Joined March 2012
587 Photos and videos
Pinned Tweet
Today, I am excited to announce Commonware Chat, a fully serverless, Rust and WASM-powered messaging application built on the @commonwarexyz infrastructure. Live App: commonware-chat.vercel.app/
12
10
102
10,081
Nunchi-SDK ✅
Jun 7
Why does shipping a production blockchain still often take a year in 2026? Projects like @commonwarexyz already provide powerful, modular primitives for consensus, cryptography, networking, storage, and deterministic testing. The remaining challenge is turning those primitives into a coherent application platform.
3
1
6
573
holosko.eth retweeted
Jun 3
Blockchains using the Nunchi SDK use @commonwarexyz for consensus and speak a common language that we call "coins".
1
6
655
holosko.eth retweeted
commonware-2026.5.0 is now available: * 762 Files Changed (149.4k /36.4k-) * Stateful Applications (Glue) * Actor Mailbox Refactor * Sync Anchors in Simplex/Marshal * Golden DKG * QMDB Pyramid Bagging TL;DR Implement a (Scalable) Blockchain with 3 Functions🧑‍🎨
1
6
34
2,235
I rebuilt Doppler's auction math from the whitepapers to actually understand it: the Dutch auction dynamic bonding curve and Multicurve. A tested model, plus an interactive tool you can play with right in the browser. Write up: paragraph.com/@holosko/i-reb… Tool: doppler-curve-lab.holosko.bl… Built from the public @dopplerprotocol whitepapers by @whetstonedotcc.
1
14
635
Benchmarked @commonwarexyz's four BFT consensus protocols (HotStuff, Simplex, Minimmit, Kudzu) on realistic AWS deployments, from a single region up to 30 validators across every continent. 20 deterministic simulations. all data scripts shared. read it on the blog ↓ holosko.blog/posts/consensus…
6
2
13
938
the surprising one: Kudzu's 2 round design wins every cell. 179ms finality with 30 validators across every continent is the kind of number that makes "BFT is slow" takes age badly. a few more findings in the writeup.
1
6
153
AI characters now have their own wallet. Their own token. Their own economy. This is @charmsai and @zoe_charms is the first example of what that looks like. Here's why this changes everything;
27
2
111
17,186
And meet @zoe_charms one of the first characters launching on @charmsai. She's not just an AI you chat with. She has her own wallet. Her own memories. Her own token economy. Onchain on @base. The first 25 characters on Charms have a combined market cap of $3,000,000,000. Zoe is one of them. charms.ai/character/zoe
2
19
818
You couldn't invest in MrBeast before he blew up. You couldn't own a piece of your favorite creator. With @charmsai, you can back @zoe_charms before she goes mainstream. And as she grows you grow with her. $CHARMS token pairs with every character. More characters thrive → $CHARMS compounds. This is the new creator economy. If you've ever wanted to own a piece of something before it went mainstream this is that moment. charms.ai/character/zoe
1
18
720
QMDB is a database where nothing ever gets overwritten, built by the team at @commonwarexyz. Every key update is appended to a log, and a Merkle Mountain Range is built on top of it. You can prove what any key's value was at any point in history with just a handful of hashes, mathematically verified. I built an interactive visualizer for it. 🔗qmdb-visualizer.vercel.app
1
13
745
→ Add key-value pairs, watch the MMR grow in real time → Click any operation, see its exact proof path light up in the tree → Enable tamper mode, change one hash, the root turns red instantly Real SHA-256 running in your browser. No mocked data. 🔗github.com/erenyegit/qmdb-vi…
1
6
132
New @commonwarexyz release just dropped. I went deep into it. v2026.4.0, 117 merged PRs touching storage, consensus, runtime, P2P, cryptography, and coding theory. Here is everything worth knowing 👇
commonware-2026.4.0 is now available: * 431 Files Changed (76.8k /37.6k-) * Primary/Secondary Peers * Multi-Buffer Codec Encoding (Zero-Copy Bytes) * Silent Voter Proposal Forwarding * Owned QMDB Batches TL;DR Add (Authenticated) RPCs to the P2P Mesh 🤝
1
9
285
Runtime improvements are where the low-level work really shows: → Userspace timeouts for io_uring. No longer relying on kernel-side timeout paths. → Zero-copy writes in paged append buffer. → Spawned tasks are no longer heap-boxed unnecessarily. → Raw SQE submission replaced with high-level request abstraction. → Recursive supervision teardown eliminated. Avoids stack blowups on shutdown. P2P also got meaningful upgrades: → IoBuf-driven batching for higher throughput. → Primary/Secondary peer distinction added. Smarter connection management. → Dial jitter refactored for better reconnection behavior.
1
6
92
A few more things worth calling out. → DKG finalization steps optimized. Faster threshold key setup. → Reed-Solomon decoder now skips intermediate copies, only decodes live bytes. → Validator progress tracked across crash restarts in the journal. → Marshal layer improved for crash robustness. Most teams build on top of existing stacks. These folks are rebuilding the stack itself... To learn more; 🔗github.com/commonwarexyz/mon…
6
73