Proof of work that runs on your CPU

01How nodes meet

Nothing dials by IP, and there is no clearnet transport to fall back to. A node is addressed by its own public key and reached at its own Tor hidden service, so it works behind CGNAT and no peer learns where it is.

# a node's address is its key, like a .onion
52 base32 chars .btf

# what it publishes, signed with that key
descriptor = address + encryption key + its .onion
# nobody can forge one for an address they do not own

baked seed         a pinned .onion in the binary
btfpeers.json      peers that answered before
peer exchange      descriptors handed over by peers
nostr              looked up only when nothing else has one

magic    bf 20 5c fd
-> version  101 | services | time | addr | height
<- version  101 | services | time | addr | height

# the height above is what the status page reads

02How a wallet is rebuilt

Twelve words rebuild the keys. The derivation is written down in full, with test vectors and a script that reproduces it using nothing but Python's standard library, so the words keep working even if this software does not.

$ bitflash -newphrase
# printed once, to the terminal, never to debug.log

$ bitflash -restorephrase="twelve words"
# walks 100 addresses at a time, rescanning,
# and stops when a whole batch turns up nothing

$ bitflash -backupwallet=backup.sqlite
# a single file that opens on its own

# new wallets are one self-contained wallet.sqlite;
# an old wallet.dat still works and the app
# converts it in one click. copying wallet.dat by
# hand is NOT a backup: berkeley db needs database/

$ bitflash -dumpwallet=keys.txt
$ bitflash -importwallet=keys.txt
# one key per line, no database, any machine
# it is your keys in the clear: move it, delete it

Keys made before the seed was installed are random: they do not come back from the words. That is why both backups exist, and it is where people lose money. Since 1.2.20 the wallet is a single SQLite file by default: wallet storage, and the derivation specification.

03How you mine

RandomX is memory-hard and built for general-purpose CPUs, so a laptop competes with a server and an ASIC has no advantage worth the electricity.

$ ./bitflash -nogui -gen
# or from the window: Options > Mining Mode

$ ./bitflash -nogui -gen -operator
# the pool server is in the node,
# on a second port of this node's own onion

$ mine.cmd YOUR_BTF_ADDRESS
# Bitflash Miner: XMRig + Tor, from the releases page
# a payment address, not a .btf; fee 1%
# PoW v2 from 2026-09-21 12:00 UTC on mainnet

04Parameters

Ticker

BTF

Proof of work

RandomX, memory-hard, CPU

Block time

~2 minutes

Retarget

every 30 blocks

Reward

50 BTF, halving every 210,000

Halving

~292 days

Supply

21,000,000 BTF

Maturity

120 blocks

Sig ops

20,000 per block

P2P port

8433

Testnet port

18433

Protocol

101

Premine

none

Licence

MIT

05Get it

Download v1.2.28ManualSourceDiscord

# check what you downloaded before you run it
$ sha256sum -c SHA256SUMS

# linux
$ chmod +x Bitflash-*.AppImage && ./Bitflash-*.AppImage

# windows: extract the zip, run Bitflash.exe

Consensus rules have changed since the first releases, so an old build quietly follows a different chain. Keep the node current, write the twelve words on paper, and do not put in more than you are willing to lose.

Bitflash has no company behind it. The first exchange listing is paid for by a community treasury: a 2-of-3 multisig on the chain, its balance and every movement read from the blocks.