S16 E36: Maud Bannwart & Cheng Wang on Alephium, Proof of Less Work & Bitcoin
The audio keeps playing while you browse the site.
Read the transcript16,507 words, every line links to the minute it was said
Ask this episode
We recorded this one in Frankfurt, Germany, last year, with Cheng Wang, the founder and core developer of Alephium, and Maud Bannwart, its COO, sitting across from me. Alephium is a sharded proof-of-work chain that mines with BLAKE3, keeps assets in a UTXO model and application state in an account model, and ships its own smart contract language called Ralph. The Danube upgrade, which cut block time to 8 seconds and brought groupless addresses and passkey support, was fresh enough that I managed to get its name wrong on Twitter before the interview, and I will come back to that.
My starting position was skeptical, and I said so. I wanted to know why a new chain should exist next to Bitcoin, whether a token the team invented counts as money, why sharding would not simply recentralize everything, and why Alephium never got Solana-sized adoption. Cheng and Maud took every one of those in turn, and the answers mostly came from implementation details I had not read.
A Problem Others Walked Away From
Cheng's origin story starts in academia. Before Alephium, Cheng said, other researchers had looked at sharding a proof-of-work chain and "invested a lot of resources and research into the problem and did not find a good solution." That was the invitation. "I'm very obsessed with challenges," Cheng told me.
The solution Cheng landed on is BlockFlow. Each shard is its own chain, but the chains reference one another, so the structure is a DAG. To pick the winning history, "you calculate the accumulated work from different blockchains, you aggregate them, and then you compare which DAG has the most amount of work." Hence the name: "it's a flow, it's not a blockchain."
I asked the obvious worry. If you split the chain, do you split the security? Cheng's answer leaned on proof of work being the easier consensus to shard. With proof of stake you have to distribute validators across shards and aggregate signatures. With work, Cheng said, "Our sharding algorithm is much, much simpler than the one we originally wanted to do, because we use proof of work." Cheng added that "for proof of work, actually, it favors liveness over safety," which is the tradeoff Cheng accepts.
Cheng also gave me the Bitcoin comparison that made the throughput argument click. "If you reduce the block time of Bitcoin to five minutes per block, then you basically scale the throughput by two times. And the security won't be impacted." On Bitcoin itself, though, Cheng is not asking for that change: "I personally think that one megabyte and also 10 minute block time is great for Bitcoin." Alephium's job, as Cheng sees it, is elsewhere: "We have to focus more on the utility, on the application side."
UTXOs for Assets, Accounts for State
The data model was my favorite technical thread of the night. Alephium uses UTXOs for assets, and Cheng's top reason is that "the biggest one is the immutability." The payoff shows up in parallel processing: "You won't have race conditions if you have multiple processes to try to use the same UTXO because it's immutable." Contract state is a different animal. "For that, we use the account model," Cheng said, after confirming "So it's like Bitcoin" for the asset side.
I asked whether UTXOs buy you privacy as well. Cheng's reply was that the primitives are there: "theoretically, you can build either mixnets or Tornado Cash." Nothing is native, but the model does not get in the way. I told Cheng it still did not compel me, and Cheng went straight back to throughput, saying that "the throughput of the blockchain is very high," so the design leaves room for that kind of application.
I Said Taproot, and I Was Wrong
Here is where I got corrected. I had been describing Alephium's contract and wallet features as riding on Taproot, and I said as much ("It takes Taproot"). Cheng set me straight: "in Alephium we don't really have Taproot." Cheng was good-humored about it and about the memory lapse too, admitting that "after one or a few months I clearly forgot about the underlying details" of some earlier design work. On Taproot itself Cheng's view is that "Taproot is a hack solution for Bitcoin specifically," something you do when you cannot change the base layer cleanly.
The Danube naming was my other blunder. I had run a Twitter poll with the wrong name for the upgrade, and Maud brought it up with a grin: "You actually named our latest upgrade." The team had actually considered calling it the Mediterranean upgrade, Maud told me, and dropped the idea because nobody could pronounce it. Danube it is. What it does matters more than what it was called: block time down to 8 seconds, groupless addresses, and passkey support.
Groupless addresses are about hiding the sharding from the user. Cheng described the goal as having "basically made the sharded blockchain the same user experience as a single blockchain." If I am holding a passkey wallet, I should not need to know which shard my coins live on. For a project whose whole architecture is about chains plural, that is the design target I would grade them on.
Why Not the EVM?
I asked why Alephium is not EVM compatible, the question every smart contract chain gets from its first skeptic. Maud admitted there is a real cost, and it is mainly commercial: "Where it's a bit more complicated to have made the choice of not being EVM compatible is integration with services." Wallets, exchanges and indexers all expect Ethereum tooling. On the developer side, Maud said, "we've never had anyone who tried the tech stack of Alephium as a developer and said, it's too complicated." Maud's view of learning a new language was relaxed: "as a software developer, knowing several language, it's part of the life of a software developers. Like, I've never met a software developers that only knew one language, one framework, and that is it."
Cheng was blunter about Ethereum's virtual machine. "I think EVM is just a first mover advantage," Cheng said, and then: "I personally feel like EVM don't really have a good future." Cheng pushed it further, saying that the baggage "slows down the development and innovation, and it's not great for the long term. And that's why EVM loses in some of the games in the DeFi market." When I looked skeptical, Cheng offered a wager of sorts: "Maybe we can have another interview in five years and to see if this has become true or not."
Ralph, the language, is Cheng's answer. Cheng described it as "Rust, I would say like a more simple version of Rust," and said of Solidity that it "is definitely very easy to learn language, but it's very hard to master. It's a bit like JavaScript. JavaScript is so easy to write bugs." How fast can a newcomer pick up Ralph? "Many people say that they can get started to write something with just one to two days."
I later recorded S16 E37 with Andrew Poelstra on Simplicity and Bitcoin smart contracts.
Proof of Less Work
The mechanism I pushed hardest on was Proof of Less Work. The pitch is that once Alephium's hashrate passes one exahash, the protocol has miners burn ALPH coins in place of electricity, and the project projects an 87% drop in energy use. My first read was cynical. I asked whether that was just a reduction in mining rewards. Maud's answer was a flat no: "It's not that you just reduce the mining reward." Maud also took pains to separate it from proof of stake: "this is not a combination with proof of stake."
What it does mechanically, Maud said, is that "at similar hash rates it does less work than proof of work." Cheng added that the idea was planned "from day one before Genesis," and when I asked for the exact numbers behind the savings Cheng admitted, "I don't remember exact numbers." On whether Bitcoin could adopt it: "For Bitcoin it's very hard to change, so I don't see any time that it will be applied for Bitcoin."
Maud framed the energy angle two ways. Miners are sometimes praised as "one of the most cost efficient way to finance sustainable transition," Maud said, but Maud still thinks "there should be a cost to creating money." That last line lines up with how I think about proof of work, though I keep my reservations about burning a coin instead of a kilowatt-hour.
The mining side of the project has its own story. Alephium started GPU-mined and moved to ASICs, and Maud's view of ASIC resistance is that "First you can never really be ASIC proof, you're ASIC resistant but never proof so to maintain that ASIC resistance this is something that is costly and this is something you have to maintain through time." Alephium's ASICs are chain-specific: "which means that the Alephium ASICs cannot be used to mine anything other than Alephium." Maud was realistic about GPUs now: "If your electricity is free, if your electricity is free and you really have a lot of GPU, why not? But otherwise, probably not."
The live chat chewed on hashrate. Krev23 questioned whether mining was profitable and compared non-mineable chains to Visa, and someone else noted that hashrate had surged 42,000% from January 1, 2025 to its all-time high after ASICs arrived. The figures I have from the team: hashrate fluctuates between 20 and 25 petahashes, with a peak of 57. I mentioned the opportunistic miner who reportedly earned $5,000 with ASIC rigs in a hotel room, which is exactly the kind of story that makes me want to buy a hotel.
Fair Launch, Genesis Allocations and Who Pays the Developers
I put the fair-launch question to them directly, since I think it is where Alephium is easiest to attack. Cheng's reply was that "I don't feel like fair launch is really a big issue." Maud added the factual claim: "There was no pre-mine." Genesis allocations exist, but they fund development and keep the project independent. Maud's reasoning on funding was that without it, "it means that third party will decide what you build, when you build it and how you build it." Cheng pointed at the starting point of the chain, where "the genesis seed is based on the Bitcoin block hash and the Ethereum block hash." Maud also mentioned, regarding the early period, "for a few months, actually," which I took as the team's answer to how long the allocations were being used before they stabilized.
I argued that a token invented by a team is not money. They did not budge on the facts, and I did not budge on the principle. Cheng did say the team "did not try to raise big or sell a lot," and Maud's price policy was simple: "As a general rule, we don't talk about price." I pointed out that meme coins like dogwifhat and HarryPotterObamaSonic10Inu (ticker BITCOIN) are all over SideShift, and wondered aloud why Alephium cannot get the same love. "They have the dog with the hat," I said. "Why would they not have Alephium?"
On adoption, Maud said "the only dApp that we built ourselves as of today is the bridge," while "around 70, if not more now" third-party projects are building on the chain. I later recorded episodes with Danny Plainview and Chad Master on OP_NET and Alex Chepurnoy on Ergo.
Hacks in the Codebase and Storage at 2GB a Day
Near the end the conversation turned to the unglamorous things. Cheng is not precious about the code: "There are a lot of hacks in the codebase," Cheng said, and Maud mentioned "my ever-growing to-do list." On monetary policy Cheng said the project wants "to move to probably a tail emission model at some point," and Cheng does not buy the fee-market-only alternative: "I don't think it's going to work in the long term."
I raised the cost of running a node, since sharded chains tend to be heavy. Node storage grows by roughly 2GB per day on Alephium, which is not a small number. Maud's framing of why that matters was firm: "nodes are the network." They also run on top of Edge Wallet and similar services on the user side, and the wider ecosystem includes bridges and projects like Citrea in the Bitcoin-adjacent space, as well as P2P marketplaces like Noones. The Alephium docs are where I would point a developer curious enough to try Ralph.
One footnote I enjoyed: Cheng and a core developer of Aleph Zero did their PhD research at the same university, which made the naming overlap (Aleph is the unit for infinity, Cheng told me, chosen to represent infinite potential) a little less coincidental than it sounded when I first brought it up.
I asked Maud, on the way out, how much Maud tweets for the project. Maud's answer was short and it is where I will stop: "I have zero personality on Twitter. There's no need."
Listen to Maud Bannwart & Cheng Wang on Spotify, Apple Podcasts, YouTube, and More
Time stamps:
- 00:00:53– Introducing Cheng & Maud
- 00:02:06– Alephium's Origins & Motivation
- 00:04:36– Choosing UTXO vs. Account Model
- 00:09:50– Privacy and UTXO Model
- 00:11:03– Blockchain Scalability & Parallelization
- 00:14:41– Decentralization vs. Shard Coordination
- 00:17:31– DAG Structure Explanation
- 00:19:44– Shard Synchronization & Security
- 00:21:09– Block Time, Decentralization & Node Requirements
- 00:25:22– Blockchain Pruning & Storage
- 00:27:13– ASICs and Mining History
- 00:29:33– Taproot & Smart Contracting
- 00:33:05– Proof of Work vs. Proof of Stake
- 00:37:45– Technical Advantages of Proof of Work
- 00:39:56– DeFi on Proof of Work
- 00:42:09– Value of Proof of Work & Community
- 00:44:56– Pre-mine & Project Funding
- 00:50:41– Transparency & Genesis Block
- 00:54:11– Proof of Less Work Mechanism
- 01:00:16– Environmental Criticism & Value Proposition
- 01:04:01– Alephium’s Unique Value & Comparison
- 01:07:39– Developer Experience & EVM Compatibility
- 01:14:11– Alephium Virtual Machine & Language
- 01:19:04– Ecosystem & DApps
- 01:22:16– Killer App & Platform Potential
- 01:23:37– Danube Hard Fork & Optimizations
- 01:26:25– Mining Accessibility & Hardware
- 01:29:10– Hashrate Growth & NiceHash
- 01:31:56– Price & Market Position
- 01:35:19– Lessons for Bitcoin & Future-Proofing
- 01:38:56– Tokenized Bitcoin & UTXO Benefits
- 01:43:25– Privacy, Rollups, and Future Features
- 01:51:11– Getting Started with Alephium
- 01:53:08– Upgrade Naming & Community
- 01:53:43– Final Thoughts & Contact
Verify this episode
This is the SHA-256 fingerprint of the MP3 as published here, timestamped in the Bitcoin blockchain with OpenTimestamps. If one byte of the file changes, the fingerprint changes.
- SHA-256
18a6e8ae5e87811d91b0736bc6f32e02df2daa75657cf87802d3a82b93b1e2f0- Proof
- 2025-08-02_s16_e36__maud_bannwart_cheng_wang_on_alephium_proof_of_less_work_bitcoin.mp3.ots · all episode hashes (proof)
- Status
- Anchored in Bitcoin block 974,312
- Timestamped
- After release. The episode came out on 2 Aug 2025, but this proof was made on 3 Nov 2026 and anchored in block 974,312, so it proves this exact file existed by 3 Nov 2026, not on the release date.
How to check it yourself
- Download the MP3 and the proof into the same folder.
- Run
sha256sum 2025-08-02_s16_e36__maud_bannwart_cheng_wang_on_alephium_proof_of_less_work_bitcoin.mp3(on a Mac:shasum -a 256 2025-08-02_s16_e36__maud_bannwart_cheng_wang_on_alephium_proof_of_less_work_bitcoin.mp3). The result must match the SHA-256 above. - Install the OpenTimestamps client (
pip install opentimestamps-client) and runots verify 2025-08-02_s16_e36__maud_bannwart_cheng_wang_on_alephium_proof_of_less_work_bitcoin.mp3.ots. It hashes the MP3 again and checks the proof against your own Bitcoin node, then prints the block and the date the file existed by. Without a node,ots info 2025-08-02_s16_e36__maud_bannwart_cheng_wang_on_alephium_proof_of_less_work_bitcoin.mp3.otsshows the block height to look up in any block explorer.
Related
S1 E4 Remastered: Omar Faridi on Bitcoin in Pakistan, Brain Drain & the Case for Moderate Thinking
In January 2019, I was still an editor at Crypto Insider. Working a fixed salary and editing other people’s articles while…
S17 E15: Danny Plainview & Chad Master on OP_NET, Smart Contracts on Bitcoin L1 & the Great Stablecoin Unlock
Editor’s note: this conversation was recorded in late March 2026, in the days surrounding OP_NET’s mainnet activation. The…
S16 E45: Shai Wyborski on Bitcoin, Kaspa & Proof of Work Research
This turned out to become the longest episode in the history of the Bitcoin Takeover podcast… yet: 8 hours and 39 minutes,…
Comments
Comments run on my own server. No trackers, no tracking cookies.