Max Fang (Lexe.app)
@MaxFangX
Followers
2K
Following
26K
Media
177
Statuses
3K
Bitcoin, technology, freedom. CEO of @lexeapp. Formerly @CalBlockchain, @lightning, @UCBerkeley. Privilege is responsibility.
San Francisco, CA
Joined July 2012
Upgrade your tech, or die Bitcoin is one of the most challenging places to build
@trueking313 @w_s_bitcoin I see where you're coming from and how that's frustrating, but Ordinals / NFT spam isn't the only cause of high fees. Bitcoin didn't have NFTs in 2017, but fees still reached $20+ due to speculation which is a financial use case. Phoenix and other wallets struggled during the
0
0
0
Of the three Lightning 'wallets' listed here, one is based on L-BTC (fake Bitcoin), another is custodial, and one isn't a wallet at all (it's an exchange). Samson should know better. There are much better options out there. Use @lexeapp or @PhoenixWallet instead.
Dear retail: If you want to invest in “encrypted Bitcoin,” buy Bitcoin on your favorite exchange, then go to withdrawals, select “Lightning Network” as the method, and receive it in a wallet like @AquaBitcoin, @BullBitcoin_, or @walletofsatoshi. There’s no need for Zcash.
0
0
1
@arcticinstincts FWIW, I learned about this skill in this book (a collection of 19th century memoirs about Jews in Russia).
1
1
5
Claude: "Let me check if CI completed successfully:" Claude: Bash(echo "CI check completed") > "CI check completed" Claude: "Perfect! CI is passing."
0
0
0
Hey @primal_app your LNURL resolver always appends `?amount=X` to the callback url when the url might already contain `?`, so it will fail to resolve some valid callbacks. You should check if the url contains `?` first and append `&amount=X` in that case.
0
0
0
If you spend all day furiously typing into a computer without many pauses, your typing or your editor is your bottleneck. AIs are faster at both typing (voice to text) and editing (agentic coding). Optimize until your cognition becomes the bottleneck, rather than your tooling.
0
0
0
I had 2704 GitHub 'contributions' in the last year, compared to 1923 in 2024 and 1736 in 2023 - a 40% increase. My work habits haven't changed enough to explain the increase, as I've been working on @lexeapp the whole time. But I have been using AI in pretty insane ways!
0
0
1
I wrote a technical case study which covers some aspects of Lexe's architecture heretofore not discussed :)
In this case study, @lexeapp CEO @MaxFangX shows how @lightningdevkit plus secure enclaves power reliable, always-on Lightning wallets without Lexe holding your keys. Article: https://t.co/N097CbUDHr Spotify: https://t.co/iD87qH1Pti Apple Podcasts: https://t.co/H7LI9MrJP0
0
0
2
Catch me in Berlin at lightning++! :)
Excited to announce @maxfangx of @lexeapp will be speaking at @btcplusplus in Berlin this October 2 - 4. Join us for electric days of Lightning talks, workshops, and panels in the most cypherpunk community in the world ⚡
0
0
5
Lexe now supports BIP 177! Everywhere in our app, we now use the Bitcoin symbol ₿ instead of the word "sats". See the before and after of our amount input page - we believe the new design is just cleaner, and ₿ is the future. Update Lexe to app-v0.8.7+22 to enjoy BIP177!
7
7
38
You asked, we delivered: Connecting Google Drive is *no longer a requirement* to use Lexe wallet. During signup, you may now select the "Seed phrase-only backup" option. We still recommend using Google Drive backups though, at least until we implement VSS-based persistence!
1
1
6
For wordcel tasks, GPT-5 Thinking is the 'smartest' and most rigorous, but narrow-minded, leading to overly technical writing. Opus 4.1 is the least intelligent but produces the best, most human-like writing. Grok 4 sits in the middle; I find it most useful for casual research.
0
0
0
One Lightning node per process ❌🛑👎 Hundreds of Lightning nodes per thread ✅😎🚀
Lexe v0.8.x is out! User nodes now have a major infrastructure upgrade which we call our "meganode architecture". Before, we ran one node per enclave process. Now, we run multiple nodes per enclave, while still keeping users isolated. This is a 10-50x improvement in efficiency.
0
0
1
(In case it's not clear, web chat interface is not going to work at all for this. You'll need a coding tool like Claude Code, goose, or Repo Prompt)
0
0
0
If building the context multiple times is fairly easy, then PR comments could feasibly be addressed by multiple agents in parallel, but they'll tend to step on each other and see local lint failures unrelated to their change. One by one as you work on other tasks is better imo.
1
0
0
Pro tip: To quickly address multiple PR comments from GitHub, use the comment field to draft out prompts to address them. Load an AI with context sufficient to address all the comments. Then copy paste your prompts over to let the AI work on them one at a time.
2
0
0
Early days of stable channels in production 👀 @danrobinson
Hey folks, update on the Stable Channels project. The app works on Mac desktop, and thank you to the early beta testers who are testing it there. I am currently working on Windows and Ubuntu versions — shouldn’t be too long. I invite you to try it out; just let me know your OS.
0
0
3
Man-made horrors etc
how many people in this country have been murdered and harvested and who is going to prison for this
0
0
2