram.word.")
-				grammar.Words[strings.ToLower(wordKey)] = v
+			if grammar != nil && core.HasPrefix(fullKey, "gram.word.") {
+				wordKey := core.TrimPrefix(fullKey, "gram.word.")
+				grammar.Words[core.Lower(wordKey)] = v
 				continue
 			}
 			out[fullKey] = Message{Text: v}
@@ -127,12 +127,12 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 				if gerund, ok := v["gerund"].(string); ok {
 					forms.Gerund = gerund
 				}
-				grammar.Verbs[strings.ToLower(verbName)] = forms
+				grammar.Verbs[core.Lower(verbName)] = forms
 				continue
 			}
 
 			// Noun form object (under gram.noun.* or has gender field)
-			if grammar != nil && (strings.HasPrefix(fullKey, "gram.noun.") || isNounFormObject(v)) {
+			if grammar != nil && (core.HasPrefix(fullKey, "gram.noun.") || isNounFormObject(v)) {
 				nounName := key
 				if after, ok := strings.CutPrefix(fullKey, "gram.noun."); ok {
 					nounName = after
@@ -150,7 +150,7 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 					if gender, ok := v["gender"].(string); ok {
 						forms.Gender = gender
 					}
-					grammar.Nouns[strings.ToLower(nounName)] = forms
+					grammar.Nouns[core.Lower(nounName)] = forms
 					continue
 				}
 			}
@@ -161,7 +161,7 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 					if arr, ok := nd.([]any); ok {
 						for _, item := range arr {
 							if s, ok := item.(string); ok {
-								grammar.Signals.NounDeterminers = append(grammar.Signals.NounDeterminers, strings.ToLower(s))
+								grammar.Signals.NounDeterminers = append(grammar.Signals.NounDeterminers, core.Lower(s))
 							}
 						}
 					}
@@ -170,7 +170,7 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 					if arr, ok := va.([]any); ok {
 						for _, item := range arr {
 							if s, ok := item.(string); ok {
-								grammar.Signals.VerbAuxiliaries = append(grammar.Signals.VerbAuxiliaries, strings.ToLower(s))
+								grammar.Signals.VerbAuxiliaries = append(grammar.Signals.VerbAuxiliaries, core.Lower(s))
 							}
 						}
 					}
@@ -179,7 +179,7 @@ func flattenWithGrammar(prefix string, data map[string]any, out map[string]Messa
 					if arr, ok := vi.([]any); ok {
 						for _, item := range arr {
 							if s, ok := item.(string); ok {
-								grammar.Signals.VerbInfinitive = append(grammar.Signals.VerbInfinitive, strings.ToLower(s))
+								grammar.Signals.VerbInfinitive = append(grammar.Signals.VerbInfinitive, core.Lower(s))
 							}
 						}
 					}
diff --git a/localise.go b/localise.go
index b59da6a..794368c 100644
--- a/localise.go
+++ b/localise.go
@@ -2,8 +2,8 @@ package i18n
 
 import (
 	"os"
-	"strings"
 
+	"dappco.re/go/core"
 	"golang.org/x/text/language"
 )
 
@@ -95,8 +95,8 @@ func detectLanguage(supported []language.Tag) string {
 	if langEnv == "" {
 		return ""
 	}
-	baseLang := strings.Split(langEnv, ".")[0]
-	baseLang = strings.ReplaceAll(baseLang, "_", "-")
+	baseLang := core.Split(langEnv, ".")[0]
+	baseLang = core.Replace(baseLang, "_", "-")
 	parsedLang, err := language.Parse(baseLang)
 	if err != nil {
 		return ""
diff --git a/numbers.go b/numbers.go
index 49468f2..b84a4ea 100644
--- a/numbers.go
+++ b/numbers.go
@@ -1,15 +1,15 @@
 package i18n
 
 import (
-	"fmt"
 	"math"
 	"strconv"
-	"strings"
+
+	"dappco.re/go/core"
 )
 
 func getNumberFormat() NumberFormat {
 	lang := currentLangForGrammar()
-	if idx := strings.IndexAny(lang, "-_"); idx > 0 {
+	if idx := indexAny(lang, "-_"); idx > 0 {
 		lang = lang[:idx]
 	}
 	if fmt, ok := numberFormats[lang]; ok {
@@ -42,8 +42,8 @@ func FormatDecimalN(f float64, decimals int) string {
 	if fracInt == 0 {
 		return intStr
 	}
-	fracStr := fmt.Sprintf("%0*d", decimals, fracInt)
-	fracStr = strings.TrimRight(fracStr, "0")
+	fracStr := core.Sprintf("%0*d", decimals, fracInt)
+	fracStr = trimRight(fracStr, "0")
 	return intStr + nf.DecimalSep + fracStr
 }
 
@@ -57,7 +57,7 @@ func FormatPercent(f float64) string {
 	} else {
 		numStr = FormatDecimalN(pct, 1)
 	}
-	return fmt.Sprintf(nf.PercentFmt, numStr)
+	return core.Sprintf(nf.PercentFmt, numStr)
 }
 
 // FormatBytes formats bytes as human-readable size.
@@ -85,31 +85,31 @@ func FormatBytes(bytes int64) string {
 		value = float64(bytes) / KB
 		unit = "KB"
 	default:
-		return fmt.Sprintf("%d B", bytes)
+		return core.Sprintf("%d B", bytes)
 	}
 	intPart := int64(value)
 	fracPart := value - float64(intPart)
 	if fracPart < 0.05 {
-		return fmt.Sprintf("%d %s", intPart, unit)
+		return core.Sprintf("%d %s", intPart, unit)
 	}
 	fracDigit := int(math.Round(fracPart * 10))
 	if fracDigit == 10 {
-		return fmt.Sprintf("%d %s", intPart+1, unit)
+		return core.Sprintf("%d %s", intPart+1, unit)
 	}
-	return fmt.Sprintf("%d%s%d %s", intPart, nf.DecimalSep, fracDigit, unit)
+	return core.Sprintf("%d%s%d %s", intPart, nf.DecimalSep, fracDigit, unit)
 }
 
 // FormatOrdinal formats a number as an ordinal.
 func FormatOrdinal(n int) string {
 	lang := currentLangForGrammar()
-	if idx := strings.IndexAny(lang, "-_"); idx > 0 {
+	if idx := indexAny(lang, "-_"); idx > 0 {
 		lang = lang[:idx]
 	}
 	switch lang {
 	case "en":
 		return formatEnglishOrdinal(n)
 	default:
-		return fmt.Sprintf("%d.", n)
+		return core.Sprintf("%d.", n)
 	}
 }
 
@@ -119,17 +119,17 @@ func formatEnglishOrdinal(n int) string {
 		abs = -abs
 	}
 	if abs%100 >= 11 && abs%100 <= 13 {
-		return fmt.Sprintf("%dth", n)
+		return core.Sprintf("%dth", n)
 	}
 	switch abs % 10 {
 	case 1:
-		return fmt.Sprintf("%dst", n)
+		return core.Sprintf("%dst", n)
 	case 2:
-		return fmt.Sprintf("%dnd", n)
+		return core.Sprintf("%dnd", n)
 	case 3:
-		return fmt.Sprintf("%drd", n)
+		return core.Sprintf("%drd", n)
 	default:
-		return fmt.Sprintf("%dth", n)
+		return core.Sprintf("%dth", n)
 	}
 }
 
@@ -148,7 +148,7 @@ func formatIntWithSep(n int64, sep string) string {
 		}
 		return str
 	}
-	var result strings.Builder
+	result := core.NewBuilder()
 	for i, c := range str {
 		if i > 0 && (len(str)-i)%3 == 0 {
 			result.WriteString(sep)
@@ -160,3 +160,34 @@ func formatIntWithSep(n int64, sep string) string {
 	}
 	return result.String()
 }
+
+// indexAny returns the index of the first occurrence of any char in chars, or -1.
+func indexAny(s, chars string) int {
+	for i, c := range s {
+		for _, ch := range chars {
+			if c == ch {
+				return i
+			}
+		}
+	}
+	return -1
+}
+
+// trimRight returns s with all trailing occurrences of cutset removed.
+func trimRight(s, cutset string) string {
+	for len(s) > 0 {
+		found := false
+		r := rune(s[len(s)-1])
+		for _, c := range cutset {
+			if r == c {
+				found = true
+				break
+			}
+		}
+		if !found {
+			break
+		}
+		s = s[:len(s)-1]
+	}
+	return s
+}
diff --git a/reversal/classify_bench_test.go b/reversal/classify_bench_test.go
index 1075b3e..eb88be9 100644
--- a/reversal/classify_bench_test.go
+++ b/reversal/classify_bench_test.go
@@ -1,10 +1,10 @@
 package reversal
 
 import (
-	"fmt"
 	"sort"
-	"strings"
 	"testing"
+
+	"dappco.re/go/core"
 )
 
 // Domain categories for classification ground truth.
@@ -418,17 +418,17 @@ func TestClassification_LeaveOneOut(t *testing.T) {
 
 	// Print confusion matrix
 	t.Log("\nConfusion matrix (rows=actual, cols=predicted):")
-	var header strings.Builder
-	header.WriteString(fmt.Sprintf("  %-10s", ""))
+	header := core.NewBuilder()
+	header.WriteString(core.Sprintf("  %-10s", ""))
 	for _, d := range domains {
-		header.WriteString(fmt.Sprintf(" %10s", d[:4]))
+		header.WriteString(core.Sprintf(" %10s", d[:4]))
 	}
 	t.Log(header.String())
 	for _, actual := range domains {
-		var row strings.Builder
-		row.WriteString(fmt.Sprintf("  %-10s", actual[:4]))
+		row := core.NewBuilder()
+		row.WriteString(core.Sprintf("  %-10s", actual[:4]))
 		for _, predicted := range domains {
-			row.WriteString(fmt.Sprintf(" %10d", confusion[actual][predicted]))
+			row.WriteString(core.Sprintf(" %10d", confusion[actual][predicted]))
 		}
 		t.Log(row.String())
 	}
@@ -506,14 +506,14 @@ func TestClassification_TenseProfile(t *testing.T) {
 			}
 		}
 
-		var parts strings.Builder
-		parts.WriteString(fmt.Sprintf("%-10s verbs=%d", d, totalVerbs))
+		parts := core.NewBuilder()
+		parts.WriteString(core.Sprintf("%-10s verbs=%d", d, totalVerbs))
 		for _, tense := range tenses {
 			pct := 0.0
 			if totalVerbs > 0 {
 				pct = float64(tenseCounts[tense]) / float64(totalVerbs) * 100
 			}
-			parts.WriteString(fmt.Sprintf("  %s=%.0f%%", tense, pct))
+			parts.WriteString(core.Sprintf("  %s=%.0f%%", tense, pct))
 		}
 		t.Log(parts.String())
 	}
@@ -550,12 +550,12 @@ func TestClassification_TopVerbs(t *testing.T) {
 		sort.Slice(sorted, func(i, j int) bool { return sorted[i].count > sorted[j].count })
 
 		top := min(len(sorted), 8)
-		var verbs strings.Builder
+		verbs := core.NewBuilder()
 		for i := 0; i < top; i++ {
 			if i > 0 {
 				verbs.WriteString(", ")
 			}
-			verbs.WriteString(fmt.Sprintf("%s(%d)", sorted[i].verb, sorted[i].count))
+			verbs.WriteString(core.Sprintf("%s(%d)", sorted[i].verb, sorted[i].count))
 		}
 		t.Logf("%-10s unique=%d top: %s", d, len(verbCounts), verbs.String())
 	}
diff --git a/reversal/multiplier.go b/reversal/multiplier.go
index 3b89bbd..15e6d1d 100644
--- a/reversal/multiplier.go
+++ b/reversal/multiplier.go
@@ -1,9 +1,9 @@
 package reversal
 
 import (
-	"strings"
 	"unicode"
 
+	"dappco.re/go/core"
 	i18n "dappco.re/go/core/i18n"
 )
 
@@ -26,7 +26,7 @@ func NewMultiplierForLang(lang string) *Multiplier {
 // Expand produces: original + tense flips (past, gerund) + number flips (plural toggle) + combinations.
 // All output is deterministic and grammatically correct.
 func (m *Multiplier) Expand(text string) []string {
-	text = strings.TrimSpace(text)
+	text = core.Trim(text)
 	if text == "" {
 		return nil
 	}
@@ -140,7 +140,7 @@ func (m *Multiplier) applyVerbTransform(tokens []Token, vi int, targetTense stri
 
 	result[vi] = Token{
 		Raw:        newForm,
-		Lower:      strings.ToLower(newForm),
+		Lower:      core.Lower(newForm),
 		Type:       TokenVerb,
 		Confidence: 1.0,
 		VerbInfo: VerbMatch{
@@ -191,7 +191,7 @@ func (m *Multiplier) applyNounTransformOnTokens(tokens []Token, ni int) []Token
 
 	result[ni] = Token{
 		Raw:        newForm,
-		Lower:      strings.ToLower(newForm),
+		Lower:      core.Lower(newForm),
 		Type:       TokenNoun,
 		Confidence: 1.0,
 		NounInfo: NounMatch{
@@ -206,7 +206,7 @@ func (m *Multiplier) applyNounTransformOnTokens(tokens []Token, ni int) []Token
 
 // reconstruct joins tokens back into a string, preserving spacing.
 func reconstruct(tokens []Token) string {
-	var b strings.Builder
+	b := core.NewBuilder()
 	for i, tok := range tokens {
 		if i > 0 {
 			// Punctuation tokens that were split from the previous word
@@ -235,7 +235,7 @@ func preserveCase(original, replacement string) string {
 
 	// If the original is all uppercase (like "DELETE"), make replacement all uppercase.
 	if isAllUpper(original) && len(original) > 1 {
-		return strings.ToUpper(replacement)
+		return core.Upper(replacement)
 	}
 
 	// If the first character of the original is uppercase, capitalise the replacement.
diff --git a/reversal/tokeniser.go b/reversal/tokeniser.go
index 275462c..501665a 100644
--- a/reversal/tokeniser.go
+++ b/reversal/tokeniser.go
@@ -18,6 +18,7 @@ package reversal
 import (
 	"strings"
 
+	"dappco.re/go/core"
 	i18n "dappco.re/go/core/i18n"
 )
 
@@ -209,7 +210,7 @@ func (t *Tokeniser) buildNounIndex() {
 // Tier 3: Try reverse morphology rules and round-trip verify via
 // the forward function PluralForm().
 func (t *Tokeniser) MatchNoun(word string) (NounMatch, bool) {
-	word = strings.ToLower(strings.TrimSpace(word))
+	word = core.Lower(core.Trim(word))
 	if word == "" {
 		return NounMatch{}, false
 	}
@@ -250,27 +251,27 @@ func (t *Tokeniser) reverseRegularPlural(word string) []string {
 	var candidates []string
 
 	// Rule: consonant + "ies" → consonant + "y" (e.g., "entries" → "entry")
-	if strings.HasSuffix(word, "ies") && len(word) > 3 {
+	if core.HasSuffix(word, "ies") && len(word) > 3 {
 		base := word[:len(word)-3] + "y"
 		candidates = append(candidates, base)
 	}
 
 	// Rule: "ves" → "f" or "fe" (e.g., "wolves" → "wolf", "knives" → "knife")
-	if strings.HasSuffix(word, "ves") && len(word) > 3 {
+	if core.HasSuffix(word, "ves") && len(word) > 3 {
 		candidates = append(candidates, word[:len(word)-3]+"f")
 		candidates = append(candidates, word[:len(word)-3]+"fe")
 	}
 
 	// Rule: sibilant + "es" (e.g., "processes" → "process", "branches" → "branch")
-	if strings.HasSuffix(word, "ses") || strings.HasSuffix(word, "xes") ||
-		strings.HasSuffix(word, "zes") || strings.HasSuffix(word, "ches") ||
-		strings.HasSuffix(word, "shes") {
+	if core.HasSuffix(word, "ses") || core.HasSuffix(word, "xes") ||
+		core.HasSuffix(word, "zes") || core.HasSuffix(word, "ches") ||
+		core.HasSuffix(word, "shes") {
 		base := word[:len(word)-2] // strip "es"
 		candidates = append(candidates, base)
 	}
 
 	// Rule: drop "s" (e.g., "servers" → "server")
-	if strings.HasSuffix(word, "s") && len(word) > 1 {
+	if core.HasSuffix(word, "s") && len(word) > 1 {
 		base := word[:len(word)-1]
 		candidates = append(candidates, base)
 	}
@@ -285,7 +286,7 @@ func (t *Tokeniser) reverseRegularPlural(word string) []string {
 // Tier 3: Try reverse morphology rules and round-trip verify via
 // the forward functions PastTense() and Gerund().
 func (t *Tokeniser) MatchVerb(word string) (VerbMatch, bool) {
-	word = strings.ToLower(strings.TrimSpace(word))
+	word = core.Lower(core.Trim(word))
 	if word == "" {
 		return VerbMatch{}, false
 	}
@@ -358,7 +359,7 @@ func (t *Tokeniser) bestRoundTrip(target string, candidates []string, forward fu
 	// Priority 3: prefer candidate not ending in "e" (avoids phantom verbs
 	// with CCe endings like "walke", "processe")
 	for _, m := range matches {
-		if !strings.HasSuffix(m, "e") {
+		if !core.HasSuffix(m, "e") {
 			return m
 		}
 	}
@@ -402,12 +403,12 @@ func isVowelByte(b byte) bool {
 func (t *Tokeniser) reverseRegularPast(word string) []string {
 	var candidates []string
 
-	if !strings.HasSuffix(word, "ed") {
+	if !core.HasSuffix(word, "ed") {
 		return candidates
 	}
 
 	// Rule: consonant + "ied" → consonant + "y" (e.g., "copied" → "copy")
-	if strings.HasSuffix(word, "ied") && len(word) > 3 {
+	if core.HasSuffix(word, "ied") && len(word) > 3 {
 		base := word[:len(word)-3] + "y"
 		candidates = append(candidates, base)
 	}
@@ -448,14 +449,14 @@ func (t *Tokeniser) reverseRegularPast(word string) []string {
 func (t *Tokeniser) reverseRegularGerund(word string) []string {
 	var candidates []string
 
-	if !strings.HasSuffix(word, "ing") || len(word) < 4 {
+	if !core.HasSuffix(word, "ing") || len(word) < 4 {
 		return candidates
 	}
 
 	stem := word[:len(word)-3] // strip "ing"
 
 	// Rule: "ying" → "ie" (e.g., "dying" → "die")
-	if strings.HasSuffix(word, "ying") && len(word) > 4 {
+	if core.HasSuffix(word, "ying") && len(word) > 4 {
 		base := word[:len(word)-4] + "ie"
 		candidates = append(candidates, base)
 	}
@@ -488,15 +489,15 @@ func (t *Tokeniser) buildWordIndex() {
 	}
 	for key, display := range data.Words {
 		// Map the key itself (already lowercase)
-		t.words[strings.ToLower(key)] = key
+		t.words[core.Lower(key)] = key
 		// Map the display form (e.g., "URL" → "url", "SSH" → "ssh")
-		t.words[strings.ToLower(display)] = key
+		t.words[core.Lower(display)] = key
 	}
 }
 
 // IsDualClass returns true if the word exists in both verb and noun tables.
 func (t *Tokeniser) IsDualClass(word string) bool {
-	return t.dualClass[strings.ToLower(word)]
+	return t.dualClass[core.Lower(word)]
 }
 
 func (t *Tokeniser) buildDualClassIndex() {
@@ -519,7 +520,7 @@ func (t *Tokeniser) buildSignalIndex() {
 	// falls back per-field rather than silently disabling signals.
 	if data != nil && len(data.Signals.NounDeterminers) > 0 {
 		for _, w := range data.Signals.NounDeterminers {
-			t.nounDet[strings.ToLower(w)] = true
+			t.nounDet[core.Lower(w)] = true
 		}
 	} else {
 		for _, w := range []string{
@@ -534,7 +535,7 @@ func (t *Tokeniser) buildSignalIndex() {
 
 	if data != nil && len(data.Signals.VerbAuxiliaries) > 0 {
 		for _, w := range data.Signals.VerbAuxiliaries {
-			t.verbAux[strings.ToLower(w)] = true
+			t.verbAux[core.Lower(w)] = true
 		}
 	} else {
 		for _, w := range []string{
@@ -548,7 +549,7 @@ func (t *Tokeniser) buildSignalIndex() {
 
 	if data != nil && len(data.Signals.VerbInfinitive) > 0 {
 		for _, w := range data.Signals.VerbInfinitive {
-			t.verbInf[strings.ToLower(w)] = true
+			t.verbInf[core.Lower(w)] = true
 		}
 	} else {
 		t.verbInf["to"] = true
@@ -570,7 +571,7 @@ func defaultWeights() map[string]float64 {
 // MatchWord performs a case-insensitive lookup in the words map.
 // Returns the category key and true if found, or ("", false) otherwise.
 func (t *Tokeniser) MatchWord(word string) (string, bool) {
-	cat, ok := t.words[strings.ToLower(word)]
+	cat, ok := t.words[core.Lower(word)]
 	return cat, ok
 }
 
@@ -583,13 +584,13 @@ func (t *Tokeniser) MatchArticle(word string) (string, bool) {
 		return "", false
 	}
 
-	lower := strings.ToLower(word)
+	lower := core.Lower(word)
 
-	if lower == strings.ToLower(data.Articles.IndefiniteDefault) ||
-		lower == strings.ToLower(data.Articles.IndefiniteVowel) {
+	if lower == core.Lower(data.Articles.IndefiniteDefault) ||
+		lower == core.Lower(data.Articles.IndefiniteVowel) {
 		return "indefinite", true
 	}
-	if lower == strings.ToLower(data.Articles.Definite) {
+	if lower == core.Lower(data.Articles.Definite) {
 		return "definite", true
 	}
 
@@ -613,7 +614,7 @@ var clauseBoundaries = map[string]bool{
 // Pass 1 classifies unambiguous tokens and marks dual-class base forms.
 // Pass 2 resolves ambiguous tokens using weighted disambiguation signals.
 func (t *Tokeniser) Tokenise(text string) []Token {
-	text = strings.TrimSpace(text)
+	text = core.Trim(text)
 	if text == "" {
 		return nil
 	}
@@ -628,7 +629,7 @@ func (t *Tokeniser) Tokenise(text string) []Token {
 
 		// Classify the word portion (if any).
 		if word != "" {
-			tok := Token{Raw: raw, Lower: strings.ToLower(word)}
+			tok := Token{Raw: raw, Lower: core.Lower(word)}
 
 			if artType, ok := t.MatchArticle(word); ok {
 				// Articles are unambiguous.
@@ -943,7 +944,7 @@ func splitTrailingPunct(s string) (string, string) {
 	}
 
 	// Check for "..." suffix first (3-char pattern).
-	if strings.HasSuffix(s, "...") {
+	if core.HasSuffix(s, "...") {
 		return s[:len(s)-3], "..."
 	}
 	// Check single-char trailing punctuation.
diff --git a/service.go b/service.go
index 566d23b..5bfde0a 100644
--- a/service.go
+++ b/service.go
@@ -2,16 +2,14 @@ package i18n
 
 import (
 	"embed"
-	"encoding/json"
-	"fmt"
 	"io/fs"
 	"maps"
 	"path"
 	"slices"
-	"strings"
 	"sync"
 	"sync/atomic"
 
+	"dappco.re/go/core"
 	log "dappco.re/go/core/log"
 	"golang.org/x/text/language"
 )
@@ -185,8 +183,8 @@ func AddLoader(loader Loader) {
 
 func (s *Service) loadJSON(lang string, data []byte) error {
 	var raw map[string]any
-	if err := json.Unmarshal(data, &raw); err != nil {
-		return err
+	if r := core.JSONUnmarshal(data, &raw); !r.OK {
+		return r.Value.(error)
 	}
 	messages := make(map[string]Message)
 	grammarData := &GrammarData{
@@ -324,8 +322,8 @@ func (s *Service) resolveWithFallback(messageID string, data any) string {
 	if text := s.tryResolve(s.fallbackLang, messageID, data); text != "" {
 		return text
 	}
-	if strings.Contains(messageID, ".") {
-		parts := strings.Split(messageID, ".")
+	if core.Contains(messageID, ".") {
+		parts := core.Split(messageID, ".")
 		verb := parts[len(parts)-1]
 		commonKey := "common.action." + verb
 		if text := s.tryResolve(s.currentLang, commonKey, data); text != "" {
@@ -394,7 +392,7 @@ func (s *Service) getEffectiveFormality(data any) Formality {
 				return f
 			}
 		case string:
-			switch strings.ToLower(f) {
+			switch core.Lower(f) {
 			case "formal":
 				return FormalityFormal
 			case "informal":
@@ -408,7 +406,7 @@ func (s *Service) getEffectiveFormality(data any) Formality {
 func (s *Service) handleMissingKey(key string, args []any) string {
 	switch s.mode {
 	case ModeStrict:
-		panic(fmt.Sprintf("i18n: missing translation key %q", key))
+		panic(core.Sprintf("i18n: missing translation key %q", key))
 	case ModeCollect:
 		var argsMap map[string]any
 		if len(args) > 0 {
@@ -506,7 +504,7 @@ func (s *Service) LoadFS(fsys fs.FS, dir string) error {
 		return log.E("Service.LoadFS", "read locales directory", err)
 	}
 	for _, entry := range entries {
-		if entry.IsDir() || !strings.HasSuffix(entry.Name(), ".json") {
+		if entry.IsDir() || !core.HasSuffix(entry.Name(), ".json") {
 			continue
 		}
 		filePath := path.Join(dir, entry.Name())
@@ -514,8 +512,8 @@ func (s *Service) LoadFS(fsys fs.FS, dir string) error {
 		if err != nil {
 			return log.E("Service.LoadFS", "read locale: "+entry.Name(), err)
 		}
-		lang := strings.TrimSuffix(entry.Name(), ".json")
-		lang = strings.ReplaceAll(lang, "_", "-")
+		lang := core.TrimSuffix(entry.Name(), ".json")
+		lang = core.Replace(lang, "_", "-")
 		if err := s.loadJSON(lang, data); err != nil {
 			return log.E("Service.LoadFS", "parse locale: "+entry.Name(), err)
 		}
diff --git a/time.go b/time.go
index 241b846..06bda9d 100644
--- a/time.go
+++ b/time.go
@@ -1,8 +1,9 @@
 package i18n
 
 import (
-	"fmt"
 	"time"
+
+	"dappco.re/go/core"
 )
 
 // TimeAgo returns a localised relative time string.
@@ -28,12 +29,12 @@ func TimeAgo(t time.Time) string {
 func FormatAgo(count int, unit string) string {
 	svc := Default()
 	if svc == nil {
-		return fmt.Sprintf("%d %ss ago", count, unit)
+		return core.Sprintf("%d %ss ago", count, unit)
 	}
 	key := "time.ago." + unit
 	result := svc.T(key, map[string]any{"Count": count})
 	if result == key {
-		return fmt.Sprintf("%d %s ago", count, Pluralize(unit, count))
+		return core.Sprintf("%d %s ago", count, Pluralize(unit, count))
 	}
 	return result
 }
diff --git a/validate.go b/validate.go
index 2588531..c7f3bec 100644
--- a/validate.go
+++ b/validate.go
@@ -4,9 +4,8 @@ package i18n
 
 import (
 	"context"
-	"fmt"
-	"strings"
 
+	"dappco.re/go/core"
 	log "dappco.re/go/core/log"
 	"forge.lthn.ai/core/go-inference"
 )
@@ -45,7 +44,7 @@ type IrregularResult struct {
 
 // articlePrompt builds a fill-in-the-blank prompt for article prediction.
 func articlePrompt(noun string) string {
-	return fmt.Sprintf(
+	return core.Sprintf(
 		"Complete with the correct article (a/an/the): ___ %s. Answer with just the article:",
 		noun,
 	)
@@ -53,7 +52,7 @@ func articlePrompt(noun string) string {
 
 // irregularPrompt builds a fill-in-the-blank prompt for irregular verb prediction.
 func irregularPrompt(verb, tense string) string {
-	return fmt.Sprintf(
+	return core.Sprintf(
 		"What is the %s form of the verb '%s'? Answer with just the word:",
 		tense, verb,
 	)
@@ -61,14 +60,14 @@ func irregularPrompt(verb, tense string) string {
 
 // collectGenerated runs a single-token generation and returns the trimmed, lowercased output.
 func collectGenerated(ctx context.Context, m inference.TextModel, prompt string) (string, error) {
-	var sb strings.Builder
+	sb := core.NewBuilder()
 	for tok := range m.Generate(ctx, prompt, inference.WithMaxTokens(1), inference.WithTemperature(0.05)) {
 		sb.WriteString(tok.Text)
 	}
 	if err := m.Err(); err != nil {
 		return "", err
 	}
-	return strings.TrimSpace(strings.ToLower(sb.String())), nil
+	return core.Trim(core.Lower(sb.String())), nil
 }
 
 // ValidateArticle checks whether a given article usage is grammatically correct
@@ -80,7 +79,7 @@ func ValidateArticle(ctx context.Context, m inference.TextModel, noun string, ar
 	if err != nil {
 		return ArticleResult{}, log.E("ValidateArticle", "validate: "+noun, err)
 	}
-	given := strings.TrimSpace(strings.ToLower(article))
+	given := core.Trim(core.Lower(article))
 	return ArticleResult{
 		Noun:      noun,
 		Given:     given,
@@ -99,7 +98,7 @@ func ValidateIrregular(ctx context.Context, m inference.TextModel, verb string,
 	if err != nil {
 		return IrregularResult{}, log.E("ValidateIrregular", "validate: "+verb+" ("+tense+")", err)
 	}
-	given := strings.TrimSpace(strings.ToLower(form))
+	given := core.Trim(core.Lower(form))
 	return IrregularResult{
 		Verb:      verb,
 		Tense:     tense,
diff --git a/validate_test.go b/validate_test.go
index 5574394..d5b1718 100644
--- a/validate_test.go
+++ b/validate_test.go
@@ -4,10 +4,10 @@ package i18n
 
 import (
 	"context"
-	"fmt"
 	"iter"
 	"testing"
 
+	"dappco.re/go/core"
 	"forge.lthn.ai/core/go-inference"
 )
 
@@ -73,7 +73,7 @@ func newMockIrregularModel(forms map[string]string) *mockGenerateModel {
 // containsVerb checks if the prompt contains the verb in the expected format.
 func containsVerb(prompt, verb string) bool {
 	return len(prompt) > 0 && len(verb) > 0 &&
-		contains(prompt, fmt.Sprintf("'%s'", verb))
+		contains(prompt, core.Sprintf("'%s'", verb))
 }
 
 // contains is a simple substring check (avoids importing strings in test).