diff --git a/rfcs/Base8.md b/rfcs/Base8.md
new file mode 100644
index 0000000..1e013ac
--- /dev/null
+++ b/rfcs/Base8.md
@@ -0,0 +1,45 @@
+# Base8
+
+The multibase base8 prefix is the character `7`. This spec is derived from
+RFC4648.
+
+## Encoding
+
+Map each 3 bytes (8-bit word), with the most significant bit on the left side
+(big-endian), to 8 3-bit words as follows:
+
+`[⁰b₇₆₅₄₃₂₁₀, ¹b₇₆₅₄₃₂₁₀, ²b₇₆₅₄₃₂₁₀]`
+
+`[⁰b₇₆₅, ⁰b₄₃₂, ⁰b₁₀¹b₇, ¹b₆₅₄, ¹b₃₂₁, ¹b₀²b₇₆, ²b₅₄₃, ²b₂₁₀]`
+
+Then map their values as big-endian unsigned ints to their chars:
+
+```
+000 → '0'
+001 → '1'
+...
+111 → '7'
+```
+
+Treat missing subbits as zero and optionally pad:
+
+## Decoding
+
+Map chars to the following 3-bit words:
+
+```
+'0' → 000
+'1' → 001
+... 
+'7' → 111
+```
+
+Then map each 8 3-bit words, with the most significant bit on the left side
+(big-endian), to 3 bytes (8-bit word) as follows:
+
+`[⁰b₂₁₀, ¹b₂₁₀, ²b₂₁₀, ³b₂₁₀, ⁴b₂₁₀, ⁵b₂₁₀, ⁶b₂₁₀, ⁷b₂₁₀]`
+
+`[⁰b₂₁₀¹b₂₁₀²b₂₁, ²b₀³b₂₁₀⁴b₂₁₀⁵b₂, ⁵b₁₀⁶b₂₁₀⁷b₂₁₀]`
+
+If there are not enough bits to complete the last 8-bit word then drop that last
+incomplete 8-bit word.
diff --git a/tests/test1.csv b/tests/test1.csv
index 97e807c..c844032 100644
--- a/tests/test1.csv
+++ b/tests/test1.csv
@@ -1,6 +1,6 @@
 encoding, "Decentralize everything!!"
 base2, "001000100011001010110001101100101011011100111010001110010011000010110110001101001011110100110010100100000011001010111011001100101011100100111100101110100011010000110100101101110011001110010000100100001"
-base8, "71043126154533472162302661513646244031273145344745643206455631620441"
+base8, "72106254331267164344605543227514510062566312711713506415133463441102"
 base10, "9429328951066508984658627669258025763026247056774804621697313"
 base16, "f446563656e7472616c697a652065766572797468696e672121"
 base16upper, "F446563656E7472616C697A652065766572797468696E672121"
diff --git a/tests/test2.csv b/tests/test2.csv
index 3b95ebf..2b944a2 100644
--- a/tests/test2.csv
+++ b/tests/test2.csv
@@ -1,6 +1,6 @@
 encoding, "yes mani !"
 base2, "001111001011001010111001100100000011011010110000101101110011010010010000000100001"
-base8, "7171312714403326055632220041"
+base8, "7362625631006654133464440102"
 base10, "9573277761329450583662625"
 base16, "f796573206d616e692021"
 base16upper, "F796573206D616E692021"
diff --git a/tests/test3.csv b/tests/test3.csv
index 8ddea2b..2f79dd9 100644
--- a/tests/test3.csv
+++ b/tests/test3.csv
@@ -1,6 +1,6 @@
 encoding, "hello world"
 base2, "00110100001100101011011000110110001101111001000000111011101101111011100100110110001100100"
-base8, "7064145330661571007355734466144"
+base8, "7320625543306744035667562330620"
 base10, "9126207244316550804821666916"
 base16, "f68656c6c6f20776f726c64"
 base16upper, "F68656C6C6F20776F726C64"
diff --git a/tests/test4.csv b/tests/test4.csv
index 7fd4fc3..eb3d91f 100644
--- a/tests/test4.csv
+++ b/tests/test4.csv
@@ -1,6 +1,6 @@
 encoding, "\x00yes mani !"
 base2, "00000000001111001011001010111001100100000011011010110000101101110011010010010000000100001"
-base8, "7000171312714403326055632220041"
+base8, "7000745453462015530267151100204"
 base10, "90573277761329450583662625"
 base16, "f00796573206d616e692021"
 base16upper, "F00796573206D616E692021"